velo-plot 3.1.0 → 4.0.0-alpha.1

package/dist/plugins/forecasting.js.map

@@ -1 +1 @@
-{"version":3,"file":"forecasting.js","sources":["../../src/plugins/forecasting/algorithms.ts","../../src/plugins/forecasting/index.ts"],"sourcesContent":["/**\n * Forecasting Algorithms\n *\n * Native implementations of time series forecasting methods.\n *\n * ---------------------------------------------------------------------------\n * Method status audit (task 3.1)\n * ---------------------------------------------------------------------------\n * | Method        | Status       | Notes                                     |\n * |---------------|--------------|-------------------------------------------|\n * | sma           | implemented  | Flat projection at trailing average       |\n * | wma           | implemented  | Flat projection at weighted trailing avg  |\n * | ema           | implemented  | Flat projection at last EMA value         |\n * | expSmoothing  | implemented  | Simple exponential smoothing (SES)        |\n * | holt          | implemented  | Double exponential smoothing (trend)      |\n * | holtWinters   | implemented  | Triple exponential smoothing (seasonal)   |\n * | linear        | implemented  | OLS linear trend projection               |\n * | arima         | implemented  | ARIMA(p,d,q) via Hannan-Rissanen (simple) |\n *\n * Every public method returns a finite forecast plus 95% (configurable)\n * confidence bands derived from in-sample one-step residuals. No public\n * method throws for a supported `ForecastingMethod`.\n */\n\nimport type { ForecastingMethod, ForecastingParams, ForecastingResult } from './types';\n\n/** Internal result of a single method: forecast plus in-sample fitted values. */\ninterface MethodResult {\n  /** Point forecast for the requested horizon. */\n  yValues: number[];\n  /**\n   * In-sample one-step-ahead fitted values aligned with the history `y`.\n   * `fitted[i]` is the prediction of `y[i]` using only `y[0..i-1]`.\n   * `null` entries are ignored when computing residuals.\n   */\n  fitted?: (number | null)[];\n  /** Whether forecast uncertainty accumulates with the horizon (sqrt growth). */\n  accumulates: boolean;\n}\n\nfunction toArray(a: number[] | Float64Array | Float32Array): number[] {\n  return Array.from(a);\n}\n\n/**\n * Map a confidence level (e.g. 0.95) to a normal-distribution z multiplier.\n */\nfunction zForConfidence(confidence: number): number {\n  const table: Array<[number, number]> = [\n    [0.5, 0.674],\n    [0.68, 0.994],\n    [0.8, 1.282],\n    [0.9, 1.645],\n    [0.95, 1.96],\n    [0.975, 2.241],\n    [0.99, 2.576],\n    [0.995, 2.807],\n  ];\n  let best = table[0];\n  let bestDiff = Infinity;\n  for (const entry of table) {\n    const diff = Math.abs(entry[0] - confidence);\n    if (diff < bestDiff) {\n      bestDiff = diff;\n      best = entry;\n    }\n  }\n  return best[1];\n}\n\n/**\n * Main dispatcher for forecasting algorithms.\n */\nexport function calculateForecast(\n  x: number[] | Float64Array | Float32Array,\n  y: number[] | Float64Array | Float32Array,\n  method: ForecastingMethod,\n  horizon: number,\n  params: ForecastingParams = {},\n  confidence = 0.95\n): ForecastingResult {\n  const n = x.length;\n  if (n === 0) throw new Error('Dataset is empty');\n\n  const ya = toArray(y);\n  const xa = toArray(x);\n\n  // Determine time step (assuming regular intervals)\n  const dt = n > 1 ? (xa[n - 1] - xa[0]) / (n - 1) : 1;\n  const lastX = xa[n - 1];\n\n  let result: MethodResult;\n\n  switch (method) {\n    case 'sma':\n      result = forecastSMA(ya, horizon, params.windowSize || 10);\n      break;\n    case 'ema':\n      result = forecastEMA(ya, horizon, params.alpha ?? 0.3);\n      break;\n    case 'linear':\n      result = forecastLinear(xa, ya, horizon);\n      break;\n    case 'expSmoothing':\n      result = forecastSimpleExpSmoothing(ya, horizon, params.alpha ?? 0.3);\n      break;\n    case 'holt':\n      result = forecastHolt(ya, horizon, params.alpha ?? 0.3, params.beta ?? 0.1);\n      break;\n    case 'wma':\n      result = forecastWMA(ya, horizon, params.windowSize || 10);\n      break;\n    case 'holtWinters':\n      result = forecastHoltWinters(\n        ya,\n        horizon,\n        params.period || 12,\n        params.alpha ?? 0.3,\n        params.beta ?? 0.1,\n        params.gamma ?? 0.1\n      );\n      break;\n    case 'arima':\n      result = forecastARIMA(\n        ya,\n        horizon,\n        params.p ?? 1,\n        params.d ?? 1,\n        params.q ?? 0\n      );\n      break;\n    default:\n      throw new Error(`Method ${method} not implemented yet`);\n  }\n\n  // Generate future X values\n  const futureX = new Array(horizon);\n  for (let i = 0; i < horizon; i++) {\n    futureX[i] = lastX + (i + 1) * dt;\n  }\n\n  // Compute residual-based statistics and confidence bands.\n  const metrics = calculateFitMetrics(ya, result.fitted);\n  const z = zForConfidence(confidence);\n  const sigma = metrics.rmse;\n  const { lower, upper } = buildConfidenceBands(\n    result.yValues,\n    sigma,\n    z,\n    result.accumulates\n  );\n\n  return {\n    xValues: futureX,\n    yValues: result.yValues,\n    lowerBound: lower,\n    upperBound: upper,\n    method,\n    metadata: { ...metrics, confidence },\n  };\n}\n\n/**\n * Build symmetric confidence bands around the point forecast.\n * When `accumulates` is true the uncertainty widens with sqrt(horizon step),\n * reflecting the compounding error of trend/AR models.\n */\nfunction buildConfidenceBands(\n  yValues: number[],\n  sigma: number,\n  z: number,\n  accumulates: boolean\n): { lower: number[]; upper: number[] } {\n  const lower = new Array(yValues.length);\n  const upper = new Array(yValues.length);\n  for (let i = 0; i < yValues.length; i++) {\n    const growth = accumulates ? Math.sqrt(i + 1) : 1;\n    const margin = z * sigma * growth;\n    lower[i] = yValues[i] - margin;\n    upper[i] = yValues[i] + margin;\n  }\n  return { lower, upper };\n}\n\n// ===========================================================================\n// Moving-average family\n// ===========================================================================\n\n/**\n * Weighted Moving Average (WMA).\n * Uses linear weights on the trailing window; projects flat at last WMA value.\n */\nfunction forecastWMA(y: number[], horizon: number, window: number): MethodResult {\n  const n = y.length;\n  const effectiveWindow = Math.min(window, n);\n\n  const wmaAt = (end: number): number => {\n    // end is always >= 1 here, so the trailing window w is always >= 1.\n    const w = Math.min(effectiveWindow, end);\n    let weightedSum = 0;\n    let weightTotal = 0;\n    for (let i = 0; i < w; i++) {\n      const weight = i + 1;\n      weightedSum += y[end - w + i] * weight;\n      weightTotal += weight;\n    }\n    return weightedSum / weightTotal;\n  };\n\n  const fitted: (number | null)[] = new Array(n).fill(null);\n  for (let i = 1; i < n; i++) fitted[i] = wmaAt(i);\n\n  const wma = wmaAt(n);\n  return { yValues: new Array(horizon).fill(wma), fitted, accumulates: false };\n}\n\n/**\n * Simple Moving Average (SMA).\n * Historical SMA used for projection (flat line at last average).\n */\nfunction forecastSMA(y: number[], horizon: number, window: number): MethodResult {\n  const n = y.length;\n  const effectiveWindow = Math.min(window, n);\n\n  const smaAt = (end: number): number => {\n    // end is always >= 1 here, so the trailing window w is always >= 1.\n    const w = Math.min(effectiveWindow, end);\n    let sum = 0;\n    for (let i = end - w; i < end; i++) sum += y[i];\n    return sum / w;\n  };\n\n  const fitted: (number | null)[] = new Array(n).fill(null);\n  for (let i = 1; i < n; i++) fitted[i] = smaAt(i);\n\n  const avg = smaAt(n);\n  return { yValues: new Array(horizon).fill(avg), fitted, accumulates: false };\n}\n\n/**\n * Exponential Moving Average (EMA).\n */\nfunction forecastEMA(y: number[], horizon: number, alpha: number): MethodResult {\n  const n = y.length;\n  let ema = y[0];\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  for (let i = 1; i < n; i++) {\n    fitted[i] = ema; // one-step forecast is the previous EMA\n    ema = alpha * y[i] + (1 - alpha) * ema;\n  }\n\n  return { yValues: new Array(horizon).fill(ema), fitted, accumulates: false };\n}\n\n/**\n * Linear Trend Projection.\n * Fits y = mx + b to the historical data.\n */\nfunction forecastLinear(x: number[], y: number[], horizon: number): MethodResult {\n  const n = x.length;\n  let sumX = 0,\n    sumY = 0,\n    sumXY = 0,\n    sumXX = 0;\n\n  for (let i = 0; i < n; i++) {\n    sumX += x[i];\n    sumY += y[i];\n    sumXY += x[i] * y[i];\n    sumXX += x[i] * x[i];\n  }\n\n  const denom = n * sumXX - sumX * sumX;\n  const slope = denom !== 0 ? (n * sumXY - sumX * sumY) / denom : 0;\n  const intercept = (sumY - slope * sumX) / n;\n\n  const fitted: (number | null)[] = new Array(n);\n  for (let i = 0; i < n; i++) fitted[i] = slope * x[i] + intercept;\n\n  const lastX = x[n - 1];\n  const dt = n > 1 ? (x[n - 1] - x[0]) / (n - 1) : 1;\n  const yValues = new Array(horizon);\n  for (let i = 0; i < horizon; i++) {\n    const nextX = lastX + (i + 1) * dt;\n    yValues[i] = slope * nextX + intercept;\n  }\n\n  return { yValues, fitted, accumulates: true };\n}\n\n/**\n * Simple Exponential Smoothing (SES).\n */\nfunction forecastSimpleExpSmoothing(y: number[], horizon: number, alpha: number): MethodResult {\n  const n = y.length;\n  let level = y[0];\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  for (let i = 1; i < n; i++) {\n    fitted[i] = level;\n    level = alpha * y[i] + (1 - alpha) * level;\n  }\n\n  return { yValues: new Array(horizon).fill(level), fitted, accumulates: false };\n}\n\n/**\n * Holt's Linear Trend (Double Exponential Smoothing).\n */\nfunction forecastHolt(y: number[], horizon: number, alpha: number, beta: number): MethodResult {\n  const n = y.length;\n  if (n < 2) {\n    return { yValues: new Array(horizon).fill(y[0] || 0), fitted: [], accumulates: true };\n  }\n\n  let level = y[0];\n  let trend = y[1] - y[0];\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  for (let i = 1; i < n; i++) {\n    fitted[i] = level + trend; // one-step forecast\n    const lastLevel = level;\n    level = alpha * y[i] + (1 - alpha) * (level + trend);\n    trend = beta * (level - lastLevel) + (1 - beta) * trend;\n  }\n\n  const yValues = new Array(horizon);\n  for (let i = 0; i < horizon; i++) yValues[i] = level + (i + 1) * trend;\n\n  return { yValues, fitted, accumulates: true };\n}\n\n/**\n * Holt-Winters (Triple Exponential Smoothing, additive seasonality).\n */\nfunction forecastHoltWinters(\n  y: number[],\n  horizon: number,\n  period: number,\n  alpha: number,\n  beta: number,\n  gamma: number\n): MethodResult {\n  const n = y.length;\n  if (n < period * 2) {\n    // Fallback to Holt if not enough data for seasonality\n    return forecastHolt(y, horizon, alpha, beta);\n  }\n\n  // 1. Initial Seasonality\n  const seasonals = new Array(period);\n  const seasonAverages = new Array(Math.floor(n / period));\n  for (let i = 0; i < seasonAverages.length; i++) {\n    let sum = 0;\n    for (let j = 0; j < period; j++) sum += y[i * period + j];\n    seasonAverages[i] = sum / period;\n  }\n\n  for (let i = 0; i < period; i++) {\n    let sumOverAvg = 0;\n    for (let j = 0; j < seasonAverages.length; j++) {\n      sumOverAvg += y[j * period + i] - seasonAverages[j];\n    }\n    seasonals[i] = sumOverAvg / seasonAverages.length;\n  }\n\n  // 2. Initial Level and Trend\n  let level = seasonAverages[0];\n  let trend =\n    (seasonAverages[seasonAverages.length - 1] - seasonAverages[0]) /\n    ((seasonAverages.length - 1) * period);\n\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  // 3. Update coefficients\n  for (let i = 0; i < n; i++) {\n    const seasonal = seasonals[i % period];\n    if (i > 0) fitted[i] = level + trend + seasonal;\n    const lastLevel = level;\n    const val = y[i];\n    level = alpha * (val - seasonal) + (1 - alpha) * (level + trend);\n    trend = beta * (level - lastLevel) + (1 - beta) * trend;\n    seasonals[i % period] = gamma * (val - level) + (1 - gamma) * seasonal;\n  }\n\n  // 4. Forecast\n  const yValues = new Array(horizon);\n  for (let i = 0; i < horizon; i++) {\n    const m = i + 1;\n    yValues[i] = level + m * trend + seasonals[(n + i) % period];\n  }\n\n  return { yValues, fitted, accumulates: true };\n}\n\n// ===========================================================================\n// ARIMA(p, d, q)\n// ===========================================================================\n\n/**\n * Simple native ARIMA implementation.\n *\n * - Differences the series `d` times to achieve (approximate) stationarity.\n * - Estimates AR(p) and MA(q) coefficients using the two-stage\n *   Hannan-Rissanen procedure (long AR to recover innovations, then OLS on\n *   lagged values and lagged innovations).\n * - Forecasts recursively on the differenced scale, then integrates back.\n *\n * Falls back to Holt's linear trend when there is not enough data to fit the\n * requested orders, so the method never throws for valid input.\n */\nfunction forecastARIMA(\n  y: number[],\n  horizon: number,\n  p: number,\n  d: number,\n  q: number\n): MethodResult {\n  const n = y.length;\n  const order = Math.max(p, q);\n  // Need enough observations to difference and fit; otherwise degrade gracefully.