sovai 0.2.63__tar.gz → 0.2.65__tar.gz

{sovai-0.2.63 → sovai-0.2.65}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: sovai
-Version: 0.2.63
+Version: 0.2.65
 Summary: python for asset management
 License: Proprietary License
 Author: Derek Snow

{sovai-0.2.63 → sovai-0.2.65}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [project]
 name = "sovai"
-version = "0.2.63"
+version = "0.2.65"
 description = "python for asset management"
 authors = [
     {name = "Derek Snow", email = "d.snow@outlook.com"}

{sovai-0.2.63 → sovai-0.2.65}/sovai/extensions/anomalies.py

@@ -61,6 +61,7 @@ def prepare_and_select_neighbors(df_accounting, ticker_select, neighbours):
         df_calc = df_accounting.loc[
             df_accounting.index.get_level_values("ticker").isin(neighs)
         ]
+        df_calc = df_calc[~df_calc.index.duplicated(keep="first")]  # ← NEW LINE
         return df_calc
 
     else:

sovai-0.2.65/sovai/extensions/change_point_generator.py

@@ -0,0 +1,590 @@
+import numpy as np
+import pandas as pd
+from typing import Tuple, List, Optional # Modified to include List and Optional for new functions
+import plotly.graph_objects as go
+from plotly.subplots import make_subplots
+import dash
+from dash import dcc, html
+from dash.dependencies import Input, Output
+import random
+import socket # Added for port management
+import time # Added for port management
+import errno # Added for port management
+
+# from port_manager_utility import get_unique_port, kill_process_on_port, release_port, get_port_manager_instance
+# from utils.port_manager import get_unique_port, kill_process_on_port
+from sovai.utils.port_manager import get_unique_port, kill_process_on_port
+
+random.seed(42)
+np.random.seed(42)
+
+# Global dictionary to keep track of apps and their ports launched by this script
+app_ports = {}
+
+class ImprovedCusumDetector: # Exactly as in the first script
+    def __init__(
+        self,
+        window_size_ratio: float = 0.05,
+        threshold_factor: float = 2.5,
+        drift_factor: float = 0.05,
+        min_distance: int = 30,
+        sideways_threshold: float = 0.03,
+        warm_up_period_ratio: float = 0.1,
+    ):
+        self.window_size_ratio = window_size_ratio
+        self.threshold_factor = threshold_factor
+        self.drift_factor = drift_factor
+        self.min_distance = min_distance
+        self.sideways_threshold = sideways_threshold
+        self.warm_up_period_ratio = warm_up_period_ratio
+
+    def detect_changepoints(
+        self, signal: np.ndarray
+    ) -> Tuple[np.ndarray, np.ndarray, np.ndarray, int]:
+        n = len(signal)
+        window_size = max(int(n * self.window_size_ratio), 8)
+        warm_up_period = int(n * self.warm_up_period_ratio)
+
+        ewma = self._calculate_ewma(signal, window_size)
+        z_scores = self._calculate_rolling_z_scores(signal, ewma, window_size)
+        scores, changepoints, trend_codes = self._calculate_two_sided_cusum(
+            signal,
+            z_scores,
+            self.drift_factor,
+            self.threshold_factor,
+            self.min_distance,
+            self.sideways_threshold,
+            warm_up_period,
+        )
+
+        trends = np.where(
+            trend_codes == 1,
+            "Increasing",
+            np.where(trend_codes == -1, "Decreasing", "Sideways"),
+        )
+
+        return changepoints, trends, scores, window_size
+
+    import numba
+    import numpy as np
+
+    @staticmethod
+    @numba.njit(cache=True)
+    def _calculate_ewma(signal, window_size):
+        alpha = 2 / (window_size + 1)
+        ewma = np.zeros_like(signal)
+        ewma[0] = signal[0]
+        for i in range(1, len(signal)):
+            ewma[i] = alpha * signal[i] + (1 - alpha) * ewma[i - 1]
+        return ewma
+
+    @staticmethod
+    @numba.njit(cache=True)
+    def _calculate_rolling_z_scores(signal, ewma, window_size):
+        rolling_var = np.zeros_like(signal)
+        for i in range(window_size, len(signal)):
+            rolling_var[i] = np.var(signal[i - window_size + 1 : i + 1])
+        rolling_std = np.sqrt(rolling_var)
+        z_scores = (signal - ewma) / (rolling_std + 1e-8)
+        return z_scores
+
+    @staticmethod
+    @numba.njit(cache=True)
+    def _calculate_two_sided_cusum(
+        signal,
+        z,
+        drift,
+        threshold_factor,
