randomstatsmodels 0.1.0__tar.gz

randomstatsmodels-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Jacob
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to do so, subject to the following
+conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

randomstatsmodels-0.1.0/PKG-INFO

@@ -0,0 +1,89 @@
+Metadata-Version: 2.4
+Name: randomstatsmodels
+Version: 0.1.0
+Summary: Tools for benchmarking, metrics, and models.
+Author: Jacob Wright
+License: MIT License
+        
+        Copyright (c) 2025 Jacob
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to do so, subject to the following
+        conditions:
+        
+        The above copyright notice and this permission notice shall be included in
+        all copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+Keywords: statistics,machine learning,metrics,models
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Topic :: Scientific/Engineering :: Mathematics
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.24
+Dynamic: license-file
+
+# randomstatsmodels
+
+A tiny, modern Python package skeleton for experimenting with forecasting and statistics utilities.
+
+## Quick start
+```bash
+# from the project root
+python -m venv .venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate
+pip install -e ".[dev]"  # install in editable mode with dev extras
+randomstatsmodels --version
+```
+
+## Usage
+```python
+from randomstatsmodels.metrics import mae, rmse
+
+y_true = [1, 2, 3]
+y_pred = [1.1, 1.9, 3.2]
+print(mae(y_true, y_pred))
+```
+
+## Testing
+```bash
+pytest -q
+```
+
+## Using your models
+Put your custom models in `randomstatsmodels/user_models.py` (we copied your uploaded file there).
+
+### Python API
+```python
+from randomstatsmodels.api import predict
+# Model ref can be 'randomstatsmodels.user_models:MyModel' or an object/callable
+yhat = predict("randomstatsmodels.user_models:MyModel", X_dataframe)
+```
+
+### CLI
+```bash
+randomstatsmodels predict \        --model randomstatsmodels.user_models:MyModel \        --input data.csv \        --output preds.csv
+```
+
+The adapter accepts:
+- Objects with `.predict(X, **kwargs)`
+- Callables like `def f(X): ...`
+- Classes that can be instantiated without args and have `.predict(X)`

randomstatsmodels-0.1.0/README.md

@@ -0,0 +1,45 @@
+# randomstatsmodels
+
+A tiny, modern Python package skeleton for experimenting with forecasting and statistics utilities.
+
+## Quick start
+```bash
+# from the project root
+python -m venv .venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate
+pip install -e ".[dev]"  # install in editable mode with dev extras
+randomstatsmodels --version
+```
+
+## Usage
+```python
+from randomstatsmodels.metrics import mae, rmse
+
+y_true = [1, 2, 3]
+y_pred = [1.1, 1.9, 3.2]
+print(mae(y_true, y_pred))
+```
+
+## Testing
+```bash
+pytest -q
+```
+
+## Using your models
+Put your custom models in `randomstatsmodels/user_models.py` (we copied your uploaded file there).
+
+### Python API
+```python
+from randomstatsmodels.api import predict
+# Model ref can be 'randomstatsmodels.user_models:MyModel' or an object/callable
+yhat = predict("randomstatsmodels.user_models:MyModel", X_dataframe)
+```
+
+### CLI
+```bash
+randomstatsmodels predict \        --model randomstatsmodels.user_models:MyModel \        --input data.csv \        --output preds.csv
+```
+
+The adapter accepts:
+- Objects with `.predict(X, **kwargs)`
+- Callables like `def f(X): ...`
+- Classes that can be instantiated without args and have `.predict(X)`

randomstatsmodels-0.1.0/pyproject.toml

@@ -0,0 +1,41 @@
+[build-system]
+requires = ["setuptools>=69", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "randomstatsmodels"
+version = "0.1.0"
+description = "Tools for benchmarking, metrics, and models."
+readme = "README.md"
+license = { file = "LICENSE" }
+authors = [
+  { name = "Jacob Wright" }
+]
+requires-python = ">=3.9"
+dependencies = [
+  "numpy>=1.24"
+]
+
+keywords = ["statistics", "machine learning", "metrics", "models"]
+
+classifiers = [
+  "Development Status :: 3 - Alpha",
+  "Intended Audience :: Science/Research",
+  "Topic :: Scientific/Engineering :: Mathematics",
+  "Topic :: Scientific/Engineering :: Artificial Intelligence",
+  "License :: OSI Approved :: MIT License",
+  "Programming Language :: Python :: 3",
+  "Programming Language :: Python :: 3.9",
+  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Operating System :: OS Independent"
+]
+
+[tool.setuptools.packages.find]
+where = ["."]
+include = ["randomstatsmodels*"]
+
+[tool.setuptools.package-data]
+# if you ever add CSVs or py.typed, include them here
+randomstatsmodels = []