\n  if (n < d + order + 2 || n < 4) {\n    return forecastHolt(y, horizon, 0.3, 0.1);\n  }\n\n  // 1. Difference d times, keeping every intermediate level for integration.\n  const diffLevels: number[][] = [y.slice()];\n  for (let k = 0; k < d; k++) {\n    diffLevels.push(difference(diffLevels[k]));\n  }\n  const w = diffLevels[d]; // fully differenced (stationary) series\n  const m = w.length;\n\n  // Center the differenced series around its mean.\n  const mean = w.reduce((s, v) => s + v, 0) / m;\n  const wc = w.map((v) => v - mean);\n\n  // 2. Stage 1 - recover innovation (residual) estimates via a long AR fit.\n  let innovations = new Array(m).fill(0);\n  if (q > 0) {\n    const arLong = Math.min(Math.max(p + q + 2, 5), Math.floor(m / 2));\n    const { residuals } = fitAR(wc, arLong);\n    innovations = residuals;\n  }\n\n  // 3. Stage 2 - regress w_t on its p lags and q lagged innovations.\n  const start = Math.max(p, q);\n  const rows: number[][] = [];\n  const targets: number[] = [];\n  for (let t = start; t < m; t++) {\n    const row: number[] = [];\n    for (let i = 1; i <= p; i++) row.push(wc[t - i]);\n    for (let j = 1; j <= q; j++) row.push(innovations[t - j]);\n    rows.push(row);\n    targets.push(wc[t]);\n  }\n\n  let coeffs: number[] = [];\n  if (rows.length > 0 && rows[0].length > 0) {\n    coeffs = olsSolve(rows, targets);\n  }\n  if (coeffs.length !== p + q || coeffs.some((c) => !Number.isFinite(c))) {\n    // Fitting failed (singular/ill-conditioned) - fall back.\n    return forecastHolt(y, horizon, 0.3, 0.1);\n  }\n  const ar = coeffs.slice(0, p);\n  const ma = coeffs.slice(p, p + q);\n\n  // 4. In-sample one-step fitted values (on original scale) for residual sigma.\n  const fitted: (number | null)[] = new Array(n).fill(null);\n  const fittedResiduals = new Array(m).fill(0);\n  for (let t = start; t < m; t++) {\n    let pred = 0;\n    for (let i = 1; i <= p; i++) pred += ar[i - 1] * wc[t - i];\n    for (let j = 1; j <= q; j++) pred += ma[j - 1] * fittedResiduals[t - j];\n    fittedResiduals[t] = wc[t] - pred;\n    // Map differenced fitted value back to the original scale by adding the\n    // previous original observation chain. For d>=1 we approximate by adding\n    // the last non-differenced value; exact integration below handles forecasts.\n    fitted[t + (n - m)] = integrateFittedPoint(diffLevels, d, t, pred + mean);\n  }\n\n  // 5. Recursive forecast on the differenced/centered scale.\n  const wHist = wc.slice();\n  const eHist = fittedResiduals.slice();\n  const forecastDiff: number[] = [];\n  for (let h = 0; h < horizon; h++) {\n    let pred = 0;\n    // wHist/eHist always have at least `order` entries here, so the lag index\n    // is always in range.\n    for (let i = 1; i <= p; i++) {\n      pred += ar[i - 1] * wHist[wHist.length - i];\n    }\n    for (let j = 1; j <= q; j++) {\n      pred += ma[j - 1] * eHist[eHist.length - j];\n    }\n    wHist.push(pred);\n    eHist.push(0); // expected future innovation is zero\n    forecastDiff.push(pred + mean);\n  }\n\n  // 6. Integrate the forecast back d times to the original scale.\n  const yValues = integrateForecast(diffLevels, d, forecastDiff);\n\n  return { yValues, fitted, accumulates: true };\n}\n\n/** First difference of a series. */\nfunction difference(series: number[]): number[] {\n  const out = new Array(series.length - 1);\n  for (let i = 1; i < series.length; i++) out[i - 1] = series[i] - series[i - 1];\n  return out;\n}\n\n/**\n * Integrate a differenced forecast back to the original scale.\n * `diffLevels[0]` is the original series, `diffLevels[d]` fully differenced.\n */\nfunction integrateForecast(diffLevels: number[][], d: number, forecastDiff: number[]): number[] {\n  let current = forecastDiff.slice();\n  for (let k = d; k >= 1; k--) {\n    const parent = diffLevels[k - 1];\n    let last = parent[parent.length - 1];\n    const integrated = new Array(current.length);\n    for (let i = 0; i < current.length; i++) {\n      last += current[i];\n      integrated[i] = last;\n    }\n    current = integrated;\n  }\n  return current;\n}\n\n/**\n * Best-effort mapping of a single in-sample differenced prediction back to the\n * original scale (used only for residual-sigma estimation).\n */\nfunction integrateFittedPoint(\n  diffLevels: number[][],\n  d: number,\n  t: number,\n  diffPred: number\n): number {\n  if (d === 0) return diffPred;\n  // Add the previous observed original value chain.\n  const parent = diffLevels[d - 1];\n  // parent is one longer than the differenced series, so index t is in range.\n  const base = parent[t];\n  // For d>1 this is an approximation; sufficient for a residual scale estimate.\n  return base + diffPred;\n}\n\n/**\n * Fit an AR(order) model via ordinary least squares and return coefficients\n * plus the in-sample residuals aligned to the input series (leading entries 0).\n */\nfunction fitAR(series: number[], order: number): { coeffs: number[]; residuals: number[] } {\n  const m = series.length;\n  const rows: number[][] = [];\n  const targets: number[] = [];\n  for (let t = order; t < m; t++) {\n    const row: number[] = [];\n    for (let i = 1; i <= order; i++) row.push(series[t - i]);\n    rows.push(row);\n    targets.push(series[t]);\n  }\n  const coeffs = rows.length > 0 ? olsSolve(rows, targets) : new Array(order).fill(0);\n  const residuals = new Array(m).fill(0);\n  for (let t = order; t < m; t++) {\n    let pred = 0;\n    for (let i = 1; i <= order; i++) pred += (coeffs[i - 1] || 0) * series[t - i];\n    residuals[t] = series[t] - pred;\n  }\n  return { coeffs, residuals };\n}\n\n/**\n * Solve a linear least-squares problem `X b = y` via the normal equations\n * with Gaussian elimination. Returns [] on a singular system.\n */\nfunction olsSolve(X: number[][], y: number[]): number[] {\n  const rows = X.length;\n  const cols = X[0].length;\n\n  // Normal equations: (XᵀX) b = Xᵀy\n  const xtx: number[][] = Array.from({ length: cols }, () => new Array(cols).fill(0));\n  const xty: number[] = new Array(cols).fill(0);\n\n  for (let i = 0; i < rows; i++) {\n    for (let a = 0; a < cols; a++) {\n      xty[a] += X[i][a] * y[i];\n      for (let b = 0; b < cols; b++) {\n        xtx[a][b] += X[i][a] * X[i][b];\n      }\n    }\n  }\n\n  // Tiny ridge term to stabilise near-singular systems.\n  for (let a = 0; a < cols; a++) xtx[a][a] += 1e-8;\n\n  return gaussianSolve(xtx, xty);\n}\n\n/**\n * Solve `A x = b` for a square matrix via Gaussian elimination with partial\n * pivoting. Returns [] if the matrix is singular.\n */\nexport function gaussianSolve(A: number[][], b: number[]): number[] {\n  const n = b.length;\n  const M = A.map((row, i) => [...row, b[i]]);\n\n  for (let col = 0; col < n; col++) {\n    // Partial pivot\n    let pivot = col;\n    for (let r = col + 1; r < n; r++) {\n      if (Math.abs(M[r][col]) > Math.abs(M[pivot][col])) pivot = r;\n    }\n    if (Math.abs(M[pivot][col]) < 1e-12) return [];\n    [M[col], M[pivot]] = [M[pivot], M[col]];\n\n    // Eliminate\n    for (let r = 0; r < n; r++) {\n      if (r === col) continue;\n      const factor = M[r][col] / M[col][col];\n      for (let c = col; c <= n; c++) M[r][c] -= factor * M[col][c];\n    }\n  }\n\n  const x = new Array(n);\n  for (let i = 0; i < n; i++) x[i] = M[i][n] / M[i][i];\n  return x;\n}\n\n/**\n * Compute error metrics from in-sample one-step residuals.\n * `rmse` doubles as the residual sigma used to build confidence bands.\n */\nfunction calculateFitMetrics(\n  historical: number[],\n  fitted?: (number | null)[]\n): { mse: number; rmse: number; mae: number; r2: number } {\n  if (!fitted || fitted.length === 0) {\n    return { mse: 0, rmse: 0, mae: 0, r2: 0 };\n  }\n\n  let sse = 0;\n  let sae = 0;\n  let count = 0;\n  const observed: number[] = [];\n\n  for (let i = 0; i < fitted.length && i < historical.length; i++) {\n    const f = fitted[i];\n    if (f === null || f === undefined || !Number.isFinite(f)) continue;\n    const err = historical[i] - f;\n    sse += err * err;\n    sae += Math.abs(err);\n    observed.push(historical[i]);\n    count++;\n  }\n\n  if (count === 0) return { mse: 0, rmse: 0, mae: 0, r2: 0 };\n\n  const mse = sse / count;\n  const rmse = Math.sqrt(mse);\n  const mae = sae / count;\n\n  const mean = observed.reduce((s, v) => s + v, 0) / observed.length;\n  const sst = observed.reduce((s, v) => s + (v - mean) * (v - mean), 0);\n  const r2 = sst > 0 ? 1 - sse / sst : 0;\n\n  return {\n    mse: round(mse),\n    rmse: round(rmse),\n    mae: round(mae),\n    r2: round(r2),\n  };\n}\n\nfunction round(v: number): number {\n  return Math.round(v * 1e6) / 1e6;\n}\n","/**\n * Velo Plot - Forecasting Plugin\n * \n * Provides time series forecasting capabilities using native statistical models.\n */\n\nimport type { \n  PluginForecastingConfig, \n  ForecastingAPI, \n  ForecastingOptions, \n  ForecastingResult, \n  ForecastingVisualizationConfig \n} from './types';\nimport { calculateForecast } from './algorithms';\nimport type { \n  PluginManifest, \n  ChartPlugin, \n  PluginContext,\n  AfterRenderEvent\n} from '../types';\n\nconst manifestForecasting: PluginManifest = {\n  name: \"velo-plot-forecasting\",\n  version: \"1.0.0\",\n  description: \"Time series forecasting for velo-plot\",\n  provides: [\"analysis\", \"forecasting\"],\n  tags: [\"statistics\", \"forecasting\", \"analysis\"],\n};\n\n/**\n * Forecasting Plugin for Velo Plot\n */\nexport function PluginForecasting(\n  config: Partial<PluginForecastingConfig> = {}\n): ChartPlugin<PluginForecastingConfig> {\n  let ctx: PluginContext | null = null;\n  const activeForecasts = new Map<string, { result: ForecastingResult, config: ForecastingVisualizationConfig }>();\n\n  const api: ForecastingAPI = {\n    forecast(data: any, options: ForecastingOptions): ForecastingResult {\n      let x: any, y: any;\n\n      // Allow `forecast('seriesId', options)` by resolving the series data.\n      if (typeof data === 'string') {\n        if (!ctx) throw new Error('Plugin not initialized');\n        const series = ctx.data.getAllSeries().find(s => s.getId() === data);\n        if (!series) throw new Error(`Series not found: ${data}`);\n        const seriesData = series.getData();\n        x = seriesData.x;\n        y = seriesData.y;\n      } else if (Array.isArray(data)) {\n        y = data;\n        x = data.map((_, i) => i);\n      } else if (data.x && data.y) {\n        x = data.x;\n        y = data.y;\n      } else {\n        throw new Error(\"Invalid data format for forecasting\");\n      }\n\n      return calculateForecast(x, y, options.method, options.horizon, options.params, options.confidence);\n    },\n\n    async forecastSeries(seriesId: string, options: ForecastingOptions): Promise<ForecastingResult> {\n      if (!ctx) throw new Error(\"Plugin not initialized\");\n      \n      const series = ctx.data.getAllSeries().find(s => s.getId() === seriesId);\n      if (!series) throw new Error(`Series not found: ${seriesId}`);\n      \n      const data = series.getData();\n      return this.forecast(data, options);\n    },\n\n    visualize(result: ForecastingResult, vizConfig: ForecastingVisualizationConfig = {}) {\n      const id = `forecast-${Math.random().toString(36).substr(2, 9)}`;\n      const mergedConfig = {\n        showLine: true,\n        showConfidenceInterval: true,\n        ...config.defaultVisualization,\n        ...vizConfig\n      };\n      \n      activeForecasts.set(id, { result, config: mergedConfig });\n      ctx?.requestRender();\n      return id;\n    },\n\n    clear(id?: string) {\n      if (id) {\n        activeForecasts.delete(id);\n      } else {\n        activeForecasts.clear();\n      }\n      ctx?.requestRender();\n    }\n  };\n\n  function drawForecasts(pCtx: PluginContext) {\n    const { render, coords } = pCtx;\n    const { ctx2d } = render;\n    if (!ctx2d) return;\n\n    activeForecasts.forEach((viz) => {\n      const { result, config: vizConfig } = viz;\n      const { xValues, yValues, lowerBound, upperBound } = result;\n\n      ctx2d.save();\n\n      // 1. Draw Confidence Interval\n      if (vizConfig.showConfidenceInterval && lowerBound && upperBound) {\n        ctx2d.beginPath();\n        const opacity = vizConfig.intervalStyle?.opacity ?? 