+        min_distance,
+        sideways_threshold,
+        warm_up_period,
+    ):
+        cs_upper = np.zeros_like(z)
+        cs_lower = np.zeros_like(z)
+        changepoints = []
+        trend_codes = []
+        last_cp = -min_distance
+        for i in range(1, len(z)):
+            if i < warm_up_period:
+                continue
+            cs_upper[i] = max(0, cs_upper[i - 1] + z[i] - drift)
+            cs_lower[i] = min(0, cs_lower[i - 1] + z[i] + drift)
+            if i - last_cp >= min_distance:
+                if cs_upper[i] > threshold_factor * np.sqrt(i):
+                    changepoints.append(i)
+                    trend_codes.append(1)
+                    cs_upper[i] = 0
+                    cs_lower[i] = 0
+                    last_cp = i
+                elif cs_lower[i] < -threshold_factor * np.sqrt(i):
+                    changepoints.append(i)
+                    trend_codes.append(-1)
+                    cs_upper[i] = 0
+                    cs_lower[i] = 0
+                    last_cp = i
+                elif ( # This condition for sideways might lead to ZeroDivisionError if signal[last_cp] is 0. Preserved as per original.
+                    i > 0 
+                    and abs((signal[i] - signal[last_cp]) / signal[last_cp])
+                    < sideways_threshold
+                ):
+                    changepoints.append(i)
+                    trend_codes.append(0)
+                    cs_upper[i] = 0
+                    cs_lower[i] = 0
+                    last_cp = i
+        return (
+            np.maximum(cs_upper, -cs_lower),
+            np.array(changepoints),
+            np.array(trend_codes),
+        )
+
+def plot_cusum_results( # Exactly as in the first script
+    df_signal, signal_array, changepoints, trends, scores, ticker, feature
+):
+    fig = make_subplots(
+        rows=2,
+        cols=1,
+        shared_xaxes=True,
+        vertical_spacing=0.1,
+        subplot_titles=(f"{feature} Over Time for {ticker}", "CUSUM Score"),
+    )
+
+    fig.add_trace(
+        go.Scatter(
+            x=df_signal.index,
+            y=signal_array,
+            mode="lines",
+            line=dict(color="gold", width=2),
+            name=feature,
+        ),
+        row=1,
+        col=1,
+    )
+
+    fig.add_trace(
+        go.Scatter(
+            x=df_signal.index,
+            y=scores,
+            mode="lines",
+            line=dict(color="rgba(30, 144, 255, 0.7)", width=1.5),
+            name="CUSUM Score",
+        ),
+        row=2,
+        col=1,
+    )
+
+    colors = {"Increasing": "lime", "Decreasing": "red", "Sideways": "yellow"}
+    symbols = {
+        "Increasing": "triangle-up",
+        "Decreasing": "triangle-down",
+        "Sideways": "circle",
+    }
+
+    for trend in set(trends):
+        mask = trends == trend
+        # Original script might have index out of bounds if changepoints[mask] is empty or values are too large. Preserved.
+        fig.add_trace(
+            go.Scatter(
+                x=df_signal.index[changepoints[mask]],
+                y=signal_array[changepoints[mask]],
+                mode="markers",
+                marker=dict(color=colors[trend], size=10, symbol=symbols[trend]),
+                name=f"{trend}",
+            ),
+            row=1,
+            col=1,
+        )
+
+    fig.update_layout(
+        title=f"{feature} Change Point Detection for {ticker}",
+        plot_bgcolor="rgba(0,0,0,0.95)",
+        paper_bgcolor="rgba(0,0,0,0.95)",
+        font=dict(color="white", size=12),
+        autosize=True,
+        legend=dict(
+            orientation="h",
+            yanchor="bottom",
+            y=1.1,
+            xanchor="right",
+            x=1,
+            bgcolor="rgba(0,0,0,0.5)",
+            bordercolor="rgba(255,255,255,0.5)",
+        ),
+        showlegend=True,
+    )
+
+    fig.update_xaxes(title_text="Date", row=2, col=1, gridcolor="rgba(255,255,255,0.2)")
+    fig.update_yaxes(
+        title_text=feature, type="log", row=1, col=1, gridcolor="rgba(255,255,255,0.2)"
+    )
+    fig.update_yaxes(
+        title_text="CUSUM Score", row=2, col=1, gridcolor="rgba(255,255,255,0.2)"
+    )
+
+    for i, (cp, trend) in enumerate(zip(changepoints, trends)):
+        # Original script might have index out of bounds if cp is too large. Preserved.
+        fig.add_annotation(