randomstatsmodels-0.1.0/randomstatsmodels/__init__.py

@@ -0,0 +1,39 @@
+# randomstatsmodels/__init__.py
+# This makes the folder a Python package.
+
+from .metrics.metrics import mae, mape, smape, rmse
+from .models.models import (
+    AutoHybridForecaster,
+    AutoKNN,
+    AutoMELD,
+    AutoNEO,
+    AutoPALF,
+    AutoThetaAR,
+    AutoPolymath,
+    AutoSeasonalAR,
+    AutoFourier,
+    AutoRollingMedian,
+    AutoTrimmedMean,
+    AutoWindow,
+    AutoRankInsertion,
+)
+from .benchmarking.benchmarking import benchmark_model, benchmark_models
+
+__version__ = "0.1.0"
+__all__ = [
+    "__version__",
+    "mae",
+    "mape",
+    "smape",
+    "rmse",
+    "AutoHybridForecaster",
+    "AutoKNN",
+    "AutoMELD",
+    "AutoNEO",
+    "AutoPALF",
+    "AutoThetaAR",
+    "AutoPolymath",
+    "AutoSeasonalAR",
+    "benchmark_models",
+    "benchmark_model",
+]

randomstatsmodels-0.1.0/randomstatsmodels/benchmarking/__init__.py

@@ -0,0 +1 @@
+from .benchmarking import benchmark_model, benchmark_models

randomstatsmodels-0.1.0/randomstatsmodels/benchmarking/benchmarking.py

@@ -0,0 +1,304 @@
+import time
+import math
+import numpy as np
+from ..metrics import mae, rmse, mape, smape
+
+import time
+import numpy as np
+from ..metrics import mae, rmse, mape, smape
+
+
+def benchmark_model(model_class, data, iterations=1, h=7):
+    """
+    Benchmark the training + prediction speed of a time series model,
+    and compute MAE, RMSE, MAPE, sMAPE on the last h points.
+
+    Parameters
+    ----------
+    model_class : class
+        The model class to initialize (e.g., AutoNEO).
+    data : array-like
+        The time series data.
+    iterations : int, default=5
+        Number of times to run the benchmark.
+    h : int, default=20
+        Forecast horizon.
+    **fit_kwargs : dict
+        Additional arguments passed to model.fit().
+
+    Returns
+    -------
+    results : dict
+        {
+          "avg_total_time_s": float,
+          "avg_fit_time_s": float,
+          "avg_predict_time_s": float,
+          "avg_mae": float,
+          "avg_rmse": float,
+          "avg_mape": float,
+          "avg_smape": float,
+          "per_iteration": [
+              {
+                "fit_time_s": float,
+                "predict_time_s": float,
+                "total_time_s": float,
+                "mae": float,
+                "rmse": float,
+                "mape": float,
+                "smape": float
+              },
+              ...
+          ]
+        }
+    """
+    data = np.asarray(data)
+    assert len(data) > h, "Data length must be greater than forecast horizon h."
+
+    per_iter = []
+    for i in range(iterations):
+
+        model = model_class()  # fresh model each run
+
+        model_name = model_class.__name__
+
+        t0 = time.time()
+        model.fit(data[:-h])
+        fit_time = time.time() - t0
+
+        t1 = time.time()
+        y_pred = model.predict(h)
+
+        predict_time = time.time() - t1
+
+        total_time = fit_time + predict_time
+
+        y_true = data[-h:]
+        # Ensure shapes compatible
+        y_pred = np.asarray(y_pred).reshape(-1)[:h]
+
+        if model_name == "AutoETS":
+            y_pred = y_pred[0]["mean"]
+
+        iter_metrics = {
+            "fit_time_s": fit_time,
+            "predict_time_s": predict_time,
+            "total_time_s": total_time,
+            "mae": float(mae(y_true, y_pred)),
+            "rmse": float(rmse(y_true, y_pred)),
+            "mape": float(mape(y_true, y_pred)),
+            "smape": float(smape(y_true, y_pred)),
+        }
+        per_iter.append(iter_metrics)
+
+    # Averages
+    avg_total = float(np.mean([x["total_time_s"] for x in per_iter]))
+    avg_fit = float(np.mean([x["fit_time_s"] for x in per_iter]))