0.15;\n        const color = vizConfig.intervalStyle?.fillColor || (vizConfig.lineStyle?.color || '#3b82f6');\n        \n        ctx2d.fillStyle = color.startsWith('rgba') ? color : color + Math.floor(opacity * 255).toString(16).padStart(2, '0');\n        \n        for (let i = 0; i < xValues.length; i++) {\n          const px = coords.dataToPixelX(xValues[i]);\n          const py = coords.dataToPixelY(upperBound[i]);\n          if (i === 0) ctx2d.moveTo(px, py);\n          else ctx2d.lineTo(px, py);\n        }\n        \n        for (let i = xValues.length - 1; i >= 0; i--) {\n          const px = coords.dataToPixelX(xValues[i]);\n          const py = coords.dataToPixelY(lowerBound[i]);\n          ctx2d.lineTo(px, py);\n        }\n        \n        ctx2d.closePath();\n        ctx2d.fill();\n      }\n\n      // 2. Draw Forecast Line\n      if (vizConfig.showLine) {\n        ctx2d.beginPath();\n        ctx2d.strokeStyle = vizConfig.lineStyle?.color || '#3b82f6';\n        ctx2d.lineWidth = vizConfig.lineStyle?.width || 2;\n        if (vizConfig.lineStyle?.dash) {\n          ctx2d.setLineDash(vizConfig.lineStyle.dash);\n        } else {\n          ctx2d.setLineDash([5, 5]); // Default dashed for forecast\n        }\n\n        for (let i = 0; i < xValues.length; i++) {\n          const px = coords.dataToPixelX(xValues[i]);\n          const py = coords.dataToPixelY(yValues[i]);\n          if (i === 0) ctx2d.moveTo(px, py);\n          else ctx2d.lineTo(px, py);\n        }\n        ctx2d.stroke();\n      }\n\n      ctx2d.restore();\n    });\n  }\n\n  return {\n    manifest: manifestForecasting,\n    onInit(c: PluginContext) {\n      ctx = c;\n    },\n    onRenderOverlay(pCtx: PluginContext, _event: AfterRenderEvent) {\n      drawForecasts(pCtx);\n    },\n    api\n  };\n}\n\n// Re-export types and constants\nexport * from './types';\nexport * from './algorithms';\nexport { manifestForecasting };\n"],"names":["toArray","a","zForConfidence","confidence","table","best","bestDiff","entry","diff","calculateForecast","x","y","method","horizon","params","n","ya","xa","dt","lastX","result","forecastSMA","forecastEMA","forecastLinear","forecastSimpleExpSmoothing","forecastHolt","forecastWMA","forecastHoltWinters","forecastARIMA","futureX","i","metrics","calculateFitMetrics","z","sigma","lower","upper","buildConfidenceBands","yValues","accumulates","growth","margin","window","effectiveWindow","wmaAt","end","w","weightedSum","weightTotal","weight","fitted","wma","smaAt","sum","avg","alpha","ema","sumX","sumY","sumXY","sumXX","denom","slope","intercept","nextX","level","beta","trend","lastLevel","period","gamma","seasonals","seasonAverages","j","sumOverAvg","seasonal","val","m","p","d","q","order","diffLevels","k","difference","mean","s","v","wc","innovations","arLong","residuals","fitAR","start","rows","targets","t","row","coeffs","olsSolve","c","ar","ma","fittedResiduals","pred","integrateFittedPoint","wHist","eHist","forecastDiff","h","integrateForecast","series","out","current","parent","last","integrated","diffPred","X","cols","xtx","xty","b","gaussianSolve","A","M","col","pivot","r","factor","historical","sse","sae","count","observed","f","err","mse","rmse","mae","sst","r2","round","manifestForecasting","PluginForecasting","config","ctx","activeForecasts","api","data","options","seriesData","_","seriesId","vizConfig","id","mergedConfig","drawForecasts","pCtx","render","coords","ctx2d","viz","_a","_b","_c","_d","_e","_f","xValues","lowerBound","upperBound","opacity","color","px","py","_event"],"mappings":"AAwCA,SAASA,EAAQC,GAAqD;AACpE,SAAO,MAAM,KAAKA,CAAC;AACrB;AAKA,SAASC,EAAeC,GAA4B;AAClD,QAAMC,IAAiC;AAAA,IACrC,CAAC,KAAK,KAAK;AAAA,IACX,CAAC,MAAM,KAAK;AAAA,IACZ,CAAC,KAAK,KAAK;AAAA,IACX,CAAC,KAAK,KAAK;AAAA,IACX,CAAC,MAAM,IAAI;AAAA,IACX,CAAC,OAAO,KAAK;AAAA,IACb,CAAC,MAAM,KAAK;AAAA,IACZ,CAAC,OAAO,KAAK;AAAA,EAAA;AAEf,MAAIC,IAAOD,EAAM,CAAC,GACdE,IAAW;AACf,aAAWC,KAASH,GAAO;AACzB,UAAMI,IAAO,KAAK,IAAID,EAAM,CAAC,IAAIJ,CAAU;AAC3C,IAAIK,IAAOF,MACTA,IAAWE,GACXH,IAAOE;AAAA,EAEX;AACA,SAAOF,EAAK,CAAC;AACf;AAKO,SAASI,EACdC,GACAC,GACAC,GACAC,GACAC,IAA4B,CAAA,GAC5BX,IAAa,MACM;AACnB,QAAMY,IAAIL,EAAE;AACZ,MAAIK,MAAM,EAAG,OAAM,IAAI,MAAM,kBAAkB;AAE/C,QAAMC,IAAKhB,EAAQW,CAAC,GACdM,IAAKjB,EAAQU,CAAC,GAGdQ,IAAKH,IAAI,KAAKE,EAAGF,IAAI,CAAC,IAAIE,EAAG,CAAC,MAAMF,IAAI,KAAK,GAC7CI,IAAQF,EAAGF,IAAI,CAAC;AAEtB,MAAIK;AAEJ,UAAQR,GAAA;AAAA,IACN,KAAK;AACH,MAAAQ,IAASC,EAAYL,GAAIH,GAASC,EAAO,cAAc,EAAE;AACzD;AAAA,IACF,KAAK;AACH,MAAAM,IAASE,EAAYN,GAAIH,GAASC,EAAO,SAAS,GAAG;AACrD;AAAA,IACF,KAAK;AACH,MAAAM,IAASG,EAAeN,GAAID,GAAIH,CAAO;AACvC;AAAA,IACF,KAAK;AACH,MAAAO,IAASI,EAA2BR,GAAIH,GAASC,EAAO,SAAS,GAAG;AACpE;AAAA,IACF,KAAK;AACH,MAAAM,IAASK,EAAaT,GAAIH,GAASC,EAAO,SAAS,KAAKA,EAAO,QAAQ,GAAG;AAC1E;AAAA,IACF,KAAK;AACH,MAAAM,IAASM,EAAYV,GAAIH,GAASC,EAAO,cAAc,EAAE;AACzD;AAAA,IACF,KAAK;AACH,MAAAM,IAASO;AAAA,QACPX;AAAA,QACAH;AAAA,QACAC,EAAO,UAAU;AAAA,QACjBA,EAAO,SAAS;AAAA,QAChBA,EAAO,QAAQ;AAAA,QACfA,EAAO,SAAS;AAAA,MAAA;AAElB;AAAA,IACF,KAAK;AACH,MAAAM,IAASQ;AAAA,QACPZ;AAAA,QACAH;AAAA,QACAC,EAAO,KAAK;AAAA,QACZA,EAAO,KAAK;AAAA,QACZA,EAAO,KAAK;AAAA,MAAA;AAEd;AAAA,IACF;AACE,YAAM,IAAI,MAAM,UAAUF,CAAM,sBAAsB;AAAA,EAAA;AAI1D,QAAMiB,IAAU,IAAI,MAAMhB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB;AAC3B,IAAAD,EAAQC,CAAC,IAAIX,KAASW,IAAI,KAAKZ;AAIjC,QAAMa,IAAUC,EAAoBhB,GAAII,EAAO,MAAM,GAC/Ca,IAAI/B,EAAeC,CAAU,GAC7B+B,IAAQH,EAAQ,MAChB,EAAE,OAAAI,GAAO,OAAAC,EAAA,IAAUC;AAAA,IACvBjB,EAAO;AAAA,IACPc;AAAA,IACAD;AAAA,IACAb,EAAO;AAAA,EAAA;AAGT,SAAO;AAAA,IACL,SAASS;AAAA,IACT,SAAST,EAAO;AAAA,IAChB,YAAYe;AAAA,IACZ,YAAYC;AAAA,IACZ,QAAAxB;AAAA,IACA,UAAU,EAAE,GAAGmB,GAAS,YAAA5B,EAAA;AAAA,EAAW;AAEvC;AAOA,SAASkC,EACPC,GACAJ,GACAD,GACAM,GACsC;AACtC,QAAMJ,IAAQ,IAAI,MAAMG,EAAQ,MAAM,GAChCF,IAAQ,IAAI,MAAME,EAAQ,MAAM;AACtC,WAASR,IAAI,GAAGA,IAAIQ,EAAQ,QAAQR,KAAK;AACvC,UAAMU,IAASD,IAAc,KAAK,KAAKT,IAAI,CAAC,IAAI,GAC1CW,IAASR,IAAIC,IAAQM;AAC3B,IAAAL,EAAML,CAAC,IAAIQ,EAAQR,CAAC,IAAIW,GACxBL,EAAMN,CAAC,IAAIQ,EAAQR,CAAC,IAAIW;AAAA,EAC1B;AACA,SAAO,EAAE,OAAAN,GAAO,OAAAC,EAAA;AAClB;AAUA,SAASV,EAAYf,GAAaE,GAAiB6B,GAA8B;AAC/E,QAAM3B,IAAIJ,EAAE,QACNgC,IAAkB,KAAK,IAAID,GAAQ3B,CAAC,GAEpC6B,IAAQ,CAACC,MAAwB;AAErC,UAAMC,IAAI,KAAK,IAAIH,GAAiBE,CAAG;AACvC,QAAIE,IAAc,GACdC,IAAc;AAClB,aAASlB,IAAI,GAAGA,IAAIgB,GAAGhB,KAAK;AAC1B,YAAMmB,IAASnB,IAAI;AACnB,MAAAiB,KAAepC,EAAEkC,IAAMC,IAAIhB,CAAC,IAAImB,GAChCD,KAAeC;AAAA,IACjB;AACA,WAAOF,IAAcC;AAAA,EACvB,GAEME,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AACxD,WAASe,IAAI,GAAGA,IAAIf,GAAGe,IAAK,CAAAoB,EAAOpB,CAAC,IAAIc,EAAMd,CAAC;AAE/C,QAAMqB,IAAMP,EAAM7B,CAAC;AACnB,SAAO,EAAE,SAAS,IAAI,MAAMF,CAAO,EAAE,KAAKsC,CAAG,GAAG,QAAAD,GAAQ,aAAa,GAAA;AACvE;AAMA,SAAS7B,EAAYV,GAAaE,GAAiB6B,GAA8B;AAC/E,QAAM3B,IAAIJ,EAAE,QACNgC,IAAkB,KAAK,IAAID,GAAQ3B,CAAC,GAEpCqC,IAAQ,CAACP,MAAwB;AAErC,UAAMC,IAAI,KAAK,IAAIH,GAAiBE,CAAG;AACvC,QAAIQ,IAAM;AACV,aAASvB,IAAIe,IAAMC,GAAGhB,IAAIe,GAAKf,IAAK,CAAAuB,KAAO1C,EAAEmB,CAAC;AAC9C,WAAOuB,IAAMP;AAAA,EACf,GAEMI,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AACxD,WAASe,IAAI,GAAGA,IAAIf,GAAGe,IAAK,CAAAoB,EAAOpB,CAAC,IAAIsB,EAAMtB,CAAC;AAE/C,QAAMwB,IAAMF,EAAMrC,CAAC;AACnB,SAAO,EAAE,SAAS,IAAI,MAAMF,CAAO,EAAE,KAAKyC,CAAG,GAAG,QAAAJ,GAAQ,aAAa,GAAA;AACvE;AAKA,SAAS5B,EAAYX,GAAaE,GAAiB0C,GAA6B;AAC9E,QAAMxC,IAAIJ,EAAE;AACZ,MAAI6C,IAAM7C,EAAE,CAAC;AACb,QAAMuC,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AAExD,WAASe,IAAI,GAAGA,IAAIf,GAAGe;AACrB,IAAAoB,EAAOpB,CAAC,IAAI0B,GACZA,IAAMD,IAAQ5C,EAAEmB,CAAC,KAAK,IAAIyB,KAASC;AAGrC,SAAO,EAAE,SAAS,IAAI,MAAM3C,CAAO,EAAE,KAAK2C,CAAG,GAAG,QAAAN,GAAQ,aAAa,GAAA;AACvE;AAMA,SAAS3B,EAAeb,GAAaC,GAAaE,GAA+B;AAC/E,QAAME,IAAIL,EAAE;AACZ,MAAI+C,IAAO,GACTC,IAAO,GACPC,IAAQ,GACRC,IAAQ;AAEV,WAAS9B,IAAI,GAAGA,IAAIf,GAAGe;AACrB,IAAA2B,KAAQ/C,EAAEoB,CAAC,GACX4B,KAAQ/C,EAAEmB,CAAC,GACX6B,KAASjD,EAAEoB,CAAC,IAAInB,EAAEmB,CAAC,GACnB8B,KAASlD,EAAEoB,CAAC,IAAIpB,EAAEoB,CAAC;AAGrB,QAAM+B,IAAQ9C,IAAI6C,IAAQH,IAAOA,GAC3BK,IAAQD,MAAU,KAAK9C,IAAI4C,IAAQF,IAAOC,KAAQG,IAAQ,GAC1DE,KAAaL,IAAOI,IAAQL,KAAQ1C,GAEpCmC,IAA4B,IAAI,MAAMnC,CAAC;AAC7C,WAASe,IAAI,GAAGA,IAAIf,GAAGe,IAAK,CAAAoB,EAAOpB,CAAC,IAAIgC,IAAQpD,EAAEoB,CAAC,IAAIiC;AAEvD,QAAM5C,IAAQT,EAAEK,IAAI,CAAC,GACfG,IAAKH,IAAI,KAAKL,EAAEK,IAAI,CAAC,IAAIL,EAAE,CAAC,MAAMK,IAAI,KAAK,GAC3CuB,IAAU,IAAI,MAAMzB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB,KAAK;AAChC,UAAMkC,IAAQ7C,KAASW,IAAI,KAAKZ;AAChC,IAAAoB,EAAQR,CAAC,IAAIgC,IAAQE,IAAQD;AAAA,EAC/B;AAEA,SAAO,EAAE,SAAAzB,GAAS,QAAAY,GAAQ,aAAa,GAAA;AACzC;AAKA,SAAS1B,EAA2Bb,GAAaE,GAAiB0C,GAA6B;AAC7F,QAAMxC,IAAIJ,EAAE;AACZ,MAAIsD,IAAQtD,EAAE,CAAC;AACf,QAAMuC,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AAExD,WAASe,IAAI,GAAGA,IAAIf,GAAGe;AACrB,IAAAoB,EAAOpB,CAAC,IAAImC,GACZA,IAAQV,IAAQ5C,EAAEmB,CAAC,KAAK,IAAIyB,KAASU;AAGvC,SAAO,EAAE,SAAS,IAAI,MAAMpD,CAAO,EAAE,KAAKoD,CAAK,GAAG,QAAAf,GAAQ,aAAa,GAAA;AACzE;AAKA,SAASzB,EAAad,GAAaE,GAAiB0C,GAAeW,GAA4B;AAC7F,QAAMnD,IAAIJ,EAAE;AACZ,MAAII,IAAI;AACN,WAAO,EAAE,SAAS,IAAI,MAAMF,CAAO,EAAE,KAAKF,EAAE,CAAC,KAAK,CAAC,GAAG,QAAQ,CAAA,GAAI,aAAa,GAAA;AAGjF,MAAIsD,IAAQtD,EAAE,CAAC,GACXwD,IAAQxD,EAAE,CAAC,IAAIA,EAAE,CAAC;AACtB,QAAMuC,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AAExD,WAAS,IAAI,GAAG,IAAIA,GAAG,KAAK;AAC1B,IAAAmC,EAAO,CAAC,IAAIe,IAAQE;AACpB,UAAMC,IAAYH;AAClB,IAAAA,IAAQV,IAAQ5C,EAAE,CAAC,KAAK,IAAI4C,MAAUU,IAAQE,IAC9CA,IAAQD,KAAQD,IAAQG,MAAc,IAAIF,KAAQC;AAAA,EACpD;AAEA,QAAM7B,IAAU,IAAI,MAAMzB,CAAO;AACjC,WAAS,IAAI,GAAG,IAAIA,GAAS,OAAa,CAAC,IAAIoD,KAAS,IAAI,KAAKE;AAEjE,SAAO,EAAE,SAAA7B,GAAS,QAAAY,GAAQ,aAAa,GAAA;AACzC;AAKA,SAASvB,EACPhB,GACAE,GACAwD,GACAd,GACAW,GACAI,GACc;AACd,QAAMvD,IAAIJ,EAAE;AACZ,MAAII,IAAIsD,IAAS;AAEf,WAAO5C,EAAad,GAAGE,GAAS0C,GAAOW,CAAI;AAI7C,QAAMK,IAAY,IAAI,MAAMF,CAAM,GAC5BG,IAAiB,IAAI,MAAM,KAAK,MAAMzD,IAAIsD,CAAM,CAAC;AACvD,WAASvC,IAAI,GAAGA,IAAI0C,EAAe,QAAQ1C,KAAK;AAC9C,QAAIuB,IAAM;AACV,aAASoB,IAAI,GAAGA,IAAIJ,GAAQI,IAAK,CAAApB,KAAO1C,EAAEmB,IAAIuC,IAASI,CAAC;AACxD,IAAAD,EAAe1C,CAAC,IAAIuB,IAAMgB;AAAA,EAC5B;AAEA,WAASvC,IAAI,GAAGA,IAAIuC,GAAQvC,KAAK;AAC/B,QAAI4C,IAAa;AACjB,aAASD,IAAI,GAAGA,IAAID,EAAe,QAAQC;AACzC,MAAAC,KAAc/D,EAAE8D,IAAIJ,IAASvC,CAAC,IAAI0C,EAAeC,CAAC;AAEpD,IAAAF,EAAUzC,CAAC,IAAI4C,IAAaF,EAAe;AAAA,EAC7C;AAGA,MAAIP,IAAQO,EAAe,CAAC,GACxBL,KACDK,EAAeA,EAAe,SAAS,CAAC,IAAIA,EAAe,CAAC,OAC3DA,EAAe,SAAS,KAAKH;AAEjC,QAAMnB,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AAGxD,WAASe,IAAI,GAAGA,IAAIf,GAAGe,KAAK;AAC1B,UAAM6C,IAAWJ,EAAUzC,IAAIuC,CAAM;AACrC,IAAIvC,IAAI,MAAGoB,EAAOpB,CAAC,IAAImC,IAAQE,IAAQQ;AACvC,UAAMP,IAAYH,GACZW,IAAMjE,EAAEmB,CAAC;AACf,IAAAmC,IAAQV,KAASqB,IAAMD,MAAa,IAAIpB,MAAUU,IAAQE,IAC1DA,IAAQD,KAAQD,IAAQG,MAAc,IAAIF,KAAQC,GAClDI,EAAUzC,IAAIuC,CAAM,IAAIC,KAASM,IAAMX,MAAU,IAAIK,KAASK;AAAA,EAChE;AAGA,QAAMrC,IAAU,IAAI,MAAMzB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB,KAAK;AAChC,UAAM+C,IAAI/C,IAAI;AACd,IAAAQ,EAAQR,CAAC,IAAImC,IAAQY,IAAIV,IAAQI,GAAWxD,IAAIe,KAAKuC,CAAM;AAAA,EAC7D;AAEA,SAAO,EAAE,SAAA/B,GAAS,QAAAY,GAAQ,aAAa,GAAA;AACzC;AAkBA,SAAStB,EACPjB,GACAE,GACAiE,GACAC,GACAC,GACc;AACd,QAAMjE,IAAIJ,EAAE,QACNsE,IAAQ,KAAK,IAAIH,GAAGE,CAAC;AAE3B,MAAIjE,IAAIgE,IAAIE,IAAQ,KAAKlE,IAAI;AAC3B,WAAOU,EAAad,GAAGE,GAAS,KAAK,GAAG;AAI1C,QAAMqE,IAAyB,CAACvE,EAAE,OAAO;AACzC,WAASwE,IAAI,GAAGA,IAAIJ,GAAGI;AACrB,IAAAD,EAAW,KAAKE,EAAWF,EAAWC,CAAC,CAAC,CAAC;AAE3C,QAAMrC,IAAIoC,EAAWH,CAAC,GAChBF,IAAI/B,EAAE,QAGNuC,IAAOvC,EAAE,OAAO,CAACwC,GAAGC,MAAMD,IAAIC,GAAG,CAAC,IAAIV,GACtCW,IAAK1C,EAAE,IAAI,CAACyC,MAAMA,IAAIF,CAAI;AAGhC,MAAII,IAAc,IAAI,MAAMZ,CAAC,EAAE,KAAK,CAAC;AACrC,MAAIG,IAAI,GAAG;AACT,UAAMU,IAAS,KAAK,IAAI,KAAK,IAAIZ,IAAIE,IAAI,GAAG,CAAC,GAAG,KAAK,MAAMH,IAAI,CAAC,CAAC,GAC3D,EAAE,WAAAc,EAAA,IAAcC,EAAMJ,GAAIE,CAAM;AACtC,IAAAD,IAAcE;AAAA,EAChB;AAGA,QAAME,IAAQ,KAAK,IAAIf,GAAGE,CAAC,GACrBc,IAAmB,CAAA,GACnBC,IAAoB,CAAA;AAC1B,WAASC,IAAIH,GAAOG,IAAInB,GAAGmB,KAAK;AAC9B,UAAMC,IAAgB,CAAA;AACtB,aAASnE,IAAI,GAAGA,KAAKgD,GAAGhD,OAAS,KAAK0D,EAAGQ,IAAIlE,CAAC,CAAC;AAC/C,aAAS2C,IAAI,GAAGA,KAAKO,GAAGP,OAAS,KAAKgB,EAAYO,IAAIvB,CAAC,CAAC;AACxD,IAAAqB,EAAK,KAAKG,CAAG,GACbF,EAAQ,KAAKP,EAAGQ,CAAC,CAAC;AAAA,EACpB;AAEA,MAAIE,IAAmB,CAAA;AAIvB,MAHIJ,EAAK,SAAS,KAAKA,EAAK,CAAC,EAAE,SAAS,MACtCI,IAASC,EAASL,GAAMC,CAAO,IAE7BG,EAAO,WAAWpB,IAAIE,KAAKkB,EAAO,KAAK,CAACE,MAAM,CAAC,OAAO,SAASA,CAAC,CAAC;AAEnE,WAAO3E,EAAad,GAAGE,GAAS,KAAK,GAAG;AAE1C,QAAMwF,IAAKH,EAAO,MAAM,GAAGpB,CAAC,GACtBwB,IAAKJ,EAAO,MAAMpB,GAAGA,IAAIE,CAAC,GAG1B9B,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI,GAClDwF,IAAkB,IAAI,MAAM1B,CAAC,EAAE,KAAK,CAAC;AAC3C,WAASmB,IAAIH,GAAOG,IAAInB,GAAGmB,KAAK;AAC9B,QAAIQ,IAAO;AACX,aAAS1E,IAAI,GAAGA,KAAKgD,GAAGhD,IAAK,CAAA0E,KAAQH,EAAGvE,IAAI,CAAC,IAAI0D,EAAGQ,IAAIlE,CAAC;AACzD,aAAS2C,IAAI,GAAGA,KAAKO,GAAGP,IAAK,CAAA+B,KAAQF,EAAG7B,IAAI,CAAC,IAAI8B,EAAgBP,IAAIvB,CAAC;AACtE,IAAA8B,EAAgBP,CAAC,IAAIR,EAAGQ,CAAC,IAAIQ,GAI7BtD,EAAO8C,KAAKjF,IAAI8D,EAAE,IAAI4B,EAAqBvB,GAAYH,GAAGiB,GAAGQ,IAAOnB,CAAI;AAAA,EAC1E;AAGA,QAAMqB,IAAQlB,EAAG,MAAA,GACXmB,IAAQJ,EAAgB,MAAA,GACxBK,IAAyB,CAAA;AAC/B,WAASC,IAAI,GAAGA,IAAIhG,GAASgG,KAAK;AAChC,QAAIL,IAAO;AAGX,aAAS1E,IAAI,GAAGA,KAAKgD,GAAGhD;AACtB,MAAA0E,KAAQH,EAAGvE,IAAI,CAAC,IAAI4E,EAAMA,EAAM,SAAS5E,CAAC;AAE5C,aAAS2C,IAAI,GAAGA,KAAKO,GAAGP;AACtB,MAAA+B,KAAQF,EAAG7B,IAAI,CAAC,IAAIkC,EAAMA,EAAM,SAASlC,CAAC;AAE5C,IAAAiC,EAAM,KAAKF,CAAI,GACfG,EAAM,KAAK,CAAC,GACZC,EAAa,KAAKJ,IAAOnB,CAAI;AAAA,EAC/B;AAKA,SAAO,EAAE,SAFOyB,EAAkB5B,GAAYH,GAAG6B,CAAY,GAE3C,QAAA1D,GAAQ,aAAa,GAAA;AACzC;AAGA,SAASkC,EAAW2B,GAA4B;AAC9C,QAAMC,IAAM,IAAI,MAAMD,EAAO,SAAS,CAAC;AACvC,WAASjF,IAAI,GAAGA,IAAIiF,EAAO,QAAQjF,IAAK,CAAAkF,EAAIlF,IAAI,CAAC,IAAIiF,EAAOjF,CAAC,IAAIiF,EAAOjF,IAAI,CAAC;AAC7E,SAAOkF;AACT;AAMA,SAASF,EAAkB5B,GAAwBH,GAAW6B,GAAkC;AAC9F,MAAIK,IAAUL,EAAa,MAAA;AAC3B,WAASzB,IAAIJ,GAAGI,KAAK,GAAGA,KAAK;AAC3B,UAAM+B,IAAShC,EAAWC,IAAI,CAAC;AAC/B,QAAIgC,IAAOD,EAAOA,EAAO,SAAS,CAAC;AACnC,UAAME,IAAa,IAAI,MAAMH,EAAQ,MAAM;AAC3C,aAASnF,IAAI,GAAGA,IAAImF,EAAQ,QAAQnF;AAClC,MAAAqF,KAAQF,EAAQnF,CAAC,GACjBsF,EAAWtF,CAAC,IAAIqF;AAElB,IAAAF,IAAUG;AAAA,EACZ;AACA,SAAOH;AACT;AAMA,SAASR,EACPvB,GACAH,GACAiB,GACAqB,GACQ;AACR,SAAItC,MAAM,IAAUsC,IAELnC,EAAWH,IAAI,CAAC,EAEXiB,CAAC,IAEPqB;AAChB;AAMA,SAASzB,EAAMmB,GAAkB9B,GAA0D;AACzF,QAAMJ,IAAIkC,EAAO,QACXjB,IAAmB,CAAA,GACnBC,IAAoB,CAAA;AAC1B,WAASC,IAAIf,GAAOe,IAAInB,GAAGmB,KAAK;AAC9B,UAAMC,IAAgB,CAAA;AACtB,aAAS,IAAI,GAAG,KAAKhB,GAAO,OAAS,KAAK8B,EAAOf,IAAI,CAAC,CAAC;AACvD,IAAAF,EAAK,KAAKG,CAAG,GACbF,EAAQ,KAAKgB,EAAOf,CAAC,CAAC;AAAA,EACxB;AACA,QAAME,IAASJ,EAAK,SAAS,IAAIK,EAASL,GAAMC,CAAO,IAAI,IAAI,MAAMd,CAAK,EAAE,KAAK,CAAC,GAC5EU,IAAY,IAAI,MAAMd,CAAC,EAAE,KAAK,CAAC;AACrC,WAASmB,IAAIf,GAAOe,IAAInB,GAAGmB,KAAK;AAC9B,QAAIQ,IAAO;AACX,aAAS,IAAI,GAAG,KAAKvB,GAAO,IAAK,CAAAuB,MAASN,EAAO,IAAI,CAAC,KAAK,KAAKa,EAAOf,IAAI,CAAC;AAC5E,IAAAL,EAAUK,CAAC,IAAIe,EAAOf,CAAC,IAAIQ;AAAA,EAC7B;AACA,SAAO,EAAE,QAAAN,GAAQ,WAAAP,EAAA;AACnB;AAMA,SAASQ,EAASmB,GAAe3G,GAAuB;AACtD,QAAMmF,IAAOwB,EAAE,QACTC,IAAOD,EAAE,CAAC,EAAE,QAGZE,IAAkB,MAAM,KAAK,EAAE,QAAQD,EAAA,GAAQ,MAAM,IAAI,MAAMA,CAAI,EAAE,KAAK,CAAC,CAAC,GAC5EE,IAAgB,IAAI,MAAMF,CAAI,EAAE,KAAK,CAAC;AAE5C,WAASzF,IAAI,GAAGA,IAAIgE,GAAMhE;AACxB,aAAS7B,