+            x=df_signal.index[cp],
+            y=signal_array[cp],
+            text=f"{trend[0]} {i+1}",
+            showarrow=False,
+            yshift=10 if trend == "Increasing" else -10,
+            font=dict(color=colors[trend]),
+            row=1,
+            col=1,
+        )
+
+    return fig
+
+
+def run_cusum_dashboard(df_accounting, ticker=None, feature=None): # Modified for new port logic
+    if ticker is None:
+        ticker_options = df_accounting.index.get_level_values("ticker").unique()
+        if not ticker_options.empty:
+            ticker = ticker_options[0]
+        else:
+            print("Error: No tickers found in DataFrame for dashboard.")
+            return
+    if feature is None:
+        if not df_accounting.columns.empty:
+            feature = df_accounting.columns[0]
+        else:
+            print("Error: No features found in DataFrame for dashboard.")
+            return
+
+    app = dash.Dash(__name__)
+
+    dark_mode_colors = {
+        "background": "#1E1E1E",
+        "text": "#FFFFFF",
+        "plot_bg": "#2B2B2B",
+        "plot_gridlines": "#3A3A3A",
+    }
+
+    app.layout = html.Div(
+        [
+            html.H3(
+                "Financial Metric Change Point Detection",
+                style={
+                    "textAlign": "center",
+                    "marginBottom": "20px",
+                    "paddingTop": "10px",
+                    "color": dark_mode_colors["text"],
+                },
+            ),
+            html.Div(
+                [
+                    html.Div(
+                        [
+                            html.Label(
+                                "Select Ticker:",
+                                style={"color": dark_mode_colors["text"]},
+                            ),
+                            dcc.Dropdown(
+                                id="ticker-dropdown",
+                                options=[
+                                    {"label": t, "value": t}
+                                    for t in df_accounting.index.get_level_values(
+                                        "ticker"
+                                    ).unique()
+                                ],
+                                value=ticker,
+                            ),
+                        ],
+                        style={"width": "48%", "display": "inline-block"},
+                    ),
+                    html.Div(
+                        [
+                            html.Label(
+                                "Select Feature:",
+                                style={"color": dark_mode_colors["text"]},
+                            ),
+                            dcc.Dropdown(
+                                id="feature-dropdown",
+                                options=[
+                                    {"label": col, "value": col}
+                                    for col in df_accounting.columns
+                                ],
+                                value=feature,
+                            ),
+                        ],
+                        style={
+                            "width": "48%",
+                            "float": "right",
+                            "display": "inline-block",
+                        },
+                    ),
+                ],
+                style={"padding": "20px 0px"},
+            ),
+            dcc.Graph(id="cusum-plot", style={"height": "80vh"}),
+        ],
+        style={
+            "padding": "0 20px",
+            "margin": "0 auto",
+            "backgroundColor": dark_mode_colors["background"],
+        },
+    )
+
+    @app.callback(
+        Output("cusum-plot", "figure"),
+        [Input("ticker-dropdown", "value"), Input("feature-dropdown", "value")],
+    )
+    def update_graph(selected_ticker, selected_feature): # Exactly as in the first script's dashboard
+        df_signal = (
+            df_accounting.query(f"ticker == '{selected_ticker}'")
+            .reset_index()
+            .set_index("date")[selected_feature]
+        )
+        signal_array = df_signal.values
+        # Add a check for empty signal_array to prevent errors downstream
+        if len(signal_array) == 0:
+            fig = go.Figure()
+            fig.update_layout(
+                title_text=f"No data available for {selected_ticker} - {selected_feature}",