+    avg_predict = float(np.mean([x["predict_time_s"] for x in per_iter]))
+    avg_mae_ = float(np.mean([x["mae"] for x in per_iter]))
+    avg_rmse_ = float(np.mean([x["rmse"] for x in per_iter]))
+    avg_mape_ = float(np.mean([x["mape"] for x in per_iter]))
+    avg_smape_ = float(np.mean([x["smape"] for x in per_iter]))
+
+    print(
+        f"\nAverages over {iterations} runs --> "
+        f"fit: {avg_fit:.4f}s | predict: {avg_predict:.4f}s | total: {avg_total:.4f}s | "
+        f"MAE: {avg_mae_:.4f} | RMSE: {avg_rmse_:.4f} | MAPE: {avg_mape_:.4f} | sMAPE: {avg_smape_:.4f}"
+    )
+
+    return {
+        "avg_total_time_s": round(avg_total, 2),
+        "avg_fit_time_s": round(avg_fit, 2),
+        "avg_predict_time_s": round(avg_predict, 2),
+        "avg_mae": round(avg_mae_, 3),
+        "avg_rmse": round(avg_rmse_, 3),
+        "avg_mape": round(avg_mape_, 3),
+        "avg_smape": round(avg_smape_, 3),
+        "per_iteration": per_iter,
+    }
+
+
+import time
+import numpy as np
+from ..metrics import mae, rmse, mape, smape
+
+
+def _coerce_forecast(yp, h, model_name):
+    """
+    Coerce various model outputs to a 1D np.ndarray of length h.
+    Handles special cases like AutoETS structure.
+    """
+    # Special-case: your AutoETS wrapper shape
+    if model_name == "AutoETS":
+        # expected like [{'mean': np.array([...])}, ...] or similar
+        try:
+            yp = yp["mean"]
+
+        except Exception:
+            pass
+
+    if yp.shape[0] != h:
+        raise ValueError(f"{model_name}.predict({h}) returned length {yp.shape[0]} (expected {h}).")
+    return yp
+
+
+def _metrics_dict(y_true, y_pred):
+    return {
+        "mae": float(mae(y_true, y_pred)),
+        "rmse": float(rmse(y_true, y_pred)),
+        "mape": float(mape(y_true, y_pred)),
+        "smape": float(smape(y_true, y_pred)),
+    }
+
+
+def _avg_block(per_iter):
+    return {
+        "avg_total_time_s": (
+            round(float(np.mean([x["total_time_s"] for x in per_iter])), 2)
+            if per_iter and "total_time_s" in per_iter[0]
+            else None
+        ),
+        "avg_fit_time_s": (
+            round(float(np.mean([x["fit_time_s"] for x in per_iter])), 2)
+            if per_iter and "fit_time_s" in per_iter[0]
+            else None
+        ),
+        "avg_predict_time_s": (
+            round(float(np.mean([x["predict_time_s"] for x in per_iter])), 2)
+            if per_iter and "predict_time_s" in per_iter[0]
+            else None
+        ),
+        "avg_mae": round(float(np.mean([x["mae"] for x in per_iter])), 3) if per_iter else None,
+        "avg_rmse": round(float(np.mean([x["rmse"] for x in per_iter])), 3) if per_iter else None,
+        "avg_mape": round(float(np.mean([x["mape"] for x in per_iter])), 3) if per_iter else None,
+        "avg_smape": round(float(np.mean([x["smape"] for x in per_iter])), 3) if per_iter else None,
+    }
+
+
+def _safe_fmt(x, fmt=".3f"):
+    if x is None:
+        return "—"
+    try:
+        if isinstance(x, float) and (math.isnan(x) or math.isinf(x)):
+            return "—"
+        return format(x, fmt)
+    except Exception:
+        return "—"
+
+
+def benchmark_models(
+    model_classes,
+    data,
+    iterations=1,
+    h=7,
+    ensembles=("mean", "median"),
+    exclude_from_ensemble=None,
+):
+    """
+    (docstring unchanged)
+    """
+    data = np.asarray(data)
+    assert len(data) > h, "Data length must be greater than forecast horizon h."
+    y_true = data[-h:]
+    model_classes = list(model_classes)
+
+    if exclude_from_ensemble is None:
+        exclude_from_ensemble = []
+    exclude_names = {(cls.__name__ if not isinstance(cls, str) else cls) for cls in exclude_from_ensemble}
+
+    results = {
+        "meta": {"iterations": iterations, "h": h, "n_models": len(model_classes)},
+        "models": {},