IAAI,GAAGA,IAAIsH,GAAMtH,KAAK;AAC7B,MAAAwH,EAAIxH,CAAC,KAAKqH,EAAExF,CAAC,EAAE7B,CAAC,IAAIU,EAAEmB,CAAC;AACvB,eAAS4F,IAAI,GAAGA,IAAIH,GAAMG;AACxB,QAAAF,EAAIvH,CAAC,EAAEyH,CAAC,KAAKJ,EAAExF,CAAC,EAAE7B,CAAC,IAAIqH,EAAExF,CAAC,EAAE4F,CAAC;AAAA,IAEjC;AAIF,WAASzH,IAAI,GAAGA,IAAIsH,GAAMtH,IAAK,CAAAuH,EAAIvH,CAAC,EAAEA,CAAC,KAAK;AAE5C,SAAO0H,EAAcH,GAAKC,CAAG;AAC/B;AAMO,SAASE,EAAcC,GAAeF,GAAuB;AAClE,QAAM3G,IAAI2G,EAAE,QACNG,IAAID,EAAE,IAAI,CAAC3B,GAAKnE,MAAM,CAAC,GAAGmE,GAAKyB,EAAE5F,CAAC,CAAC,CAAC;AAE1C,WAASgG,IAAM,GAAGA,IAAM/G,GAAG+G,KAAO;AAEhC,QAAIC,IAAQD;AACZ,aAASE,IAAIF,IAAM,GAAGE,IAAIjH,GAAGiH;AAC3B,MAAI,KAAK,IAAIH,EAAEG,CAAC,EAAEF,CAAG,CAAC,IAAI,KAAK,IAAID,EAAEE,CAAK,EAAED,CAAG,CAAC,MAAGC,IAAQC;AAE7D,QAAI,KAAK,IAAIH,EAAEE,CAAK,EAAED,CAAG,CAAC,IAAI,MAAO,QAAO,CAAA;AAC5C,KAACD,EAAEC,CAAG,GAAGD,EAAEE,CAAK,CAAC,IAAI,CAACF,EAAEE,CAAK,GAAGF,EAAEC,CAAG,CAAC;AAGtC,aAASE,IAAI,GAAGA,IAAIjH,GAAGiH,KAAK;AAC1B,UAAIA,MAAMF,EAAK;AACf,YAAMG,IAASJ,EAAEG,CAAC,EAAEF,CAAG,IAAID,EAAEC,CAAG,EAAEA,CAAG;AACrC,eAAS1B,IAAI0B,GAAK1B,KAAKrF,GAAGqF,IAAK,CAAAyB,EAAEG,CAAC,EAAE5B,CAAC,KAAK6B,IAASJ,EAAEC,CAAG,EAAE1B,CAAC;AAAA,IAC7D;AAAA,EACF;AAEA,QAAM1F,IAAI,IAAI,MAAMK,CAAC;AACrB,WAASe,IAAI,GAAGA,IAAIf,GAAGe,OAAOA,CAAC,IAAI+F,EAAE/F,CAAC,EAAEf,CAAC,IAAI8G,EAAE/F,CAAC,EAAEA,CAAC;AACnD,SAAOpB;AACT;AAMA,SAASsB,EACPkG,GACAhF,GACwD;AACxD,MAAI,CAACA,KAAUA,EAAO,WAAW;AAC/B,WAAO,EAAE,KAAK,GAAG,MAAM,GAAG,KAAK,GAAG,IAAI,EAAA;AAGxC,MAAIiF,IAAM,GACNC,IAAM,GACNC,IAAQ;AACZ,QAAMC,IAAqB,CAAA;AAE3B,WAASxG,IAAI,GAAGA,IAAIoB,EAAO,UAAUpB,IAAIoG,EAAW,QAAQpG,KAAK;AAC/D,UAAMyG,IAAIrF,EAAOpB,CAAC;AAClB,QAAIyG,KAAM,QAA2B,CAAC,OAAO,SAASA,CAAC,EAAG;AAC1D,UAAMC,IAAMN,EAAWpG,CAAC,IAAIyG;AAC5B,IAAAJ,KAAOK,IAAMA,GACbJ,KAAO,KAAK,IAAII,CAAG,GACnBF,EAAS,KAAKJ,EAAWpG,CAAC,CAAC,GAC3BuG;AAAA,EACF;AAEA,MAAIA,MAAU,EAAG,QAAO,EAAE,KAAK,GAAG,MAAM,GAAG,KAAK,GAAG,IAAI,EAAA;AAEvD,QAAMI,IAAMN,IAAME,GACZK,IAAO,KAAK,KAAKD,CAAG,GACpBE,IAAMP,IAAMC,GAEZhD,IAAOiD,EAAS,OAAO,CAAChD,GAAGC,MAAMD,IAAIC,GAAG,CAAC,IAAI+C,EAAS,QACtDM,IAAMN,EAAS,OAAO,CAAChD,GAAGC,MAAMD,KAAKC,IAAIF,MAASE,IAAIF,IAAO,CAAC,GAC9DwD,IAAKD,IAAM,IAAI,IAAIT,IAAMS,IAAM;AAErC,SAAO;AAAA,IACL,KAAKE,EAAML,CAAG;AAAA,IACd,MAAMK,EAAMJ,CAAI;AAAA,IAChB,KAAKI,EAAMH,CAAG;AAAA,IACd,IAAIG,EAAMD,CAAE;AAAA,EAAA;AAEhB;AAEA,SAASC,EAAMvD,GAAmB;AAChC,SAAO,KAAK,MAAMA,IAAI,GAAG,IAAI;AAC/B;ACppBA,MAAMwD,IAAsC;AAAA,EAC1C,MAAM;AAAA,EACN,SAAS;AAAA,EACT,aAAa;AAAA,EACb,UAAU,CAAC,YAAY,aAAa;AAAA,EACpC,MAAM,CAAC,cAAc,eAAe,UAAU;AAChD;AAKO,SAASC,EACdC,IAA2C,IACL;AACtC,MAAIC,IAA4B;AAChC,QAAMC,wBAAsB,IAAA,GAEtBC,IAAsB;AAAA,IAC1B,SAASC,GAAWC,GAAgD;AAClE,UAAI5I,GAAQC;AAGZ,UAAI,OAAO0I,KAAS,UAAU;AAC5B,YAAI,CAACH,EAAK,OAAM,IAAI,MAAM,wBAAwB;AAClD,cAAMnC,IAASmC,EAAI,KAAK,aAAA,EAAe,KAAK,CAAA5D,MAAKA,EAAE,MAAA,MAAY+D,CAAI;AACnE,YAAI,CAACtC,EAAQ,OAAM,IAAI,MAAM,qBAAqBsC,CAAI,EAAE;AACxD,cAAME,IAAaxC,EAAO,QAAA;AAC1B,QAAArG,IAAI6I,EAAW,GACf5I,IAAI4I,EAAW;AAAA,MACjB,WAAW,MAAM,QAAQF,CAAI;AAC3B,QAAA1I,IAAI0I,GACJ3I,IAAI2I,EAAK,IAAI,CAACG,GAAG1H,MAAMA,CAAC;AAAA,eACfuH,EAAK,KAAKA,EAAK;AACxB,QAAA3I,IAAI2I,EAAK,GACT1I,IAAI0I,EAAK;AAAA;AAET,cAAM,IAAI,MAAM,qCAAqC;AAGvD,aAAO5I,EAAkBC,GAAGC,GAAG2I,EAAQ,QAAQA,EAAQ,SAASA,EAAQ,QAAQA,EAAQ,UAAU;AAAA,IACpG;AAAA,IAEA,MAAM,eAAeG,GAAkBH,GAAyD;AAC9F,UAAI,CAACJ,EAAK,OAAM,IAAI,MAAM,wBAAwB;AAElD,YAAMnC,IAASmC,EAAI,KAAK,aAAA,EAAe,KAAK,CAAA5D,MAAKA,EAAE,MAAA,MAAYmE,CAAQ;AACvE,UAAI,CAAC1C,EAAQ,OAAM,IAAI,MAAM,qBAAqB0C,CAAQ,EAAE;AAE5D,YAAMJ,IAAOtC,EAAO,QAAA;AACpB,aAAO,KAAK,SAASsC,GAAMC,CAAO;AAAA,IACpC;AAAA,IAEA,UAAUlI,GAA2BsI,IAA4C,IAAI;AACnF,YAAMC,IAAK,YAAY,KAAK,OAAA,EAAS,SAAS,EAAE,EAAE,OAAO,GAAG,CAAC,CAAC,IACxDC,IAAe;AAAA,QACnB,UAAU;AAAA,QACV,wBAAwB;AAAA,QACxB,GAAGX,EAAO;AAAA,QACV,GAAGS;AAAA,MAAA;AAGL,aAAAP,EAAgB,IAAIQ,GAAI,EAAE,QAAAvI,GAAQ,QAAQwI,GAAc,GACxDV,KAAA,QAAAA,EAAK,iBACES;AAAA,IACT;AAAA,IAEA,MAAMA,GAAa;AACjB,MAAIA,IACFR,EAAgB,OAAOQ,CAAE,IAEzBR,EAAgB,MAAA,GAElBD,KAAA,QAAAA,EAAK;AAAA,IACP;AAAA,EAAA;AAGF,WAASW,EAAcC,GAAqB;AAC1C,UAAM,EAAE,QAAAC,GAAQ,QAAAC,EAAA,IAAWF,GACrB,EAAE,OAAAG,MAAUF;AAClB,IAAKE,KAELd,EAAgB,QAAQ,CAACe,MAAQ;AD9DrC,UAAAC,GAAAC,GAAAC,GAAAC,GAAAC,GAAAC;AC+DM,YAAM,EAAE,QAAApJ,GAAQ,QAAQsI,EAAA,IAAcQ,GAChC,EAAE,SAAAO,GAAS,SAAAnI,GAAS,YAAAoI,GAAY,YAAAC,MAAevJ;AAKrD,UAHA6I,EAAM,KAAA,GAGFP,EAAU,0BAA0BgB,KAAcC,GAAY;AAChE,QAAAV,EAAM,UAAA;AACN,cAAMW,MAAUT,IAAAT,EAAU,kBAAV,gBAAAS,EAAyB,YAAW,MAC9CU,MAAQT,IAAAV,EAAU,kBAAV,gBAAAU,EAAyB,gBAAcC,IAAAX,EAAU,cAAV,gBAAAW,EAAqB,UAAS;AAEnF,QAAAJ,EAAM,YAAYY,EAAM,WAAW,MAAM,IAAIA,IAAQA,IAAQ,KAAK,MAAMD,IAAU,GAAG,EAAE,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG;AAEnH,iBAAS9I,IAAI,GAAGA,IAAI2I,EAAQ,QAAQ3I,KAAK;AACvC,gBAAMgJ,IAAKd,EAAO,aAAaS,EAAQ3I,CAAC,CAAC,GACnCiJ,IAAKf,EAAO,aAAaW,EAAW7I,CAAC,CAAC;AAC5C,UAAIA,MAAM,IAAGmI,EAAM,OAAOa,GAAIC,CAAE,IAC3Bd,EAAM,OAAOa,GAAIC,CAAE;AAAA,QAC1B;AAEA,iBAASjJ,IAAI2I,EAAQ,SAAS,GAAG3I,KAAK,GAAGA,KAAK;AAC5C,gBAAMgJ,IAAKd,EAAO,aAAaS,EAAQ3I,CAAC,CAAC,GACnCiJ,IAAKf,EAAO,aAAaU,EAAW5I,CAAC,CAAC;AAC5C,UAAAmI,EAAM,OAAOa,GAAIC,CAAE;AAAA,QACrB;AAEA,QAAAd,EAAM,UAAA,GACNA,EAAM,KAAA;AAAA,MACR;AAGA,UAAIP,EAAU,UAAU;AACtB,QAAAO,EAAM,UAAA,GACNA,EAAM,gBAAcK,IAAAZ,EAAU,cAAV,gBAAAY,EAAqB,UAAS,WAClDL,EAAM,cAAYM,IAAAb,EAAU,cAAV,gBAAAa,EAAqB,UAAS,IAC5CC,IAAAd,EAAU,cAAV,QAAAc,EAAqB,OACvBP,EAAM,YAAYP,EAAU,UAAU,IAAI,IAE1CO,EAAM,YAAY,CAAC,GAAG,CAAC,CAAC;AAG1B,iBAASnI,IAAI,GAAGA,IAAI2I,EAAQ,QAAQ3I,KAAK;AACvC,gBAAMgJ,IAAKd,EAAO,aAAaS,EAAQ3I,CAAC,CAAC,GACnCiJ,IAAKf,EAAO,aAAa1H,EAAQR,CAAC,CAAC;AACzC,UAAIA,MAAM,IAAGmI,EAAM,OAAOa,GAAIC,CAAE,IAC3Bd,EAAM,OAAOa,GAAIC,CAAE;AAAA,QAC1B;AACA,QAAAd,EAAM,OAAA;AAAA,MACR;AAEA,MAAAA,EAAM,QAAA;AAAA,IACR,CAAC;AAAA,EACH;AAEA,SAAO;AAAA,IACL,UAAUlB;AAAA,IACV,OAAO3C,GAAkB;AACvB,MAAA8C,IAAM9C;AAAA,IACR;AAAA,IACA,gBAAgB0D,GAAqBkB,GAA0B;AAC7D,MAAAnB,EAAcC,CAAI;AAAA,IACpB;AAAA,IACA,KAAAV;AAAA,EAAA;AAEJ;"}
+{"version":3,"file":"forecasting.js","sources":["../../src/plugins/forecasting/algorithms.ts","../../src/plugins/forecasting/index.ts"],"sourcesContent":["/**\n * Forecasting Algorithms\n *\n * Native implementations of time series forecasting methods.\n *\n * ---------------------------------------------------------------------------\n * Method status audit (task 3.1)\n * ---------------------------------------------------------------------------\n * | Method        | Status       | Notes                                     |\n * |---------------|--------------|-------------------------------------------|\n * | sma           | implemented  | Flat projection at trailing average       |\n * | wma           | implemented  | Flat projection at weighted trailing avg  |\n * | ema           | implemented  | Flat projection at last EMA value         |\n * | expSmoothing  | implemented  | Simple exponential smoothing (SES)        |\n * | holt          | implemented  | Double exponential smoothing (trend)      |\n * | holtWinters   | implemented  | Triple exponential smoothing (seasonal)   |\n * | linear        | implemented  | OLS linear trend projection               |\n * | arima         | implemented  | ARIMA(p,d,q) via Hannan-Rissanen (simple) |\n *\n * Every public method returns a finite forecast plus 95% (configurable)\n * confidence bands derived from in-sample one-step residuals. No public\n * method throws for a supported `ForecastingMethod`.\n */\n\nimport type { ForecastingMethod, ForecastingParams, ForecastingResult } from './types';\n\n/** Internal result of a single method: forecast plus in-sample fitted values. */\ninterface MethodResult {\n  /** Point forecast for the requested horizon. */\n  yValues: number[];\n  /**\n   * In-sample one-step-ahead fitted values aligned with the history `y`.\n   * `fitted[i]` is the prediction of `y[i]` using only `y[0..i-1]`.\n   * `null` entries are ignored when computing residuals.\n   */\n  fitted?: (number | null)[];\n  /** Whether forecast uncertainty accumulates with the horizon (sqrt growth). */\n  accumulates: boolean;\n}\n\nfunction toArray(a: number[] | Float64Array | Float32Array): number[] {\n  return Array.from(a);\n}\n\n/**\n * Map a confidence level (e.g. 0.95) to a normal-distribution z multiplier.\n */\nfunction zForConfidence(confidence: number): number {\n  const table: Array<[number, number]> = [\n    [0.5, 0.674],\n    [0.68, 0.994],\n    [0.8, 1.282],\n    [0.9, 1.645],\n    [0.95, 1.96],\n    [0.975, 2.241],\n    [0.99, 2.576],\n    [0.995, 2.807],\n  ];\n  let best = table[0];\n  let bestDiff = Infinity;\n  for (const entry of table) {\n    const diff = Math.abs(entry[0] - confidence);\n    if (diff < bestDiff) {\n      bestDiff = diff;\n      best = entry;\n    }\n  }\n  return best[1];\n}\n\n/**\n * Main dispatcher for forecasting algorithms.\n */\nexport function calculateForecast(\n  x: number[] | Float64Array | Float32Array,\n  y: number[] | Float64Array | Float32Array,\n  method: ForecastingMethod,\n  horizon: number,\n  params: ForecastingParams = {},\n  confidence = 0.95\n): ForecastingResult {\n  const n = x.length;\n  if (n === 0) throw new Error('Dataset is empty');\n\n  const ya = toArray(y);\n  const xa = toArray(x);\n\n  // Determine time step (assuming regular intervals)\n  const dt = n > 1 ? (xa[n - 1] - xa[0]) / (n - 1) : 1;\n  const lastX = xa[n - 1];\n\n  let result: MethodResult;\n\n  switch (method) {\n    case 'sma':\n      result = forecastSMA(ya, horizon, params.windowSize || 10);\n      break;\n    case 'ema':\n      result = forecastEMA(ya, horizon, params.alpha ?? 0.3);\n      break;\n    case 'linear':\n      result = forecastLinear(xa, ya, horizon);\n      break;\n    case 'expSmoothing':\n      result = forecastSimpleExpSmoothing(ya, horizon, params.alpha ?? 0.3);\n      break;\n    case 'holt':\n      result = forecastHolt(ya, horizon, params.alpha ?? 0.3, params.beta ?? 0.1);\n      break;\n    case 'wma':\n      result = forecastWMA(ya, horizon, params.windowSize || 10);\n      break;\n    case 'holtWinters':\n      result = forecastHoltWinters(\n        ya,\n        horizon,\n        params.period || 12,\n        params.alpha ?? 0.3,\n        params.beta ?? 0.1,\n        params.gamma ?? 0.1\n      );\n      break;\n    case 'arima':\n      result = forecastARIMA(\n        ya,\n        horizon,\n        params.p ?? 1,\n        params.d ?? 1,\n        params.q ?? 0\n      );\n      break;\n    default:\n      throw new Error(`Method ${method} not implemented yet`);\n  }\n\n  // Generate future X values\n  const futureX = new Array(horizon);\n  for (let i = 0; i < horizon; i++) {\n    futureX[i] = lastX + (i + 1) * dt;\n  }\n\n  // Compute residual-based statistics and confidence bands.\n  const metrics = calculateFitMetrics(ya, result.fitted);\n  const z = zForConfidence(confidence);\n  const sigma = metrics.rmse;\n  const { lower, upper } = buildConfidenceBands(\n    result.yValues,\n    sigma,\n    z,\n    result.accumulates\n  );\n\n  return {\n    xValues: futureX,\n    yValues: result.yValues,\n    lowerBound: lower,\n    upperBound: upper,\n    method,\n    metadata: { ...metrics, confidence },\n  };\n}\n\n/**\n * Build symmetric confidence bands around the point forecast.\n * When `accumulates` is true the uncertainty widens with sqrt(horizon step),\n * reflecting the compounding error of trend/AR models.\n */\nfunction buildConfidenceBands(\n  yValues: number[],\n  sigma: number,\n  z: number,\n  accumulates: boolean\n): { lower: number[]; upper: number[] } {\n  const lower = new Array(yValues.length);\n  const upper = new Array(yValues.length);\n  for (let i = 0; i < yValues.length; i++) {\n    const growth = accumulates ? Math.sqrt(i + 1) : 1;\n    const margin = z * sigma * growth;\n    lower[i] = yValues[i] - margin;\n    upper[i] = yValues[i] + margin;\n  }\n  return { lower, upper };\n}\n\n// ===========================================================================\n// Moving-average family\n// ===========================================================================\n\n/**\n * Weighted Moving Average (WMA).\n * Uses linear weights on the trailing window; projects flat at last WMA value.\n */\nfunction forecastWMA(y: number[], horizon: number, window: number): MethodResult {\n  const n = y.length;\n  const effectiveWindow = Math.min(window, n);\n\n  const wmaAt = (end: number): number => {\n    // end is always >= 1 here, so the trailing window w is always >= 1.\n    const w = Math.min(effectiveWindow, end);\n    let weightedSum = 0;\n    let weightTotal = 0;\n    for (let i = 0; i < w; i++) {\n      const weight = i + 1;\n      weightedSum += y[end - w + i] * weight;\n      weightTotal += weight;\n    }\n    return weightedSum / weightTotal;\n  };\n\n  const fitted: (number | null)[] = new Array(n).fill(null);\n  for (let i = 1; i < n; i++) fitted[i] = wmaAt(i);\n\n  const wma = wmaAt(n);\n  return { yValues: new Array(horizon).fill(wma), fitted, accumulates: false };\n}\n\n/**\n * Simple Moving Average (SMA).\n * Historical SMA used for projection (flat line at last average).\n */\nfunction forecastSMA(y: number[], horizon: number, window: number): MethodResult {\n  const n = y.length;\n  const effectiveWindow = Math.min(window, n);\n\n  const smaAt = (end: number): number => {\n    // end is always >= 1 here, so the trailing window w is always >= 1.\n    const w = Math.min(effectiveWindow, end);\n    let sum = 0;\n    for (let i = end - w; i < end; i++) sum += y[i];\n    return sum / w;\n  };\n\n  const fitted: (number | null)[] = new Array(n).fill(null);\n  for (let i = 1; i < n; i++) fitted[i] = smaAt(i);\n\n  const avg = smaAt(n);\n  return { yValues: new Array(horizon).fill(avg), fitted, accumulates: false };\n}\n\n/**\n * Exponential Moving Average (EMA).