+                plot_bgcolor=dark_mode_colors["plot_bg"],
+                paper_bgcolor=dark_mode_colors["plot_bg"],
+                font_color=dark_mode_colors["text"],
+            )
+            return fig
+            
+        detector = ImprovedCusumDetector()
+        changepoints, trends, scores, window_size = detector.detect_changepoints(
+            signal_array
+        )
+        fig = plot_cusum_results(
+            df_signal,
+            signal_array,
+            changepoints,
+            trends,
+            scores,
+            selected_ticker,
+            selected_feature,
+        )
+        fig.update_layout(
+            plot_bgcolor=dark_mode_colors["plot_bg"],
+            paper_bgcolor=dark_mode_colors["plot_bg"],
+            font_color=dark_mode_colors["text"],
+        )
+        fig.update_xaxes(gridcolor=dark_mode_colors["plot_gridlines"])
+        fig.update_yaxes(gridcolor=dark_mode_colors["plot_gridlines"])
+        return fig
+
+    app_name = f"cumsum-app-{ticker}-{feature}".replace(" ","_") # Using more specific app name
+    
+    # --- Robust Server Launch (from the second script) ---
+    max_retries = 3
+    for attempt in range(max_retries):
+        port = get_unique_port(app_name)
+        try:
+            print(f"Attempting to launch Dash app '{app_name}' on port {port} (Attempt {attempt + 1}/{max_retries})")
+            # Using host='0.0.0.0' to make it accessible on the network
+            app.run_server(debug=False, port=port, host='0.0.0.0') 
+            print(f"Dash app '{app_name}' running on http://127.0.0.1:{port}/ or http://<your-ip>:{port}/")
+            break # Success
+        except socket.error as e:
+            if e.errno == errno.EADDRINUSE: # Address already in use
+                print(f"Port {port} is in use. Attempting to kill blocking process(es).")
+                if kill_process_on_port(port):
+                    print(f"Successfully killed process(es) on port {port}. Retrying Dash launch...")
+                    time.sleep(2) # Give OS a moment to free the port
+                else:
+                    print(f"Could not kill process(es) on port {port} or no process found. Trying a different port.")
+                    # Invalidate this port for this app_name so get_unique_port tries a new one
+                    if app_name in app_ports and app_ports[app_name] == port:
+                        del app_ports[app_name] 
+                
+                if attempt == max_retries - 1:
+                    print(f"Failed to launch Dash app '{app_name}' after {max_retries} attempts. Port {port} remains occupied.")
+                    # Not returning here, as the original run_server call was the return
+                    # This will effectively mean the function finishes if all retries fail.
+                    # To match original behavior of returning the server object, this structure would need adjustment
+                    # However, run_server is blocking, so returning it means the rest of the script doesn't run until server stops.
+                    # This current structure runs the server and blocks, or prints failure and continues if it can't.
+                    # The original `return app.run_server(...)` would also block.
+                    # If it fails to launch due to port, original would raise error immediately.
+                    # This structure tries to recover or fails after retries.
+                    # To make it fully equivalent on failure, we should re-raise here.
+                    raise RuntimeError(f"Failed to launch Dash app '{app_name}' after {max_retries} attempts. Port {port} remains occupied.")
+            else:
+                print(f"An unexpected socket error occurred: {e}")
+                raise # Re-raise other socket errors
+        except Exception as e:
+            print(f"An unexpected error occurred during app launch: {e}")
+            raise
+    else: # Executed if the loop completes without break (i.e., all retries failed before exception)
+         print(f"CRITICAL: Dash app '{app_name}' could not be started after {max_retries} attempts due to port issues without raising an exception in the loop (should not happen).")
+         # This else block for the for loop is unlikely to be hit if EADDRINUSE always occurs and leads to a raise on final attempt.