+        "ensembles": {},
+    }
+
+    # Prepare per-model storage
+    per_model_iters = {cls.__name__: [] for cls in model_classes}
+    failed_models = set()
+
+    # Per-iteration: collect predictions for ensembles
+    ens_iters_preds = []  # list per iteration: 2D array [n_models_used x h]
+
+    for i in range(iterations):
+        iter_preds = []
+        for cls in model_classes:
+            model_name = cls.__name__
+            try:
+                # fresh model each run
+                t0 = time.time()
+                model = cls()
+                model.fit(data[:-h])
+                fit_time = time.time() - t0
+
+                t1 = time.time()
+                y_pred = model.predict(h)
+                predict_time = time.time() - t1
+
+                y_pred = _coerce_forecast(y_pred, h, model_name)
+                if np.isnan(y_pred).any():
+                    print(f"Skipping model {model_name}: NaN in predictions")
+                    failed_models.add(model_name)
+                    continue
+
+                total_time = fit_time + predict_time
+
+                # metrics
+                m = _metrics_dict(y_true, y_pred)
+
+                per_model_iters[model_name].append(
+                    {
+                        "fit_time_s": float(fit_time),
+                        "predict_time_s": float(predict_time),
+                        "total_time_s": float(total_time),
+                        **m,
+                    }
+                )
+
+                # include in ensemble only if not excluded
+                if model_name not in exclude_names:
+                    iter_preds.append(y_pred)
+
+            except Exception as e:
+                print(f"Skipping model {model_name}: {e}")
+                failed_models.add(model_name)
+                continue
+
+        # Store stacked predictions for ensembles this iteration
+        if iter_preds:
+            ens_iters_preds.append(np.vstack(iter_preds))  # shape: (n_used_models, h)
+
+    # Aggregate per-model (only include models with at least one valid iteration)
+    for model_name, per_iter in per_model_iters.items():
+        if not per_iter:
+            # Do not include empty models to avoid None in summary formatting
+            continue
+        results["models"][model_name] = {**_avg_block(per_iter), "per_iteration": per_iter}
+
+    # Compute ensembles (metrics only; no timing)
+    valid_ens = set([e.lower() for e in ensembles]) if ensembles else set()
+    for ens_type in ("mean", "median"):
+        if ens_type in valid_ens and ens_iters_preds:
+            per_iter_metrics = []
+            for stacked in ens_iters_preds:
+                if stacked.size == 0:
+                    continue
+                if ens_type == "mean":
+                    y_ens = np.nanmean(stacked, axis=0)
+                else:  # median
+                    y_ens = np.nanmedian(stacked, axis=0)
+                per_iter_metrics.append(_metrics_dict(y_true, y_ens))
+
+            if per_iter_metrics:
+                results["ensembles"][ens_type] = {
+                    "avg_mae": round(float(np.mean([x["mae"] for x in per_iter_metrics])), 3),
+                    "avg_rmse": round(float(np.mean([x["rmse"] for x in per_iter_metrics])), 3),
+                    "avg_mape": round(float(np.mean([x["mape"] for x in per_iter_metrics])), 3),
+                    "avg_smape": round(float(np.mean([x["smape"] for x in per_iter_metrics])), 3),
+                    "per_iteration": per_iter_metrics,
+                }
+
+    # Optional: brief console summary (use safe formatting)
+    print(f"\nBenchmark over {iterations} runs (h={h})")
+    if failed_models:
+        print(f"(Some models were skipped due to errors/NaNs: {sorted(failed_models)})")
+
+    return results