\n */\nfunction forecastEMA(y: number[], horizon: number, alpha: number): MethodResult {\n  const n = y.length;\n  let ema = y[0];\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  for (let i = 1; i < n; i++) {\n    fitted[i] = ema; // one-step forecast is the previous EMA\n    ema = alpha * y[i] + (1 - alpha) * ema;\n  }\n\n  return { yValues: new Array(horizon).fill(ema), fitted, accumulates: false };\n}\n\n/**\n * Linear Trend Projection.\n * Fits y = mx + b to the historical data.\n */\nfunction forecastLinear(x: number[], y: number[], horizon: number): MethodResult {\n  const n = x.length;\n  let sumX = 0,\n    sumY = 0,\n    sumXY = 0,\n    sumXX = 0;\n\n  for (let i = 0; i < n; i++) {\n    sumX += x[i];\n    sumY += y[i];\n    sumXY += x[i] * y[i];\n    sumXX += x[i] * x[i];\n  }\n\n  const denom = n * sumXX - sumX * sumX;\n  const slope = denom !== 0 ? (n * sumXY - sumX * sumY) / denom : 0;\n  const intercept = (sumY - slope * sumX) / n;\n\n  const fitted: (number | null)[] = new Array(n);\n  for (let i = 0; i < n; i++) fitted[i] = slope * x[i] + intercept;\n\n  const lastX = x[n - 1];\n  const dt = n > 1 ? (x[n - 1] - x[0]) / (n - 1) : 1;\n  const yValues = new Array(horizon);\n  for (let i = 0; i < horizon; i++) {\n    const nextX = lastX + (i + 1) * dt;\n    yValues[i] = slope * nextX + intercept;\n  }\n\n  return { yValues, fitted, accumulates: true };\n}\n\n/**\n * Simple Exponential Smoothing (SES).\n */\nfunction forecastSimpleExpSmoothing(y: number[], horizon: number, alpha: number): MethodResult {\n  const n = y.length;\n  let level = y[0];\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  for (let i = 1; i < n; i++) {\n    fitted[i] = level;\n    level = alpha * y[i] + (1 - alpha) * level;\n  }\n\n  return { yValues: new Array(horizon).fill(level), fitted, accumulates: false };\n}\n\n/**\n * Holt's Linear Trend (Double Exponential Smoothing).\n */\nfunction forecastHolt(y: number[], horizon: number, alpha: number, beta: number): MethodResult {\n  const n = y.length;\n  if (n < 2) {\n    return { yValues: new Array(horizon).fill(y[0] || 0), fitted: [], accumulates: true };\n  }\n\n  let level = y[0];\n  let trend = y[1] - y[0];\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  for (let i = 1; i < n; i++) {\n    fitted[i] = level + trend; // one-step forecast\n    const lastLevel = level;\n    level = alpha * y[i] + (1 - alpha) * (level + trend);\n    trend = beta * (level - lastLevel) + (1 - beta) * trend;\n  }\n\n  const yValues = new Array(horizon);\n  for (let i = 0; i < horizon; i++) yValues[i] = level + (i + 1) * trend;\n\n  return { yValues, fitted, accumulates: true };\n}\n\n/**\n * Holt-Winters (Triple Exponential Smoothing, additive seasonality).\n */\nfunction forecastHoltWinters(\n  y: number[],\n  horizon: number,\n  period: number,\n  alpha: number,\n  beta: number,\n  gamma: number\n): MethodResult {\n  const n = y.length;\n  if (n < period * 2) {\n    // Fallback to Holt if not enough data for seasonality\n    return forecastHolt(y, horizon, alpha, beta);\n  }\n\n  // 1. Initial Seasonality\n  const seasonals = new Array(period);\n  const seasonAverages = new Array(Math.floor(n / period));\n  for (let i = 0; i < seasonAverages.length; i++) {\n    let sum = 0;\n    for (let j = 0; j < period; j++) sum += y[i * period + j];\n    seasonAverages[i] = sum / period;\n  }\n\n  for (let i = 0; i < period; i++) {\n    let sumOverAvg = 0;\n    for (let j = 0; j < seasonAverages.length; j++) {\n      sumOverAvg += y[j * period + i] - seasonAverages[j];\n    }\n    seasonals[i] = sumOverAvg / seasonAverages.length;\n  }\n\n  // 2. Initial Level and Trend\n  let level = seasonAverages[0];\n  let trend =\n    (seasonAverages[seasonAverages.length - 1] - seasonAverages[0]) /\n    ((seasonAverages.length - 1) * period);\n\n  const fitted: (number | null)[] = new Array(n).fill(null);\n\n  // 3. Update coefficients\n  for (let i = 0; i < n; i++) {\n    const seasonal = seasonals[i % period];\n    if (i > 0) fitted[i] = level + trend + seasonal;\n    const lastLevel = level;\n    const val = y[i];\n    level = alpha * (val - seasonal) + (1 - alpha) * (level + trend);\n    trend = beta * (level - lastLevel) + (1 - beta) * trend;\n    seasonals[i % period] = gamma * (val - level) + (1 - gamma) * seasonal;\n  }\n\n  // 4. Forecast\n  const yValues = new Array(horizon);\n  for (let i = 0; i < horizon; i++) {\n    const m = i + 1;\n    yValues[i] = level + m * trend + seasonals[(n + i) % period];\n  }\n\n  return { yValues, fitted, accumulates: true };\n}\n\n// ===========================================================================\n// ARIMA(p, d, q)\n// ===========================================================================\n\n/**\n * Simple native ARIMA implementation.\n *\n * - Differences the series `d` times to achieve (approximate) stationarity.\n * - Estimates AR(p) and MA(q) coefficients using the two-stage\n *   Hannan-Rissanen procedure (long AR to recover innovations, then OLS on\n *   lagged values and lagged innovations).\n * - Forecasts recursively on the differenced scale, then integrates back.\n *\n * Falls back to Holt's linear trend when there is not enough data to fit the\n * requested orders, so the method never throws for valid input.\n */\nfunction forecastARIMA(\n  y: number[],\n  horizon: number,\n  p: number,\n  d: number,\n  q: number\n): MethodResult {\n  const n = y.length;\n  const order = Math.max(p, q);\n  // Need enough observations to difference and fit; otherwise degrade gracefully.\n  if (n < d + order + 2 || n < 4) {\n    return forecastHolt(y, horizon, 0.3, 0.1);\n  }\n\n  // 1. Difference d times, keeping every intermediate level for integration.\n  const diffLevels: number[][] = [y.slice()];\n  for (let k = 0; k < d; k++) {\n    diffLevels.push(difference(diffLevels[k]));\n  }\n  const w = diffLevels[d]; // fully differenced (stationary) series\n  const m = w.length;\n\n  // Center the differenced series around its mean.\n  const mean = w.reduce((s, v) => s + v, 0) / m;\n  const wc = w.map((v) => v - mean);\n\n  // 2. Stage 1 - recover innovation (residual) estimates via a long AR fit.\n  let innovations = new Array(m).fill(0);\n  if (q > 0) {\n    const arLong = Math.min(Math.max(p + q + 2, 5), Math.floor(m / 2));\n    const { residuals } = fitAR(wc, arLong);\n    innovations = residuals;\n  }\n\n  // 3. Stage 2 - regress w_t on its p lags and q lagged innovations.\n  const start = Math.max(p, q);\n  const rows: number[][] = [];\n  const targets: number[] = [];\n  for (let t = start; t < m; t++) {\n    const row: number[] = [];\n    for (let i = 1; i <= p; i++) row.push(wc[t - i]);\n    for (let j = 1; j <= q; j++) row.push(innovations[t - j]);\n    rows.push(row);\n    targets.push(wc[t]);\n  }\n\n  let coeffs: number[] = [];\n  if (rows.length > 0 && rows[0].length > 0) {\n    coeffs = olsSolve(rows, targets);\n  }\n  if (coeffs.length !== p + q || coeffs.some((c) => !Number.isFinite(c))) {\n    // Fitting failed (singular/ill-conditioned) - fall back.\n    return forecastHolt(y, horizon, 0.3, 0.1);\n  }\n  const ar = coeffs.slice(0, p);\n  const ma = coeffs.slice(p, p + q);\n\n  // 4. In-sample one-step fitted values (on original scale) for residual sigma.\n  const fitted: (number | null)[] = new Array(n).fill(null);\n  const fittedResiduals = new Array(m).fill(0);\n  for (let t = start; t < m; t++) {\n    let pred = 0;\n    for (let i = 1; i <= p; i++) pred += ar[i - 1] * wc[t - i];\n    for (let j = 1; j <= q; j++) pred += ma[j - 1] * fittedResiduals[t - j];\n    fittedResiduals[t] = wc[t] - pred;\n    // Map differenced fitted value back to the original scale by adding the\n    // previous original observation chain. For d>=1 we approximate by adding\n    // the last non-differenced value; exact integration below handles forecasts.\n    fitted[t + (n - m)] = integrateFittedPoint(diffLevels, d, t, pred + mean);\n  }\n\n  // 5. Recursive forecast on the differenced/centered scale.\n  const wHist = wc.slice();\n  const eHist = fittedResiduals.slice();\n  const forecastDiff: number[] = [];\n  for (let h = 0; h < horizon; h++) {\n    let pred = 0;\n    // wHist/eHist always have at least `order` entries here, so the lag index\n    // is always in range.\n    for (let i = 1; i <= p; i++) {\n      pred += ar[i - 1] * wHist[wHist.length - i];\n    }\n    for (let j = 1; j <= q; j++) {\n      pred += ma[j - 1] * eHist[eHist.length - j];\n    }\n    wHist.push(pred);\n    eHist.push(0); // expected future innovation is zero\n    forecastDiff.push(pred + mean);\n  }\n\n  // 6. Integrate the forecast back d times to the original scale.\n  const yValues = integrateForecast(diffLevels, d, forecastDiff);\n\n  return { yValues, fitted, accumulates: true };\n}\n\n/** First difference of a series. */\nfunction difference(series: number[]): number[] {\n  const out = new Array(series.length - 1);\n  for (let i = 1; i < series.length; i++) out[i - 1] = series[i] - series[i - 1];\n  return out;\n}\n\n/**\n * Integrate a differenced forecast back to the original scale.\n * `diffLevels[0]` is the original series, `diffLevels[d]` fully differenced.\n */\nfunction integrateForecast(diffLevels: number[][], d: number, forecastDiff: number[]): number[] {\n  let current = forecastDiff.slice();\n  for (let k = d; k >= 1; k--) {\n    const parent = diffLevels[k - 1];\n    let last = parent[parent.length - 1];\n    const integrated = new Array(current.length);\n    for (let i = 0; i < current.length; i++) {\n      last += current[i];\n      integrated[i] = last;\n    }\n    current = integrated;\n  }\n  return current;\n}\n\n/**\n * Best-effort mapping of a single in-sample differenced prediction back to the\n * original scale (used only for residual-sigma estimation).\n */\nfunction integrateFittedPoint(\n  diffLevels: number[][],\n  d: number,\n  t: number,\n  diffPred: number\n): number {\n  if (d === 0) return diffPred;\n  // Add the previous observed original value chain.\n  const parent = diffLevels[d - 1];\n  // parent is one longer than the differenced series, so index t is in range.\n  const base = parent[t];\n  // For d>1 this is an approximation; sufficient for a residual scale estimate.\n  return base + diffPred;\n}\n\n/**\n * Fit an AR(order) model via ordinary least squares and return coefficients\n * plus the in-sample residuals aligned to the input series (leading entries 0).\n */\nfunction fitAR(series: number[], order: number): { coeffs: number[]; residuals: number[] } {\n  const m = series.length;\n  const rows: number[][] = [];\n  const targets: number[] = [];\n  for (let t = order; t < m; t++) {\n    const row: number[] = [];\n    for (let i = 1; i <= order; i++) row.push(series[t - i]);\n    rows.push(row);\n    targets.push(series[t]);\n  }\n  const coeffs = rows.length > 0 ? olsSolve(rows, targets) : new Array(order).fill(0);\n  const residuals = new Array(m).fill(0);\n  for (let t = order; t < m; t++) {\n    let pred = 0;\n    for (let i = 1; i <= order; i++) pred += (coeffs[i - 1] || 0) * series[t - i];\n    residuals[t] = series[t] - pred;\n  }\n  return { coeffs, residuals };\n}\n\n/**\n * Solve a linear least-squares problem `X b = y` via the normal equations\n * with Gaussian elimination. Returns [] on a singular system.\n */\nfunction olsSolve(X: number[][], y: number[]): number[] {\n  const rows = X.length;\n  const cols = X[0].length;\n\n  // Normal equations: (XᵀX) b = Xᵀy\n  const xtx: number[][] = Array.from({ length: cols }, () => new Array(cols).fill(0));\n  const xty: number[] = new Array(cols).fill(0);\n\n  for (let i = 0; i < rows; i++) {\n    for (let a = 0; a < cols; a++) {\n      xty[a] += X[i][a] * y[i];\n      for (let b = 0; b < cols; b++) {\n        xtx[a][b] += X[i][a] * X[i][b];\n      }\n    }\n  }\n\n  // Tiny ridge term to stabilise near-singular systems.\n  for (let a = 0; a < cols; a++) xtx[a][a] += 1e-8;\n\n  return gaussianSolve(xtx, xty);\n}\n\n/**\n * Solve `A x = b` for a square matrix via Gaussian elimination with partial\n * pivoting. Returns [] if the matrix is singular.\n */\nexport function gaussianSolve(A: number[][], b: number[]): number[] {\n  const n = b.length;\n  const M = A.map((row, i) => [...row, b[i]]);\n\n  for (let col = 0; col < n; col++) {\n    // Partial pivot\n    let pivot = col;\n    for (let r = col + 1; r < n; r++) {\n      if (Math.abs(M[r][col]) > Math.abs(M[pivot][col])) pivot = r;\n    }\n    if (Math.abs(M[pivot][col]) < 1e-12) return [];\n    [M[col], M[pivot]] = [M[pivot], M[col]];\n\n    // Eliminate\n    for (let r = 0; r < n; r++) {\n      if (r === col) continue;\n      const factor = M[r][col] / M[col][col];\n      for (let c = col; c <= n; c++) M[r][c] -= factor * M[col][c];\n    }\n  }\n\n  const x = new Array(n);\n  for (let i = 0; i < n; i++) x[i] = M[i][n] / M[i][i];\n  return x;\n}\n\n/**\n * Compute error metrics from in-sample one-step residuals.\n * `rmse` doubles as the residual sigma used to build confidence bands.\n */\nfunction calculateFitMetrics(\n  historical: number[],\n  fitted?: (number | null)[]\n): { mse: number; rmse: number; mae: number; r2: number } {\n  if (!fitted || fitted.length === 0) {\n    return { mse: 0, rmse: 0, mae: 0, r2: 0 };\n  }\n\n  let sse = 0;\n  let sae = 0;\n  let count = 0;\n  const observed: number[] = [];\n\n  for (let i = 0; i < fitted.length && i < historical.length; i++) {\n    const f = fitted[i];\n    if (f === null || f === undefined || !Number.isFinite(f)) continue;\n    const err = historical[i] - f;\n    sse += err * err;\n    sae += Math.abs(err);\n    observed.push(historical[i]);\n    count++;\n  }\n\n  if (count === 0) return { mse: 0, rmse: 0, mae: 0, r2: 0 };\n\n  const mse = sse / count;\n  const rmse = Math.sqrt(mse);\n  const mae = sae / count;\n\n  const mean = observed.reduce((s, v) => s + v, 0) / observed.length;\n  const sst = observed.reduce((s, v) => s + (v - mean) * (v - mean), 0);\n  const r2 = sst > 0 ? 