+         # If get_unique_port exhausted options and raised an error, that would be caught by the generic Exception.
+
+
+def perform_cusum_analysis(df_accounting, ticker=None, feature=None): # Exactly as in the first script
+    # Get the first ticker and feature if not provided
+    if ticker is None:
+        ticker = df_accounting.index.get_level_values("ticker").unique()[0]
+    if feature is None:
+        feature = df_accounting.columns[0]
+
+    # Extract the signal for the specified ticker and feature
+    df_signal = (
+        df_accounting.query(f"ticker == '{ticker}'")
+        .reset_index()
+        .set_index("date")[feature]
+    )
+
+    detector = ImprovedCusumDetector()
+    signal_array = df_signal.values
+
+    # Added basic check for empty signal to prevent error in detect_changepoints
+    if len(signal_array) == 0:
+        print(f"Warning: No data for {ticker} - {feature} in perform_cusum_analysis. Returning empty results.")
+        empty_analysis_df = pd.DataFrame(columns=["Date", "Trend", "Signal_Value", "CUSUM_Score", "Duration", "Duration_Days", "Percent_Change"])
+        empty_stats_df = pd.DataFrame({
+            "Statistic": [
+                "Ticker", "Feature", "Total_Change_Points", "Average_Duration_Days", 
+                "Average_Percent_Change", "Increasing_Trends", "Decreasing_Trends", 
+                "Sideways_Trends", "Max_CUSUM_Score", "Min_CUSUM_Score", "Window_Size"],
+            "Value": [ticker, feature, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+        })
+        empty_analysis_df.attrs["stats"] = empty_stats_df
+        return empty_analysis_df
+
+
+    changepoints, trends, scores, window_size = detector.detect_changepoints(
+        signal_array
+    )
+    
+    # Handle case with no changepoints to avoid errors
+    if len(changepoints) == 0:
+        analysis_df = pd.DataFrame(columns=["Date", "Trend", "Signal_Value", "CUSUM_Score"])
+    else:
+        # Original script might have index out of bounds or issues if changepoints are empty or too large. Preserved.
+        # CUSUM_Score indexing `changepoints - 1` could be problematic if changepoints contains 0.
+        # `_calculate_two_sided_cusum` starts `i` from 1, so changepoints should be >= 1.
+        # However, if warm_up_period is very large, changepoints could be empty.
+        analysis_df = pd.DataFrame(
+            {
+                "Date": df_signal.index[changepoints],
+                "Trend": trends,
+                "Signal_Value": signal_array[changepoints],
+                "CUSUM_Score": scores[changepoints - 1],
+            }
+        )
+
+    # Calculate duration between change points
+    if not analysis_df.empty:
+        analysis_df["Duration"] = (
+            analysis_df["Date"].diff().shift(-1).fillna(pd.Timedelta(days=0))
+        )
+        analysis_df["Duration_Days"] = analysis_df["Duration"].dt.days
+
+        # Calculate percentage change between change points
+        analysis_df["Percent_Change"] = (
+            analysis_df["Signal_Value"].pct_change().shift(-1) * 100
+        ).fillna(0)
+    else: # Ensure columns exist if analysis_df is empty
+        analysis_df["Duration"] = pd.Series(dtype='timedelta64[ns]')
+        analysis_df["Duration_Days"] = pd.Series(dtype='int')
+        analysis_df["Percent_Change"] = pd.Series(dtype='float')
+
+
+    # Overall statistics
+    # Original `scores[scores > 0].min()` could fail if scores[scores > 0] is empty.
+    min_cusum_score_val = 0
+    if any(scores > 0):
+        min_positive_scores = scores[scores > 0]
+        if len(min_positive_scores) > 0:
+            min_cusum_score_val = min_positive_scores.min()
+            
+    overall_stats = pd.DataFrame(
+        {
+            "Statistic": [
+                "Ticker",
+                "Feature",
+                "Total_Change_Points",
+                "Average_Duration_Days",
+                "Average_Percent_Change",
+                "Increasing_Trends",
+                "Decreasing_Trends",
+                "Sideways_Trends",
+                "Max_CUSUM_Score",
+                "Min_CUSUM_Score", # Original name was Min_CUSUM_Score
+                "Window_Size",
+            ],
+            "Value": [
+                ticker,
+                feature,
+                len(changepoints),