randomstatsmodels-0.1.0/randomstatsmodels/metrics/__init__.py

@@ -0,0 +1 @@
+from .metrics import mae, rmse, mape, smape

randomstatsmodels-0.1.0/randomstatsmodels/metrics/metrics.py

@@ -0,0 +1,27 @@
+import numpy as np
+
+
+def mae(y_true, y_pred):
+    y_true = np.asarray(y_true, float)
+    y_pred = np.asarray(y_pred, float)
+    return np.mean(np.abs(y_true - y_pred))
+
+
+def rmse(y_true, y_pred):
+    y_true = np.asarray(y_true, float)
+    y_pred = np.asarray(y_pred, float)
+    return np.sqrt(np.mean((y_true - y_pred) ** 2))
+
+
+def mape(y_true, y_pred, epsilon=1e-8):
+    y_true = np.asarray(y_true, float)
+    y_pred = np.asarray(y_pred, float)
+    denom = np.maximum(np.abs(y_true), epsilon)
+    return np.mean(np.abs((y_true - y_pred) / denom)) * 100.0
+
+
+def smape(y_true, y_pred, epsilon=1e-8):
+    y_true = np.asarray(y_true, float)
+    y_pred = np.asarray(y_pred, float)
+    denom = np.maximum((np.abs(y_true) + np.abs(y_pred)) / 2.0, epsilon)
+    return np.mean(np.abs(y_true - y_pred) / denom) * 100.0

randomstatsmodels-0.1.0/randomstatsmodels/models/__init__.py

@@ -0,0 +1,15 @@
+from .models import (
+    AutoHybridForecaster,
+    AutoKNN,
+    AutoMELD,
+    AutoNEO,
+    AutoPALF,
+    AutoThetaAR,
+    AutoPolymath,
+    AutoSeasonalAR,
+    AutoFourier,
+    AutoRollingMedian,
+    AutoTrimmedMean,
+    AutoWindow,
+    AutoRankInsertion,
+)

randomstatsmodels-0.1.0/randomstatsmodels/models/model_utils.py

@@ -0,0 +1,47 @@
+import numpy as np
+
+
+def _weighted_quantile(values, weights, q):
+    values = np.asarray(values, float)
+    weights = np.asarray(weights, float)
+    srt = np.argsort(values)
+    v, w = values[srt], weights[srt]
+    cw = np.cumsum(w) / np.sum(w)
+    idx = np.searchsorted(cw, q, side="left")
+    idx = np.clip(idx, 0, len(v) - 1)
+    return float(v[idx])
+
+
+def _golden_section_minimize(f, a, b, tol=1e-6, max_iter=200):
+    phi = (1 + 5**0.5) / 2
+    invphi = 1 / phi
+    c = b - invphi * (b - a)
+    d = a + invphi * (b - a)
+    fc = f(c)
+    fd = f(d)
+    for _ in range(max_iter):
+        if abs(b - a) < tol:
+            break
+        if fc < fd:
+            b, d, fd = d, c, fc
+            c = b - invphi * (b - a)
+            fc = f(c)
+        else:
+            a, c, fc = c, d, fd
+            d = a + invphi * (b - a)
+            fd = f(d)
+    return (a + b) / 2
+
+
+def _penalty_value(r, kind="l2", delta=1.0, tau=0.5):
+    if kind == "l2":
+        return 0.5 * r * r
+    elif kind == "l1":
+        return np.abs(r)
+    elif kind == "huber":
+        a = np.abs(r)
+        return np.where(a <= delta, 0.5 * r * r, delta * (a - 0.5 * delta))
+    elif kind == "pinball":
+        return np.where(r >= 0, tau * r, (tau - 1.0) * r)
+    else:
+        raise ValueError("Unknown penalty")