1 - sse / sst : 0;\n\n  return {\n    mse: round(mse),\n    rmse: round(rmse),\n    mae: round(mae),\n    r2: round(r2),\n  };\n}\n\nfunction round(v: number): number {\n  return Math.round(v * 1e6) / 1e6;\n}\n","/**\n * Velo Plot - Forecasting Plugin\n * \n * Provides time series forecasting capabilities using native statistical models.\n */\n\nimport type { \n  PluginForecastingConfig, \n  ForecastingAPI, \n  ForecastingOptions, \n  ForecastingResult, \n  ForecastingVisualizationConfig \n} from './types';\nimport { calculateForecast } from './algorithms';\nimport { exportForecastOverlay } from '../../core/chart/exporter/svg/plugins/regression';\nimport type { \n  PluginManifest, \n  ChartPlugin, \n  PluginContext,\n  AfterRenderEvent\n} from '../types';\n\nconst manifestForecasting: PluginManifest = {\n  name: \"velo-plot-forecasting\",\n  version: \"1.0.0\",\n  description: \"Time series forecasting for velo-plot\",\n  provides: [\"analysis\", \"forecasting\"],\n  tags: [\"statistics\", \"forecasting\", \"analysis\"],\n};\n\n/**\n * Forecasting Plugin for Velo Plot\n */\nexport function PluginForecasting(\n  config: Partial<PluginForecastingConfig> = {}\n): ChartPlugin<PluginForecastingConfig> {\n  let ctx: PluginContext | null = null;\n  const activeForecasts = new Map<string, { result: ForecastingResult, config: ForecastingVisualizationConfig }>();\n\n  const api: ForecastingAPI = {\n    forecast(data: any, options: ForecastingOptions): ForecastingResult {\n      let x: any, y: any;\n\n      // Allow `forecast('seriesId', options)` by resolving the series data.\n      if (typeof data === 'string') {\n        if (!ctx) throw new Error('Plugin not initialized');\n        const series = ctx.data.getAllSeries().find(s => s.getId() === data);\n        if (!series) throw new Error(`Series not found: ${data}`);\n        const seriesData = series.getData();\n        x = seriesData.x;\n        y = seriesData.y;\n      } else if (Array.isArray(data)) {\n        y = data;\n        x = data.map((_, i) => i);\n      } else if (data.x && data.y) {\n        x = data.x;\n        y = data.y;\n      } else {\n        throw new Error(\"Invalid data format for forecasting\");\n      }\n\n      return calculateForecast(x, y, options.method, options.horizon, options.params, options.confidence);\n    },\n\n    async forecastSeries(seriesId: string, options: ForecastingOptions): Promise<ForecastingResult> {\n      if (!ctx) throw new Error(\"Plugin not initialized\");\n      \n      const series = ctx.data.getAllSeries().find(s => s.getId() === seriesId);\n      if (!series) throw new Error(`Series not found: ${seriesId}`);\n      \n      const data = series.getData();\n      return this.forecast(data, options);\n    },\n\n    visualize(result: ForecastingResult, vizConfig: ForecastingVisualizationConfig = {}) {\n      const id = `forecast-${Math.random().toString(36).substr(2, 9)}`;\n      const mergedConfig = {\n        showLine: true,\n        showConfidenceInterval: true,\n        ...config.defaultVisualization,\n        ...vizConfig\n      };\n      \n      activeForecasts.set(id, { result, config: mergedConfig });\n      ctx?.requestRender();\n      return id;\n    },\n\n    clear(id?: string) {\n      if (id) {\n        activeForecasts.delete(id);\n      } else {\n        activeForecasts.clear();\n      }\n      ctx?.requestRender();\n    }\n  };\n\n  function drawForecasts(pCtx: PluginContext) {\n    const { render, coords } = pCtx;\n    const { ctx2d } = render;\n    if (!ctx2d) return;\n\n    activeForecasts.forEach((viz) => {\n      const { result, config: vizConfig } = viz;\n      const { xValues, yValues, lowerBound, upperBound } = result;\n\n      ctx2d.save();\n\n      // 1. Draw Confidence Interval\n      if (vizConfig.showConfidenceInterval && lowerBound && upperBound) {\n        ctx2d.beginPath();\n        const opacity = vizConfig.intervalStyle?.opacity ?? 0.15;\n        const color = vizConfig.intervalStyle?.fillColor || (vizConfig.lineStyle?.color || '#3b82f6');\n        \n        ctx2d.fillStyle = color.startsWith('rgba') ? color : color + Math.floor(opacity * 255).toString(16).padStart(2, '0');\n        \n        for (let i = 0; i < xValues.length; i++) {\n          const px = coords.dataToPixelX(xValues[i]);\n          const py = coords.dataToPixelY(upperBound[i]);\n          if (i === 0) ctx2d.moveTo(px, py);\n          else ctx2d.lineTo(px, py);\n        }\n        \n        for (let i = xValues.length - 1; i >= 0; i--) {\n          const px = coords.dataToPixelX(xValues[i]);\n          const py = coords.dataToPixelY(lowerBound[i]);\n          ctx2d.lineTo(px, py);\n        }\n        \n        ctx2d.closePath();\n        ctx2d.fill();\n      }\n\n      // 2. Draw Forecast Line\n      if (vizConfig.showLine) {\n        ctx2d.beginPath();\n        ctx2d.strokeStyle = vizConfig.lineStyle?.color || '#3b82f6';\n        ctx2d.lineWidth = vizConfig.lineStyle?.width || 2;\n        if (vizConfig.lineStyle?.dash) {\n          ctx2d.setLineDash(vizConfig.lineStyle.dash);\n        } else {\n          ctx2d.setLineDash([5, 5]); // Default dashed for forecast\n        }\n\n        for (let i = 0; i < xValues.length; i++) {\n          const px = coords.dataToPixelX(xValues[i]);\n          const py = coords.dataToPixelY(yValues[i]);\n          if (i === 0) ctx2d.moveTo(px, py);\n          else ctx2d.lineTo(px, py);\n        }\n        ctx2d.stroke();\n      }\n\n      ctx2d.restore();\n    });\n  }\n\n  return {\n    manifest: manifestForecasting,\n    onInit(c: PluginContext) {\n      ctx = c;\n    },\n    onRenderOverlay(pCtx: PluginContext, _event: AfterRenderEvent) {\n      drawForecasts(pCtx);\n    },\n    onExportSVG(svgCtx) {\n      if (!svgCtx.builder || svgCtx.exportContext?.options.includeOverlays === false) return;\n      activeForecasts.forEach(({ result, config: vizConfig }) => {\n        const { xValues, yValues, lowerBound, upperBound } = result;\n        const n = xValues.length;\n        const points: Array<{ x: number; y: number }> = [];\n        for (let i = 0; i < n; i++) {\n          points.push({ x: xValues[i], y: yValues[i] });\n        }\n        const ci =\n          vizConfig.showConfidenceInterval && lowerBound && upperBound\n            ? {\n                upper: Array.from({ length: n }, (_, i) => ({ x: xValues[i], y: upperBound[i] })),\n                lower: Array.from({ length: n }, (_, i) => ({ x: xValues[i], y: lowerBound[i] })),\n              }\n            : undefined;\n        exportForecastOverlay(svgCtx, points, ci, vizConfig.lineStyle?.color ?? \"#3b82f6\");\n      });\n    },\n    api\n  };\n}\n\n// Re-export types and constants\nexport * from './types';\nexport * from './algorithms';\nexport { manifestForecasting };\n"],"names":["toArray","a","zForConfidence","confidence","table","best","bestDiff","entry","diff","calculateForecast","x","y","method","horizon","params","n","ya","xa","dt","lastX","result","forecastSMA","forecastEMA","forecastLinear","forecastSimpleExpSmoothing","forecastHolt","forecastWMA","forecastHoltWinters","forecastARIMA","futureX","i","metrics","calculateFitMetrics","z","sigma","lower","upper","buildConfidenceBands","yValues","accumulates","growth","margin","window","effectiveWindow","wmaAt","end","w","weightedSum","weightTotal","weight","fitted","wma","smaAt","sum","avg","alpha","ema","sumX","sumY","sumXY","sumXX","denom","slope","intercept","nextX","level","beta","trend","lastLevel","period","gamma","seasonals","seasonAverages","j","sumOverAvg","seasonal","val","m","p","d","q","order","diffLevels","k","difference","mean","s","v","wc","innovations","arLong","residuals","fitAR","start","rows","targets","t","row","coeffs","olsSolve","c","ar","ma","fittedResiduals","pred","integrateFittedPoint","wHist","eHist","forecastDiff","h","integrateForecast","series","out","current","parent","last","integrated","diffPred","X","cols","xtx","xty","b","gaussianSolve","A","M","col","pivot","r","factor","historical","sse","sae","count","observed","err","mse","rmse","mae","sst","r2","round","manifestForecasting","PluginForecasting","config","ctx","activeForecasts","api","data","options","seriesData","_","seriesId","vizConfig","id","mergedConfig","drawForecasts","pCtx","render","coords","ctx2d","viz","xValues","lowerBound","upperBound","opacity","_a","color","_b","_c","px","py","_d","_e","_f","_event","svgCtx","points","ci","exportForecastOverlay"],"mappings":";AAwCA,SAASA,EAAQC,GAAqD;AACpE,SAAO,MAAM,KAAKA,CAAC;AACrB;AAKA,SAASC,EAAeC,GAA4B;AAClD,QAAMC,IAAiC;AAAA,IACrC,CAAC,KAAK,KAAK;AAAA,IACX,CAAC,MAAM,KAAK;AAAA,IACZ,CAAC,KAAK,KAAK;AAAA,IACX,CAAC,KAAK,KAAK;AAAA,IACX,CAAC,MAAM,IAAI;AAAA,IACX,CAAC,OAAO,KAAK;AAAA,IACb,CAAC,MAAM,KAAK;AAAA,IACZ,CAAC,OAAO,KAAK;AAAA,EAAA;AAEf,MAAIC,IAAOD,EAAM,CAAC,GACdE,IAAW;AACf,aAAWC,KAASH,GAAO;AACzB,UAAMI,IAAO,KAAK,IAAID,EAAM,CAAC,IAAIJ,CAAU;AAC3C,IAAIK,IAAOF,MACTA,IAAWE,GACXH,IAAOE;AAAA,EAEX;AACA,SAAOF,EAAK,CAAC;AACf;AAKO,SAASI,EACdC,GACAC,GACAC,GACAC,GACAC,IAA4B,CAAA,GAC5BX,IAAa,MACM;AACnB,QAAMY,IAAIL,EAAE;AACZ,MAAIK,MAAM,EAAG,OAAM,IAAI,MAAM,kBAAkB;AAE/C,QAAMC,IAAKhB,EAAQW,CAAC,GACdM,IAAKjB,EAAQU,CAAC,GAGdQ,IAAKH,IAAI,KAAKE,EAAGF,IAAI,CAAC,IAAIE,EAAG,CAAC,MAAMF,IAAI,KAAK,GAC7CI,IAAQF,EAAGF,IAAI,CAAC;AAEtB,MAAIK;AAEJ,UAAQR,GAAA;AAAA,IACN,KAAK;AACH,MAAAQ,IAASC,EAAYL,GAAIH,GAASC,EAAO,cAAc,EAAE;AACzD;AAAA,IACF,KAAK;AACH,MAAAM,IAASE,EAAYN,GAAIH,GAASC,EAAO,SAAS,GAAG;AACrD;AAAA,IACF,KAAK;AACH,MAAAM,IAASG,EAAeN,GAAID,GAAIH,CAAO;AACvC;AAAA,IACF,KAAK;AACH,MAAAO,IAASI,EAA2BR,GAAIH,GAASC,EAAO,SAAS,GAAG;AACpE;AAAA,IACF,KAAK;AACH,MAAAM,IAASK,EAAaT,GAAIH,GAASC,EAAO,SAAS,KAAKA,EAAO,QAAQ,GAAG;AAC1E;AAAA,IACF,KAAK;AACH,MAAAM,IAASM,EAAYV,GAAIH,GAASC,EAAO,cAAc,EAAE;AACzD;AAAA,IACF,KAAK;AACH,MAAAM,IAASO;AAAA,QACPX;AAAA,QACAH;AAAA,QACAC,EAAO,UAAU;AAAA,QACjBA,EAAO,SAAS;AAAA,QAChBA,EAAO,QAAQ;AAAA,QACfA,EAAO,SAAS;AAAA,MAAA;AAElB;AAAA,IACF,KAAK;AACH,MAAAM,IAASQ;AAAA,QACPZ;AAAA,QACAH;AAAA,QACAC,EAAO,KAAK;AAAA,QACZA,EAAO,KAAK;AAAA,QACZA,EAAO,KAAK;AAAA,MAAA;AAEd;AAAA,IACF;AACE,YAAM,IAAI,MAAM,UAAUF,CAAM,sBAAsB;AAAA,EAAA;AAI1D,QAAMiB,IAAU,IAAI,MAAMhB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB;AAC3B,IAAAD,EAAQC,CAAC,IAAIX,KAASW,IAAI,KAAKZ;AAIjC,QAAMa,IAAUC,EAAoBhB,GAAII,EAAO,MAAM,GAC/Ca,IAAI/B,EAAeC,CAAU,GAC7B+B,IAAQH,EAAQ,MAChB,EAAE,OAAAI,GAAO,OAAAC,EAAA,IAAUC;AAAA,IACvBjB,EAAO;AAAA,IACPc;AAAA,IACAD;AAAA,IACAb,EAAO;AAAA,EAAA;AAGT,SAAO;AAAA,IACL,SAASS;AAAA,IACT,SAAST,EAAO;AAAA,IAChB,YAAYe;AAAA,IACZ,YAAYC;AAAA,IACZ,QAAAxB;AAAA,IACA,UAAU,EAAE,GAAGmB,GAAS,YAAA5B,EAAA;AAAA,EAAW;AAEvC;AAOA,SAASkC,EACPC,GACAJ,GACAD,GACAM,GACsC;AACtC,QAAMJ,IAAQ,IAAI,MAAMG,EAAQ,MAAM,GAChCF,IAAQ,IAAI,MAAME,EAAQ,MAAM;AACtC,WAASR,IAAI,GAAGA,IAAIQ,EAAQ,QAAQR,KAAK;AACvC,UAAMU,IAASD,IAAc,KAAK,KAAKT,IAAI,CAAC,IAAI,GAC1CW,IAASR,IAAIC,IAAQM;AAC3B,IAAAL,EAAML,CAAC,IAAIQ,EAAQR,CAAC,IAAIW,GACxBL,EAAMN,CAAC,IAAIQ,EAAQR,CAAC,IAAIW;AAAA,EAC1B;AACA,SAAO,EAAE,OAAAN,GAAO,OAAAC,EAAA;AAClB;AAUA,SAASV,EAAYf,GAAaE,GAAiB6B,GAA8B;AAC/E,QAAM,IAAI/B,EAAE,QACNgC,IAAkB,KAAK,IAAID,GAAQ,CAAC,GAEpCE,IAAQ,CAACC,MAAwB;AAErC,UAAMC,IAAI,KAAK,IAAIH,GAAiBE,CAAG;AACvC,QAAIE,IAAc,GACdC,IAAc;AAClB,aAASlB,IAAI,GAAGA,IAAIgB,GAAGhB,KAAK;AAC1B,YAAMmB,IAASnB,IAAI;AACnB,MAAAiB,KAAepC,EAAEkC,IAAMC,IAAIhB,CAAC,IAAImB,GAChCD,KAAeC;AAAA,IACjB;AACA,WAAOF,IAAcC;AAAA,EACvB,GAEME,IAA4B,IAAI,MAAM,CAAC,EAAE,KAAK,IAAI;AACxD,WAASpB,IAAI,GAAGA,IAAI,GAAGA,IAAK,CAAAoB,EAAOpB,CAAC,IAAIc,EAAMd,CAAC;AAE/C,QAAMqB,IAAMP,EAAM,CAAC;AACnB,SAAO,EAAE,SAAS,IAAI,MAAM/B,CAAO,EAAE,KAAKsC,CAAG,GAAG,QAAAD,GAAQ,aAAa,GAAA;AACvE;AAMA,SAAS7B,EAAYV,GAAaE,GAAiB6B,GAA8B;AAC/E,QAAM,IAAI/B,EAAE,QACNgC,IAAkB,KAAK,IAAID,GAAQ,CAAC,GAEpCU,IAAQ,CAACP,MAAwB;AAErC,UAAMC,IAAI,KAAK,IAAIH,GAAiBE,CAAG;AACvC,QAAIQ,IAAM;AACV,aAASvB,IAAIe,IAAMC,GAAGhB,IAAIe,GAAKf,IAAK,CAAAuB,KAAO1C,EAAEmB,CAAC;AAC9C,WAAOuB,IAAMP;AAAA,EACf,GAEMI,IAA4B,IAAI,MAAM,CAAC,EAAE,KAAK,IAAI;AACxD,WAASpB,IAAI,GAAGA,IAAI,GAAGA,IAAK,CAAAoB,EAAOpB,CAAC,IAAIsB,EAAMtB,CAAC;AAE/C,QAAMwB,IAAMF,EAAM,CAAC;AACnB,SAAO,EAAE,SAAS,IAAI,MAAMvC,CAAO,EAAE,KAAKyC,CAAG,GAAG,QAAAJ,GAAQ,aAAa,GAAA;AACvE;AAKA,SAAS5B,EAAYX,GAAaE,GAAiB0C,GAA6B;AAC9E,QAAM,IAAI5C,EAAE;AACZ,MAAI6C,IAAM7C,EAAE,CAAC;AACb,QAAMuC,IAA4B,IAAI,MAAM,CAAC,EAAE,KAAK,IAAI;AAExD,WAASpB,IAAI,GAAGA,IAAI,GAAGA;AACrB,IAAAoB,EAAOpB,CAAC,IAAI0B,GACZA,IAAMD,IAAQ5C,EAAEmB,CAAC,KAAK,IAAIyB,KAASC;AAGrC,SAAO,EAAE,SAAS,IAAI,MAAM3C,CAAO,EAAE,KAAK2C,CAAG,GAAG,QAAAN,GAAQ,aAAa,GAAA;AACvE;AAMA,SAAS3B,EAAeb,GAAaC,GAAaE,GAA+B;AAC/E,QAAM,IAAIH,EAAE;AACZ,MAAI+C,IAAO,GACTC,IAAO,GACPC,IAAQ,GACRC,IAAQ;AAEV,WAAS9B,IAAI,GAAGA,IAAI,GAAGA;AACrB,IAAA2B,KAAQ/C,EAAEoB,CAAC,GACX4B,KAAQ/C,EAAEmB,CAAC,GACX6B,KAASjD,EAAEoB,CAAC,IAAInB,EAAEmB,CAAC,GACnB8B,KAASlD,EAAEoB,CAAC,IAAIpB,EAAEoB,CAAC;AAGrB,QAAM+B,IAAQ,IAAID,IAAQH,IAAOA,GAC3BK,IAAQD,MAAU,KAAK,IAAIF,IAAQF,IAAOC,KAAQG,IAAQ,GAC1DE,KAAaL,IAAOI,IAAQL,KAAQ,GAEpCP,IAA4B,IAAI,MAAM,CAAC;AAC7C,WAASpB,IAAI,GAAGA,IAAI,GAAGA,IAAK,CAAAoB,EAAOpB,CAAC,IAAIgC,IAAQpD,EAAEoB,CAAC,IAAIiC;AAEvD,QAAM5C,IAAQT,EAAE,IAAI,CAAC,GACfQ,IAAK,IAAI,KAAKR,EAAE,IAAI,CAAC,IAAIA,EAAE,CAAC,MAAM,IAAI,KAAK,GAC3C4B,IAAU,IAAI,MAAMzB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB,KAAK;AAChC,UAAMkC,IAAQ7C,KAASW,IAAI,KAAKZ;AAChC,IAAAoB,EAAQR,CAAC,IAAIgC,IAAQE,IAAQD;AAAA,EAC/B;AAEA,SAAO,EAAE,SAAAzB,GAAS,QAAAY,GAAQ,aAAa,GAAA;AACzC;AAKA,SAAS1B,EAA2Bb,GAAaE,GAAiB0C,GAA6B;AAC7F,QAAM,IAAI5C,EAAE;AACZ,MAAIsD,IAAQtD,EAAE,CAAC;AACf,QAAMuC,IAA4B,IAAI,MAAM,CAAC,EAAE,KAAK,IAAI;AAExD,WAASpB,IAAI,GAAGA,IAAI,GAAGA;AACrB,IAAAoB,EAAOpB,CAAC,IAAImC,GACZA,IAAQV,IAAQ5C,EAAEmB,CAAC,KAAK,IAAIyB,KAASU;AAGvC,SAAO,EAAE,SAAS,IAAI,MAAMpD,CAAO,EAAE,KAAKoD,CAAK,GAAG,QAAAf,GAAQ,aAAa,GAAA;AACzE;AAKA,SAASzB,EAAad,GAAaE,GAAiB0C,GAAeW,GAA4B;AAC7F,QAAMnD,IAAIJ,EAAE;AACZ,MAAII,IAAI;AACN,WAAO,EAAE,SAAS,IAAI,MAAMF,CAAO,EAAE,KAAKF,EAAE,CAAC,KAAK,CAAC,GAAG,QAAQ,CAAA,GAAI,aAAa,GAAA;AAGjF,MAAIsD,IAAQtD,EAAE,CAAC,GACXwD,IAAQxD,EAAE,CAAC,IAAIA,EAAE,CAAC;AACtB,QAAMuC,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AAExD,WAASe,IAAI,GAAGA,IAAIf,GAAGe,KAAK;AAC1B,IAAAoB,EAAOpB,CAAC,IAAImC,IAAQE;AACpB,UAAMC,IAAYH;AAClB,IAAAA,IAAQV,IAAQ5C,EAAEmB,CAAC,KAAK,IAAIyB,MAAUU,IAAQE,IAC9CA,IAAQD,KAAQD,IAAQG,MAAc,IAAIF,KAAQC;AAAA,EACpD;AAEA,QAAM7B,IAAU,IAAI,MAAMzB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB,OAAaA,CAAC,IAAImC,KAASnC,IAAI,KAAKqC;AAEjE,SAAO,EAAE,SAAA7B,GAAS,QAAAY,GAAQ,aAAa,GAAA;AACzC;AAKA,SAASvB,EACPhB,GACAE,GACAwD,GACAd,GACAW,GACAI,GACc;AACd,QAAMvD,IAAIJ,EAAE;AACZ,MAAII,IAAIsD,IAAS;AAEf,WAAO5C,EAAad,GAAGE,GAAS0C,GAAOW,CAAI;AAI7C,QAAMK,IAAY,IAAI,MAAMF,CAAM,GAC5BG,IAAiB,IAAI,MAAM,KAAK,MAAMzD,IAAIsD,CAAM,CAAC;AACvD,WAASvC,IAAI,GAAGA,IAAI0C,EAAe,QAAQ1C,KAAK;AAC9C,QAAIuB,IAAM;AACV,aAASoB,IAAI,GAAGA,IAAIJ,GAAQI,IAAK,CAAApB,KAAO1C,EAAEmB,IAAIuC,IAASI,CAAC;AACxD,IAAAD,EAAe1C,CAAC,IAAIuB,IAAMgB;AAAA,EAC5B;AAEA,WAASvC,IAAI,GAAGA,IAAIuC,GAAQvC,KAAK;AAC/B,QAAI4C,IAAa;AACjB,aAASD,IAAI,GAAGA,IAAID,EAAe,QAAQC;AACzC,MAAAC,KAAc/D,EAAE8D,IAAIJ,IAASvC,CAAC,IAAI0C,EAAeC,CAAC;AAEpD,IAAAF,EAAUzC,CAAC,IAAI4C,IAAaF,EAAe;AAAA,EAC7C;AAGA,MAAIP,IAAQO,EAAe,CAAC,GACxBL,KACDK,EAAeA,EAAe,SAAS,CAAC,IAAIA,EAAe,CAAC,OAC3DA,EAAe,SAAS,KAAKH;AAEjC,QAAMnB,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI;AAGxD,WAASe,IAAI,GAAGA,IAAIf,GAAGe,KAAK;AAC1B,UAAM6C,IAAWJ,EAAUzC,IAAIuC,CAAM;AACrC,IAAIvC,IAAI,MAAGoB,EAAOpB,CAAC,IAAImC,IAAQE,IAAQQ;AACvC,UAAMP,IAAYH,GACZW,IAAMjE,EAAEmB,CAAC;AACf,IAAAmC,IAAQV,KAASqB,IAAMD,MAAa,IAAIpB,MAAUU,IAAQE,IAC1DA,IAAQD,KAAQD,IAAQG,MAAc,IAAIF,KAAQC,GAClDI,EAAUzC,IAAIuC,CAAM,IAAIC,KAASM,IAAMX,MAAU,IAAIK,KAASK;AAAA,EAChE;AAGA,QAAMrC,IAAU,IAAI,MAAMzB,CAAO;AACjC,WAASiB,IAAI,GAAGA,IAAIjB,GAASiB,KAAK;AAChC,UAAM+C,IAAI/C,IAAI;AACd,IAAAQ,EAAQR,CAAC,IAAImC,IAAQY,IAAIV,IAAQI,GAAWxD,IAAIe,KAAKuC,CAAM;AAAA,EAC7D;AAEA,SAAO,EAAE,SAAA/B,GAAS,QAAAY,GAAQ,aAAa,GAAA;AACzC;AAkBA,SAAStB,EACPjB,GACAE,GACAiE,GACAC,GACAC,GACc;AACd,QAAMjE,IAAIJ,EAAE,QACNsE,IAAQ,KAAK,IAAIH,GAAGE,CAAC;AAE3B,MAAIjE,IAAIgE,IAAIE,IAAQ,KAAKlE,IAAI;AAC3B,WAAOU,EAAad,GAAGE,GAAS,KAAK,GAAG;AAI1C,QAAMqE,IAAyB,CAACvE,EAAE,OAAO;AACzC,WAASwE,IAAI,GAAGA,IAAIJ,GAAGI;AACrB,IAAAD,EAAW,KAAKE,EAAWF,EAAWC,CAAC,CAAC,CAAC;AAE3C,QAAMrC,IAAIoC,EAAWH,CAAC,GAChBF,IAAI/B,EAAE,QAGNuC,IAAOvC,EAAE,OAAO,CAACwC,GAAGC,MAAMD,IAAIC,GAAG,CAAC,IAAIV,GACtCW,IAAK1C,EAAE,IAAI,CAACyC,MAAMA,IAAIF,CAAI;AAGhC,MAAII,IAAc,IAAI,MAAMZ,CAAC,EAAE,KAAK,CAAC;AACrC,MAAIG,IAAI,GAAG;AACT,UAAMU,IAAS,KAAK,IAAI,KAAK,IAAIZ,IAAIE,IAAI,GAAG,CAAC,GAAG,KAAK,MAAMH,IAAI,CAAC,CAAC,GAC3D,EAAE,WAAAc,EAAA,IAAcC,EAAMJ,GAAIE,CAAM;AACtC,IAAAD,IAAcE;AAAA,EAChB;AAGA,QAAME,IAAQ,KAAK,IAAIf,GAAGE,CAAC,GACrBc,IAAmB,CAAA,GACnBC,IAAoB,CAAA;AAC1B,WAASC,IAAIH,GAAOG,IAAInB,GAAGmB,KAAK;AAC9B,UAAMC,IAAgB,CAAA;AACtB,aAASnE,IAAI,GAAGA,KAAKgD,GAAGhD,OAAS,KAAK0D,EAAGQ,IAAIlE,CAAC,CAAC;AAC/C,aAAS2C,IAAI,GAAGA,KAAKO,GAAGP,OAAS,KAAKgB,EAAYO,IAAIvB,CAAC,CAAC;AACxD,IAAAqB,EAAK,KAAKG,CAAG,GACbF,EAAQ,KAAKP,EAAGQ,CAAC,CAAC;AAAA,EACpB;AAEA,MAAIE,IAAmB,CAAA;AAIvB,MAHIJ,EAAK,SAAS,KAAKA,EAAK,CAAC,EAAE,SAAS,MACtCI,IAASC,EAASL,GAAMC,CAAO,IAE7BG,EAAO,WAAWpB,IAAIE,KAAKkB,EAAO,KAAK,CAACE,MAAM,CAAC,OAAO,SAASA,CAAC,CAAC;AAEnE,WAAO3E,EAAad,GAAGE,GAAS,KAAK,GAAG;AAE1C,QAAMwF,IAAKH,EAAO,MAAM,GAAGpB,CAAC,GACtBwB,IAAKJ,EAAO,MAAMpB,GAAGA,IAAIE,CAAC,GAG1B9B,IAA4B,IAAI,MAAMnC,CAAC,EAAE,KAAK,IAAI,GAClDwF,IAAkB,IAAI,MAAM1B,CAAC,EAAE,KAAK,CAAC;AAC3C,WAASmB,IAAIH,GAAOG,IAAInB,GAAGmB,KAAK;AAC9B,QAAIQ,IAAO;AACX,aAAS1E,IAAI,GAAGA,KAAKgD,GAAGhD,IAAK,CAAA0E,KAAQH,EAAGvE,IAAI,CAAC,IAAI0D,EAAGQ,IAAIlE,CAAC;AACzD,aAAS2C,IAAI,GAAGA,KAAKO,GAAGP,IAAK,CAAA+B,KAAQF,EAAG7B,IAAI,CAAC,IAAI8B,EAAgBP,IAAIvB,CAAC;AACtE,IAAA8B,EAAgBP,CAAC,IAAIR,EAAGQ,CAAC,IAAIQ,GAI7BtD,EAAO8C,KAAKjF,IAAI8D,EAAE,IAAI4B,EAAqBvB,GAAYH,GAAGiB,GAAGQ,IAAOnB,CAAI;AAAA,EAC1E;AAGA,QAAMqB,IAAQlB,EAAG,MAAA,GACXmB,IAAQJ,EAAgB,MAAA,GACxBK,IAAyB,CAAA;AAC/B,WAASC,IAAI,GAAGA,IAAIhG,GAASgG,KAAK;AAChC,QAAIL,IAAO;AAGX,aAAS1E,IAAI,GAAGA,KAAKgD,GAAGhD;AACtB,MAAA0E,KAAQH,EAAGvE,IAAI,CAAC,IAAI4E,EAAMA,EAAM,SAAS5E,CAAC;AAE5C,aAAS2C,IAAI,GAAGA,KAAKO,GAAGP;AACtB,MAAA+B,KAAQF,EAAG7B,IAAI,CAAC,IAAIkC,EAAMA,EAAM,SAASlC,CAAC;AAE5C,IAAAiC,EAAM,KAAKF,CAAI,GACfG,EAAM,KAAK,CAAC,GACZC,EAAa,KAAKJ,IAAOnB,CAAI;AAAA,EAC/B;AAKA,SAAO,EAAE,SAFOyB,EAAkB5B,GAAYH,GAAG6B,CAAY,GAE3C,QAAA1D,GAAQ,aAAa,GAAA;AACzC;AAGA,SAASkC,EAAW2B,GAA4B;AAC9C,QAAMC,IAAM,IAAI,MAAMD,EAAO,SAAS,CAAC;AACvC,WAASjF,IAAI,GAAGA,IAAIiF,EAAO,QAAQjF,IAAK,CAAAkF,EAAIlF,IAAI,CAAC,IAAIiF,EAAOjF,CAAC,IAAIiF,EAAOjF,IAAI,CAAC;AAC7E,SAAOkF;AACT;AAMA,SAASF,EAAkB5B,GAAwBH,GAAW6B,GAAkC;AAC9F,MAAIK,IAAUL,EAAa,MAAA;AAC3B,WAASzB,IAAIJ,GAAGI,KAAK,GAAGA,KAAK;AAC3B,UAAM+B,IAAShC,EAAWC,IAAI,CAAC;AAC/B,QAAIgC,IAAOD,EAAOA,EAAO,SAAS,CAAC;AACnC,UAAME,IAAa,IAAI,MAAMH,EAAQ,MAAM;AAC3C,aAASnF,IAAI,GAAGA,IAAImF,EAAQ,QAAQnF;AAClC,MAAAqF,KAAQF,EAAQnF,CAAC,GACjBsF,EAAWtF,CAAC,IAAIqF;AAElB,IAAAF,IAAUG;AAAA,EACZ;AACA,SAAOH;AACT;AAMA,SAASR,EACPvB,GACAH,GACAiB,GACAqB,GACQ;AACR,SAAItC,MAAM,IAAUsC,IAELnC,EAAWH,IAAI,CAAC,EAEXiB,CAAC,IAEPqB;AAChB;AAMA,SAASzB,EAAMmB,GAAkB9B,GAA0D;AACzF,QAAMJ,IAAIkC,EAAO,QACXjB,IAAmB,CAAA,GACnBC,IAAoB,CAAA;AAC1B,WAASC,IAAIf,GAAOe,IAAInB,GAAGmB,KAAK;AAC9B,UAAMC,IAAgB,CAAA;AACtB,aAASnE,IAAI,GAAGA,KAAKmD,GAAOnD,OAAS,KAAKiF,EAAOf,IAAIlE,CAAC,CAAC;AACvD,IAAAgE,EAAK,KAAKG,CAAG,GACbF,EAAQ,KAAKgB,EAAOf,CAAC,CAAC;AAAA,EACxB;AACA,QAAME,IAASJ,EAAK,SAAS,IAAIK,EAASL,GAAMC,CAAO,IAAI,IAAI,MAAMd,CAAK,EAAE,KAAK,CAAC,GAC5EU,IAAY,IAAI,MAAMd,CAAC,EAAE,KAAK,CAAC;AACrC,WAASmB,IAAIf,GAAOe,IAAInB,GAAGmB,KAAK;AAC9B,QAAIQ,IAAO;AACX,aAAS1E,IAAI,GAAGA,KAAKmD,GAAOnD,IAAK,CAAA0E,MAASN,EAAOpE,IAAI,CAAC,KAAK,KAAKiF,EAAOf,IAAIlE,CAAC;AAC5E,IAAA6D,EAAUK,CAAC,IAAIe,EAAOf,CAAC,IAAIQ;AAAA,EAC7B;AACA,SAAO,EAAE,QAAAN,GAAQ,WAAAP,EAAA;AACnB;AAMA,SAASQ,EAASmB,GAAe3G,GAAuB;AACtD,QAAMmF,IAAOwB,EAAE,QACTC,IAAOD,EAAE,CAAC,EAAE,QAGZE,IAAkB,MAAM,KAAK,EAAE,QAAQD,EAAA,GAAQ,MAAM,IAAI,MAAMA,CAAI,EAAE,KAAK,CAAC,CAAC,GAC5EE,IAAgB,IAAI,MAAMF,CAAI,EAAE,KAAK,CAAC;AAE5C,WAASzF,IAAI,GAAGA,IAAIgE,GAAMhE;AACxB,aAAS7B,IAAI,GAAGA,IAAIsH,GAAMtH,KAAK;AAC7B,MAAAwH,EAAIxH,CAAC,KAAKqH,EAAExF,CAAC,EAAE7B,CAAC,IAAIU,EAAEmB,CAAC;AACvB,eAAS4F,IAAI,GAAGA,IAAIH,GAAMG;AACxB,QAAAF,EAAIvH,CAAC,EAAEyH,CAAC,KAAKJ,EAAExF,CAAC,EAAE7B,CAAC,IAAIqH,EAAExF,CAAC,EAAE4F,CAAC;AAAA,IAEjC;AAIF,WAASzH,IAAI,GAAGA,IAAIsH,GAAMtH,IAAK,CAAAuH,EAAIvH,CAAC,EAAEA,CAAC,KAAK;AAE5C,SAAO0H,EAAcH,GAAKC,CAAG;AAC/B;AAMO,SAASE,EAAcC,GAAeF,GAAuB;AAClE,QAAM3G,IAAI2G,EAAE,QACNG,IAAID,EAAE,IAAI,CAAC3B,GAAKnE,MAAM,CAAC,GAAGmE,GAAKyB,EAAE5F,CAAC,CAAC,CAAC;AAE1C,WAASgG,IAAM,GAAGA,IAAM/G,GAAG+G,KAAO;AAEhC,QAAIC,IAAQD;AACZ,aAASE,IAAIF,IAAM,GAAGE,IAAIjH,GAAGiH;AAC3B,MAAI,KAAK,IAAIH,EAAEG,CAAC,EAAEF,CAAG,CAAC,IAAI,KAAK,IAAID,EAAEE,CAAK,EAAED,CAAG,CAAC,MAAGC,IAAQC;AAE7D,QAAI,KAAK,IAAIH,EAAEE,CAAK,EAAED,CAAG,CAAC,IAAI,MAAO,QAAO,CAAA;AAC5C,KAACD,EAAEC,CAAG,GAAGD,EAAEE,CAAK,CAAC,IAAI,CAACF,EAAEE,CAAK,GAAGF,EAAEC,CAAG,CAAC;AAGtC,aAASE,IAAI,GAAGA,IAAIjH,GAAGiH,KAAK;AAC1B,UAAIA,MAAMF,EAAK;AACf,YAAMG,IAASJ,EAAEG,CAAC,EAAEF,CAAG,IAAID,EAAEC,CAAG,EAAEA,CAAG;AACrC,eAAS1B,IAAI0B,GAAK1B,KAAKrF,GAAGqF,IAAK,CAAAyB,EAAEG,CAAC,EAAE5B,CAAC,KAAK6B,IAASJ,EAAEC,CAAG,EAAE1B,CAAC;AAAA,IAC7D;AAAA,EACF;AAEA,QAAM1F,IAAI,IAAI,MAAMK,CAAC;AACrB,WAASe,IAAI,GAAGA,IAAIf,GAAGe,OAAOA,CAAC,IAAI+F,EAAE/F,CAAC,EAAEf,CAAC,IAAI8G,EAAE/F,CAAC,EAAEA,CAAC;AACnD,SAAOpB;AACT;AAMA,SAASsB,EACPkG,GACAhF,GACwD;AACxD,MAAI,CAACA,KAAUA,EAAO,WAAW;AAC/B,WAAO,EAAE,KAAK,GAAG,MAAM,GAAG,KAAK,GAAG,IAAI,EAAA;AAGxC,MAAIiF,IAAM,GACNC,IAAM,GACNC,IAAQ;AACZ,QAAMC,IAAqB,CAAA;AAE3B,WAASxG,IAAI,GAAGA,IAAIoB,EAAO,UAAUpB,IAAIoG,EAAW,QAAQpG,KAAK;AAC/D,UAAM,IAAIoB,EAAOpB,CAAC;AAClB,QAAI,KAAM,QAA2B,CAAC,OAAO,SAAS,CAAC,EAAG;AAC1D,UAAMyG,IAAML,EAAWpG,CAAC,IAAI;AAC5B,IAAAqG,KAAOI,IAAMA,GACbH,KAAO,KAAK,IAAIG,CAAG,GACnBD,EAAS,KAAKJ,EAAWpG,CAAC,CAAC,GAC3BuG;AAAA,EACF;AAEA,MAAIA,MAAU,EAAG,QAAO,EAAE,KAAK,GAAG,MAAM,GAAG,KAAK,GAAG,IAAI,EAAA;AAEvD,QAAMG,IAAML,IAAME,GACZI,IAAO,KAAK,KAAKD,CAAG,GACpBE,IAAMN,IAAMC,GAEZhD,IAAOiD,EAAS,OAAO,CAAChD,GAAGC,MAAMD,IAAIC,GAAG,CAAC,IAAI+C,EAAS,QACtDK,IAAML,EAAS,OAAO,CAAChD,GAAGC,MAAMD,KAAKC,IAAIF,MAASE,IAAIF,IAAO,CAAC,GAC9DuD,IAAKD,IAAM,IAAI,IAAIR,IAAMQ,IAAM;AAErC,SAAO;AAAA,IACL,KAAKE,EAAML,CAAG;AAAA,IACd,MAAMK,EAAMJ,CAAI;AAAA,IAChB,KAAKI,EAAMH,CAAG;AAAA,IACd,IAAIG,EAAMD,CAAE;AAAA,EAAA;AAEhB;AAEA,SAASC,EAAMtD,GAAmB;AAChC,SAAO,KAAK,MAAMA,IAAI,GAAG,IAAI;AAC/B;ACnpBA,MAAMuD,IAAsC;AAAA,EAC1C,MAAM;AAAA,EACN,SAAS;AAAA,EACT,aAAa;AAAA,EACb,UAAU,CAAC,YAAY,aAAa;AAAA,EACpC,MAAM,CAAC,cAAc,eAAe,UAAU;AAChD;AAKO,SAASC,EACdC,IAA2C,IACL;AACtC,MAAIC,IAA4B;AAChC,QAAMC,wBAAsB,IAAA,GAEtBC,IAAsB;AAAA,IAC1B,SAASC,GAAWC,GAAgD;AAClE,UAAI3I,GAAQC;AAGZ,UAAI,OAAOyI,KAAS,UAAU;AAC5B,YAAI,CAACH,EAAK,OAAM,IAAI,MAAM,wBAAwB;AAClD,cAAMlC,IAASkC,EAAI,KAAK,aAAA,EAAe,KAAK,CAAA3D,MAAKA,EAAE,MAAA,MAAY8D,CAAI;AACnE,YAAI,CAACrC,EAAQ,OAAM,IAAI,MAAM,qBAAqBqC,CAAI,EAAE;AACxD,cAAME,IAAavC,EAAO,QAAA;AAC1B,QAAArG,IAAI4I,EAAW,GACf3I,IAAI2I,EAAW;AAAA,MACjB,WAAW,MAAM,QAAQF,CAAI;AAC3B,QAAAzI,IAAIyI,GACJ1I,IAAI0I,EAAK,IAAI,CAACG,GAAGzH,MAAMA,CAAC;AAAA,eACfsH,EAAK,KAAKA,EAAK;AACxB,QAAA1I,IAAI0I,EAAK,GACTzI,IAAIyI,EAAK;AAAA;AAET,cAAM,IAAI,MAAM,qCAAqC;AAGvD,aAAO3I,EAAkBC,GAAGC,GAAG0I,EAAQ,QAAQA,EAAQ,SAASA,EAAQ,QAAQA,EAAQ,UAAU;AAAA,IACpG;AAAA,IAEA,MAAM,eAAeG,GAAkBH,GAAyD;AAC9F,UAAI,CAACJ,EAAK,OAAM,IAAI,MAAM,wBAAwB;AAElD,YAAMlC,IAASkC,EAAI,KAAK,aAAA,EAAe,KAAK,CAAA3D,MAAKA,EAAE,MAAA,MAAYkE,CAAQ;AACvE,UAAI,CAACzC,EAAQ,OAAM,IAAI,MAAM,qBAAqByC,CAAQ,EAAE;AAE5D,YAAMJ,IAAOrC,EAAO,QAAA;AACpB,aAAO,KAAK,SAASqC,GAAMC,CAAO;AAAA,IACpC;AAAA,IAEA,UAAUjI,GAA2BqI,IAA4C,IAAI;AACnF,YAAMC,IAAK,YAAY,KAAK,OAAA,EAAS,SAAS,EAAE,EAAE,OAAO,GAAG,CAAC,CAAC,IACxDC,IAAe;AAAA,QACnB,UAAU;AAAA,QACV,wBAAwB;AAAA,QACxB,GAAGX,EAAO;AAAA,QACV,GAAGS;AAAA,MAAA;AAGL,aAAAP,EAAgB,IAAIQ,GAAI,EAAE,QAAAtI,GAAQ,QAAQuI,GAAc,GACxDV,KAAA,QAAAA,EAAK,iBACES;AAAA,IACT;AAAA,IAEA,MAAMA,GAAa;AACjB,MAAIA,IACFR,EAAgB,OAAOQ,CAAE,IAEzBR,EAAgB,MAAA,GAElBD,KAAA,QAAAA,EAAK;AAAA,IACP;AAAA,EAAA;AAGF,WAASW,EAAcC,GAAqB;AAC1C,UAAM,EAAE,QAAAC,GAAQ,QAAAC,EAAA,IAAWF,GACrB,EAAE,OAAAG,MAAUF;AAClB,IAAKE,KAELd,EAAgB,QAAQ,CAACe,MAAQ;;AAC/B,YAAM,EAAE,QAAA7I,GAAQ,QAAQqI,EAAA,IAAcQ,GAChC,EAAE,SAAAC,GAAS,SAAA5H,GAAS,YAAA6H,GAAY,YAAAC,MAAehJ;AAKrD,UAHA4I,EAAM,KAAA,GAGFP,EAAU,0BAA0BU,KAAcC,GAAY;AAChE,QAAAJ,EAAM,UAAA;AACN,cAAMK,MAAUC,IAAAb,EAAU,kBAAV,gBAAAa,EAAyB,YAAW,MAC9CC,MAAQC,IAAAf,EAAU,kBAAV,gBAAAe,EAAyB,gBAAcC,IAAAhB,EAAU,cAAV,gBAAAgB,EAAqB,UAAS;AAEnF,QAAAT,EAAM,YAAYO,EAAM,WAAW,MAAM,IAAIA,IAAQA,IAAQ,KAAK,MAAMF,IAAU,GAAG,EAAE,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG;AAEnH,iBAASvI,IAAI,GAAGA,IAAIoI,EAAQ,QAAQpI,KAAK;AACvC,gBAAM4I,IAAKX,EAAO,aAAaG,EAAQpI,CAAC,CAAC,GACnC6I,IAAKZ,EAAO,aAAaK,EAAWtI,CAAC,CAAC;AAC5C,UAAIA,MAAM,IAAGkI,EAAM,OAAOU,GAAIC,CAAE,IAC3BX,EAAM,OAAOU,GAAIC,CAAE;AAAA,QAC1B;AAEA,iBAAS7I,IAAIoI,EAAQ,SAAS,GAAGpI,KAAK,GAAGA,KAAK;AAC5C,gBAAM4I,IAAKX,EAAO,aAAaG,EAAQpI,CAAC,CAAC,GACnC6I,IAAKZ,EAAO,aAAaI,EAAWrI,CAAC,CAAC;AAC5C,UAAAkI,EAAM,OAAOU,GAAIC,CAAE;AAAA,QACrB;AAEA,QAAAX,EAAM,UAAA,GACNA,EAAM,KAAA;AAAA,MACR;AAGA,UAAIP,EAAU,UAAU;AACtB,QAAAO,EAAM,UAAA,GACNA,EAAM,gBAAcY,IAAAnB,EAAU,cAAV,gBAAAmB,EAAqB,UAAS,WAClDZ,EAAM,cAAYa,IAAApB,EAAU,cAAV,gBAAAoB,EAAqB,UAAS,IAC5CC,IAAArB,EAAU,cAAV,QAAAqB,EAAqB,OACvBd,EAAM,YAAYP,EAAU,UAAU,IAAI,IAE1CO,EAAM,YAAY,CAAC,GAAG,CAAC,CAAC;AAG1B,iBAASlI,IAAI,GAAGA,IAAIoI,EAAQ,QAAQpI,KAAK;AACvC,gBAAM4I,IAAKX,EAAO,aAAaG,EAAQpI,CAAC,CAAC,GACnC6I,IAAKZ,EAAO,aAAazH,EAAQR,CAAC,CAAC;AACzC,UAAIA,MAAM,IAAGkI,EAAM,OAAOU,GAAIC,CAAE,IAC3BX,EAAM,OAAOU,GAAIC,CAAE;AAAA,QAC1B;AACA,QAAAX,EAAM,OAAA;AAAA,MACR;AAEA,MAAAA,EAAM,QAAA;AAAA,IACR,CAAC;AAAA,EACH;AAEA,SAAO;AAAA,IACL,UAAUlB;AAAA,IACV,OAAO1C,GAAkB;AACvB,MAAA6C,IAAM7C;AAAA,IACR;AAAA,IACA,gBAAgByD,GAAqBkB,GAA0B;AAC7D,MAAAnB,EAAcC,CAAI;AAAA,IACpB;AAAA,IACA,YAAYmB,GAAQ;;AAClB,MAAI,CAACA,EAAO,aAAWV,IAAAU,EAAO,kBAAP,gBAAAV,EAAsB,QAAQ,qBAAoB,MACzEpB,EAAgB,QAAQ,CAAC,EAAE,QAAA9H,GAAQ,QAAQqI,QAAgB;;AACzD,cAAM,EAAE,SAAAS,GAAS,SAAA5H,GAAS,YAAA6H,GAAY,YAAAC,MAAehJ,GAC/CL,IAAImJ,EAAQ,QACZe,IAA0C,CAAA;AAChD,iBAASnJ,IAAI,GAAGA,IAAIf,GAAGe;AACrB,UAAAmJ,EAAO,KAAK,EAAE,GAAGf,EAAQpI,CAAC,GAAG,GAAGQ,EAAQR,CAAC,GAAG;AAE9C,cAAMoJ,IACJzB,EAAU,0BAA0BU,KAAcC,IAC9C;AAAA,UACE,OAAO,MAAM,KAAK,EAAE,QAAQrJ,EAAA,GAAK,CAACwI,GAAGzH,OAAO,EAAE,GAAGoI,EAAQpI,CAAC,GAAG,GAAGsI,EAAWtI,CAAC,IAAI;AAAA,UAChF,OAAO,MAAM,KAAK,EAAE,QAAQf,EAAA,GAAK,CAACwI,GAAGzH,OAAO,EAAE,GAAGoI,EAAQpI,CAAC,GAAG,GAAGqI,EAAWrI,CAAC,IAAI;AAAA,QAAA,IAElF;AACN,QAAAqJ,EAAsBH,GAAQC,GAAQC,KAAIZ,IAAAb,EAAU,cAAV,gBAAAa,EAAqB,UAAS,SAAS;AAAA,MACnF,CAAC;AAAA,IACH;AAAA,IACA,KAAAnB;AAAA,EAAA;AAEJ;"}

package/dist/plugins/ml-integration.js

@@ -1,11 +1,12 @@
-var k = Object.defineProperty;
-var q = (i, t, n) => t in i ? k(i, t, { enumerable: !0, configurable: !0, writable: !0, value: n }) : i[t] = n;
-var h = (i, t, n) => q(i, typeof t != "symbol" ? t + "" : t, n);
-class x {
+var X = Object.defineProperty;
+var k = (i, t, n) => t in i ? X(i, t, { enumerable: !0, configurable: !0, writable: !0, value: n }) : i[t] = n;
+var d = (i, t, n) => k(i, typeof t != "symbol" ? t + "" : t, n);
+import { e as q } from "../regression-HNdyk1Dm.js";
+class L {
   constructor(t) {
-    h(this, "network");
-    h(this, "config");
-    h(this, "ready", !1);
+    d(this, "network");
+    d(this, "config");
+    d(this, "ready", !1);
     this.config = t, this.network = {
       layers: [],
       learningRate: 0.01,
@@ -22,13 +23,13 @@ class x {
     return this.config.type;
   }
   async predict(t) {
-    const n = performance.now(), e = Array.from(t.data), s = this.forwardPass(e), l = performance.now() - n;
+    const n = performance.now(), e = Array.from(t.data), s = this.forwardPass(e), c = performance.now() - n;
     return {
       modelId: this.config.id,
       output: new Float32Array(s),
       outputShape: [s.length],
       timestamp: Date.now(),
-      processingTime: l
+      processingTime: c
     };
   }
   forwardPass(t) {
@@ -39,10 +40,10 @@ class x {
   }
   denseLayer(t, n, e) {
     const s = [];
-    for (let l = 0; l < n.length; l++) {
-      let r = e[l];
+    for (let c = 0; c < n.length; c++) {
+      let r = e[c];
       for (let o = 0; o < t.length; o++)
-        r += t[o] * n[l][o];
+        r += t[o] * n[c][o];
       s.push(r);
     }
     return s;
@@ -73,13 +74,13 @@ class x {
     this.ready = !1;
   }
 }
-class I {
+class D {
   constructor(t) {
-    h(this, "config");
-    h(this, "coefficients", []);
-    h(this, "intercept", 0);
-    h(this, "ready", !1);
-    h(this, "lastTraining", null);
+    d(this, "config");
+    d(this, "coefficients", []);
+    d(this, "intercept", 0);
+    d(this, "ready", !1);
+    d(this, "lastTraining", null);
     this.config = t;
   }
   get id() {
@@ -96,13 +97,13 @@ class I {
       const o = r;
       let a = this.intercept;
       return this.coefficients.length > 0 && (a += this.coefficients[0] * o), a;
-    }), l = performance.now() - n;
+    }), c = performance.now() - n;
     return {
       modelId: this.config.id,
       output: new Float32Array(s),
       outputShape: [s.length],
       timestamp: Date.now(),
-      processingTime: l
+      processingTime: c
     };
   }
   /** Predict from full feature rows (multivariate). */
@@ -118,10 +119,10 @@ class I {
     const n = t.x.length, e = [];
     for (let u = 0; u < n; u++)
       e.push([1, ...t.x[u]]);
-    const s = F(b(e), e), l = F(b(e), t.y.map((u) => [u])), r = L(s), o = F(r, l);
+    const s = A(b(e), e), c = A(b(e), t.y.map((u) => [u])), r = z(s), o = A(r, c);
     this.intercept = o[0][0], this.coefficients = o.slice(1).map((u) => u[0]), this.ready = !0;
-    const a = this.predictRows(t.x), c = t.y.map((u, p) => u - a[p]), g = t.y.reduce((u, p) => u + p, 0) / (n || 1), m = c.reduce((u, p) => u + p * p, 0), d = t.y.reduce((u, p) => u + (p - g) ** 2, 0), M = d > 0 ? 1 - m / d : 0, y = Math.sqrt(m / (n || 1));
-    return this.lastTraining = { coefficients: this.coefficients, intercept: this.intercept, fitted: a, residuals: c, r2: M, rmse: y }, this.lastTraining;
+    const a = this.predictRows(t.x), l = t.y.map((u, h) => u - a[h]), g = t.y.reduce((u, h) => u + h, 0) / (n || 1), m = l.reduce((u, h) => u + h * h, 0), f = t.y.reduce((u, h) => u + (h - g) ** 2, 0), y = f > 0 ? 1 - m / f : 0, w = Math.sqrt(m / (n || 1));
+    return this.lastTraining = { coefficients: this.coefficients, intercept: this.intercept, fitted: a, residuals: l, r2: y, rmse: w }, this.lastTraining;
   }
   getTrainingResult() {
     return this.lastTraining;
@@ -142,33 +143,33 @@ class I {
 function b(i) {
   return i[0].map((t, n) => i.map((e) => e[n]));
 }
-function F(i, t) {
+function A(i, t) {
   const n = [];
   for (let e = 0; e < i.length; e++) {
     n[e] = [];
     for (let s = 0; s < t[0].length; s++) {
-      let l = 0;
+      let c = 0;
       for (let r = 0; r < t.length; r++)
-        l += i[e][r] * t[r][s];
-      n[e][s] = l;
+        c += i[e][r] * t[r][s];
+      n[e][s] = c;
     }
   }
   return n;
 }
-function L(i) {
+function z(i) {
   const t = i.length, n = i.map((e, s) => [
     ...e,
-    ...Array.from({ length: t }, (l, r) => s === r ? 1 : 0)
+    ...Array.from({ length: t }, (c, r) => s === r ? 1 : 0)
   ]);
   for (let e = 0; e < t; e++) {
     let s = e;
     for (let r = e + 1; r < t; r++)
       Math.abs(n[r][e]) > Math.abs(n[s][e]) && (s = r);
     if (Math.abs(n[s][e]) < 1e-12)
-      return i.map((r, o) => r.map((a, c) => o === c ? 1 : 0));
+      return i.map((r, o) => r.map((a, l) => o === l ? 1 : 0));
     [n[e], n[s]] = [n[s], n[e]];
-    const l = n[e][e];
-    for (let r = 0; r < 2 * t; r++) n[e][r] /= l;
+    const c = n[e][e];
+    for (let r = 0; r < 2 * t; r++) n[e][r] /= c;
     for (let r = 0; r < t; r++) {
       if (r === e) continue;
       const o = n[r][e];
@@ -177,13 +178,13 @@ function L(i) {
   }
   return n.map((e) => e.slice(t));
 }
-class z {
+class E {
   constructor(t) {
-    h(this, "config");
-    h(this, "filterType");
-    h(this, "cutoffFrequency");
-    h(this, "sampleRate");
-    h(this, "ready", !1);
+    d(this, "config");
+    d(this, "filterType");
+    d(this, "cutoffFrequency");
+    d(this, "sampleRate");
+    d(this, "ready", !1);
     var n, e, s;
     this.config = t, this.filterType = ((n = t.metadata) == null ? void 0 : n.filterType) || "lowpass", this.cutoffFrequency = ((e = t.metadata) == null ? void 0 : e.cutoffFrequency) || 1e3, this.sampleRate = ((s = t.metadata) == null ? void 0 : s.sampleRate) || 44100;
   }
@@ -197,13 +198,13 @@ class z {
     return this.config.type;
   }
   async predict(t) {
-    const n = performance.now(), e = Array.from(t.data), s = this.applyFilter(e), l = performance.now() - n;
+    const n = performance.now(), e = Array.from(t.data), s = this.applyFilter(e), c = performance.now() - n;
     return {
       modelId: this.config.id,
       output: new Float32Array(s),
       outputShape: [s.length],
       timestamp: Date.now(),
-      processingTime: l
+      processingTime: c
     };
   }
   applyFilter(t) {
@@ -220,8 +221,8 @@ class z {
   }
   lowpassFilter(t) {
     const n = 2 * Math.PI * this.cutoffFrequency / this.sampleRate, e = n / (1 + n), s = [t[0]];
-    for (let l = 1; l < t.length; l++)
-      s.push(e * t[l] + (1 - e) * s[l - 1]);
+    for (let c = 1; c < t.length; c++)
+      s.push(e * t[c] + (1 - e) * s[c - 1]);
     return s;
   }
   highpassFilter(t) {
@@ -247,57 +248,57 @@ class z {
     this.ready = !1;
   }
 }
-function X(i) {
+function I(i) {
   switch (i.type) {
     case "neural-network":
-      return new x(i);
+      return new L(i);
     case "linear-regression":
-      return new I(i);
+      return new D(i);
     case "signal-processor":
-      return new z(i);
+      return new E(i);
     default:
-      return new I(i);
+      return new D(i);
   }
 }
 function V(i) {
   const t = i.length, n = new Array(t), e = new Array(t);
   for (let s = 0; s < t; s++) {
-    let l = 0, r = 0;
+    let c = 0, r = 0;
     for (let o = 0; o < t; o++) {
       const a = -2 * Math.PI * s * o / t;
-      l += i[o] * Math.cos(a), r += i[o] * Math.sin(a);
+      c += i[o] * Math.cos(a), r += i[o] * Math.sin(a);
     }
-    n[s] = l, e[s] = r;
+    n[s] = c, e[s] = r;
   }
   return { real: n, imag: e };
 }
-function R(i) {
+function P(i) {
   return i.reduce((t, n) => t + n, 0) / i.length;
 }
 function Y(i) {
-  const t = R(i), n = i.reduce((e, s) => e + Math.pow(s - t, 2), 0) / i.length;
+  const t = P(i), n = i.reduce((e, s) => e + Math.pow(s - t, 2), 0) / i.length;
   return Math.sqrt(n);
 }
 function $(i, t) {
-  const n = i.length, e = R(i), s = R(t);
-  let l = 0, r = 0, o = 0;
+  const n = i.length, e = P(i), s = P(t);
+  let c = 0, r = 0, o = 0;
   for (let a = 0; a < n; a++) {
-    const c = i[a] - e, g = t[a] - s;
-    l += c * g, r += c * c, o += g * g;
+    const l = i[a] - e, g = t[a] - s;
+    c += l * g, r += l * l, o += g * g;
   }
-  return l / Math.sqrt(r * o);
+  return c / Math.sqrt(r * o);
 }
-const E = {
+const N = {
   name: "velo-plot-ml-integration",
   version: "1.0.0",
   description: "Machine Learning Integration Bridge for Velo Plot",
   provides: ["ml", "prediction", "inference"],
   tags: ["ml", "ai", "analysis"]
 };
-function _(i = {}) {
+function j(i = {}) {
   let t = null;
   const n = /* @__PURE__ */ new Map(), e = /* @__PURE__ */ new Map();
-  i.models && i.models.forEach((r) => n.set(r.id, X(r)));
+  i.models && i.models.forEach((r) => n.set(r.id, I(r)));
   const s = {
     registerModel(r) {
       if (!r.id) {
@@ -309,10 +310,10 @@ function _(i = {}) {
     async runInference(r, o) {
       const a = n.get(r);
       if (!a) throw new Error(`Model not found: ${r}`);
-      const c = t == null ? void 0 : t.data.getSeriesData(o);
-      if (!c) throw new Error(`Series not found: ${o}`);
-      const g = new Float32Array(c.y), m = await a.predict({ data: g });
-      return m.xValues = c.x, m;
+      const l = t == null ? void 0 : t.data.getSeriesData(o);
+      if (!l) throw new Error(`Series not found: ${o}`);
+      const g = new Float32Array(l.y), m = await a.predict({ data: g });
+      return m.xValues = l.x, m;
     },
     visualizeResults(r, o = {}) {
       const a = `ml-viz-${Math.random().toString(36).substr(2, 9)}`;
@@ -338,7 +339,7 @@ function _(i = {}) {
      */
     trainModel(r, o) {
       let a = n.get(r);
-      if (a || (a = X({ id: r, name: r, type: "linear-regression" }), n.set(r, a)), typeof a.train != "function")
+      if (a || (a = I({ id: r, name: r, type: "linear-regression" }), n.set(r, a)), typeof a.train != "function")
         throw new Error(`Model '${r}' does not support training`);
       return a.train(o);
     },
@@ -354,53 +355,66 @@ function _(i = {}) {
     },
     stats: {
       fft: V,
-      mean: R,
+      mean: P,
       standardDeviation: Y,
       correlation: $
     }
   };
-  function l(r) {
-    const { render: o, coords: a } = r, { ctx2d: c } = o;
-    c && e.forEach((g) => {
-      var S, P, A, D;
-      const { result: m, config: d } = g, { output: M, xValues: y, confidence: u } = m;
-      if (!y) return;
-      const p = Math.min(y.length, M.length);
-      if (c.save(), d.showConfidenceInterval && u && u.length === p) {
-        c.beginPath();
-        const w = d.intervalOpacity ?? 0.2;
-        c.fillStyle = (((S = d.lineStyle) == null ? void 0 : S.color) || "#3b82f6") + Math.floor(w * 255).toString(16).padStart(2, "0");
-        for (let f = 0; f < p; f++) {
-          const v = a.dataToPixelX(y[f]), T = a.dataToPixelY(M[f] + u[f]);
-          f === 0 ? c.moveTo(v, T) : c.lineTo(v, T);
+  function c(r) {
+    const { render: o, coords: a } = r, { ctx2d: l } = o;
+    l && e.forEach((g) => {
+      var R, v, M, x;
+      const { result: m, config: f } = g, { output: y, xValues: w, confidence: u } = m;
+      if (!w) return;
+      const h = Math.min(w.length, y.length);
+      if (l.save(), f.showConfidenceInterval && u && u.length === h) {
+        l.beginPath();
+        const T = f.intervalOpacity ?? 0.2;
+        l.fillStyle = (((R = f.lineStyle) == null ? void 0 : R.color) || "#3b82f6") + Math.floor(T * 255).toString(16).padStart(2, "0");
+        for (let p = 0; p < h; p++) {
+          const F = a.dataToPixelX(w[p]), S = a.dataToPixelY(y[p] + u[p]);
+          p === 0 ? l.moveTo(F, S) : l.lineTo(F, S);
         }
-        for (let f = p - 1; f >= 0; f--) {
-          const v = a.dataToPixelX(y[f]), T = a.dataToPixelY(M[f] - u[f]);
-          c.lineTo(v, T);
+        for (let p = h - 1; p >= 0; p--) {
+          const F = a.dataToPixelX(w[p]), S = a.dataToPixelY(y[p] - u[p]);
+          l.lineTo(F, S);
         }
-        c.closePath(), c.fill();
+        l.closePath(), l.fill();
       }
-      c.beginPath(), c.strokeStyle = ((P = d.lineStyle) == null ? void 0 : P.color) || "#3b82f6", c.lineWidth = ((A = d.lineStyle) == null ? void 0 : A.width) || 2, (D = d.lineStyle) != null && D.dash && c.setLineDash(d.lineStyle.dash);
-      for (let w = 0; w < p; w++) {
-        const f = a.dataToPixelX(y[w]), v = a.dataToPixelY(M[w]);
-        w === 0 ? c.moveTo(f, v) : c.lineTo(f, v);
+      l.beginPath(), l.strokeStyle = ((v = f.lineStyle) == null ? void 0 : v.color) || "#3b82f6", l.lineWidth = ((M = f.lineStyle) == null ? void 0 : M.width) || 2, (x = f.lineStyle) != null && x.dash && l.setLineDash(f.lineStyle.dash);
+      for (let T = 0; T < h; T++) {
+        const p = a.dataToPixelX(w[T]), F = a.dataToPixelY(y[T]);
+        T === 0 ? l.moveTo(p, F) : l.lineTo(p, F);
       }
-      c.stroke(), c.restore();
+      l.stroke(), l.restore();
     });
   }
   return {
-    manifest: E,
+    manifest: N,
     onInit(r) {
       t = r;
     },
     onRenderOverlay(r, o) {
-      l(r);
+      c(r);
+    },
+    onExportSVG(r) {
+      var o;
+      !r.builder || ((o = r.exportContext) == null ? void 0 : o.options.includeOverlays) === !1 || e.forEach((a) => {
+        var R;
+        const { result: l, config: g } = a, { output: m, xValues: f, confidence: y } = l;
+        if (!(f != null && f.length)) return;
+        const w = Math.min(f.length, m.length), u = Array.from({ length: w }, (v, M) => ({ x: f[M], y: m[M] })), h = g.showConfidenceInterval && (y == null ? void 0 : y.length) === w ? {
+          upper: u.map((v, M) => ({ x: v.x, y: v.y + y[M] })),
+          lower: u.map((v, M) => ({ x: v.x, y: v.y - y[M] }))
+        } : void 0;
+        q(r, u, h, ((R = g.lineStyle) == null ? void 0 : R.color) ?? "#3b82f6");
+      });
     },
     api: s
   };
 }
 export {
-  _ as PluginMLIntegration,
-  _ as default
+  j as PluginMLIntegration,
+  j as default
 };
 //# sourceMappingURL=ml-integration.js.map