+                analysis_df["Duration_Days"].mean() if not analysis_df.empty else 0,
+                analysis_df["Percent_Change"].mean() if not analysis_df.empty else 0,
+                sum(trends == "Increasing") if len(trends) > 0 else 0,
+                sum(trends == "Decreasing") if len(trends) > 0 else 0,
+                sum(trends == "Sideways") if len(trends) > 0 else 0,
+                scores.max() if len(scores) > 0 else 0,
+                min_cusum_score_val,
+                window_size,
+            ],
+        }
+    )
+    # Fill NaN for mean values if analysis_df was empty or had single row for stats
+    overall_stats['Value'] = overall_stats['Value'].fillna(0)
+
+    analysis_df.attrs["stats"] = overall_stats
+
+    return analysis_df
+
+
+if __name__ == "__main__":
+    # Create a dummy df_accounting for demonstration as in the second script
+    # This helps in testing the updated script functionality.
+    dates = pd.to_datetime(pd.date_range(start="2020-01-01", periods=300, freq="B"))
+    data_aapl = np.random.lognormal(mean=0.001, sigma=0.02, size=300).cumprod() * 100
+    data_msft = np.random.lognormal(mean=0.0005, sigma=0.025, size=300).cumprod() * 150
+    
+    data_aapl[100:150] *= 1.3 
+    data_aapl[200:250] *= 0.7 
+    
+    data_msft[50:80] *= 0.8   
+    data_msft[150:200] *= 1.5 
+    data_msft[250:] += 20 
+
+    df_aapl = pd.DataFrame({"Close": data_aapl, "Volume": np.random.randint(1000, 5000, 300)}, index=dates)
+    df_aapl["ticker"] = "AAPL"
+    
+    df_msft = pd.DataFrame({"Close": data_msft, "Volume": np.random.randint(2000, 7000, 300)}, index=dates)
+    df_msft["ticker"] = "MSFT"
+
+    # Ensure df_accounting structure matches what functions expect (MultiIndex with 'ticker', 'date')
+    df_accounting_demo = pd.concat([
+        df_aapl.reset_index().rename(columns={'index': 'date'}).set_index(['ticker', 'date']),
+        df_msft.reset_index().rename(columns={'index': 'date'}).set_index(['ticker', 'date'])
+    ])
+
+    print("---- Running CUSUM Analysis for AAPL Close ----")
+    analysis_results_aapl = perform_cusum_analysis(df_accounting_demo, ticker="AAPL", feature="Close")
+    if not analysis_results_aapl.empty: # Check if DataFrame is not empty
+        print(analysis_results_aapl)
+        print("\nOverall Stats (AAPL Close):")
+        print(analysis_results_aapl.attrs["stats"])
+    else:
+        print("No analysis results for AAPL Close.")
+
+
+    print("\n---- Running CUSUM Analysis for MSFT Volume ----")
+    analysis_results_msft_vol = perform_cusum_analysis(df_accounting_demo, ticker="MSFT", feature="Volume")
+    if not analysis_results_msft_vol.empty: # Check if DataFrame is not empty
+        print(analysis_results_msft_vol)
+        print("\nOverall Stats (MSFT Volume):")
+        print(analysis_results_msft_vol.attrs["stats"])
+    else:
+        print("No analysis results for MSFT Volume.")
+    
+    print("\n---- Launching Dashboard ----")
+    print("Note: If a port is busy, the script will attempt to kill the process.")
+    print("Ensure you are comfortable with this, especially if not in a dev environment.")
+    
+    try:
+        # df_accounting_demo needs to be passed to the dashboard
+        run_cusum_dashboard(df_accounting_demo, ticker="AAPL", feature="Close")
+    except Exception as e:
+        print(f"Dashboard launch failed: {e}")

{sovai-0.2.63 → sovai-0.2.65}/sovai/extensions/clustering.py

@@ -151,6 +151,9 @@ def segment_series(df_accounting, features_select=None):
 
 
 def cluster(df_mega, features_select=None):
+
+    df_mega = df_mega[~df_mega.index.duplicated(keep="first")]  # ← NEW LINE
+
     days = df_mega.index.get_level_values("date").nunique()
 
     (
@@ -205,7 +208,7 @@ def cluster(df_mega, features_select=None):
     output_df["labels"] = output_df["labels"].apply(lambda x: f"Centroid {int(x)}")
 
     # Group by 'ticker' and sum across rows
-    summed_series = output_df.groupby("ticker").sum().sum(axis=1)
+    summed_series = output_df.groupby("ticker").sum(numeric_only=True).sum(axis=1)
 
     # Find the indices where the value is 0.00
     zero_indices = summed_series[summed_series == 0.00].index.tolist()

{sovai-0.2.63 → sovai-0.2.65}/sovai/extensions/dimensionality_reduction.py

@@ -4,7 +4,7 @@ from sklearn.decomposition import PCA, TruncatedSVD, FactorAnalysis
 from sklearn.random_projection import GaussianRandomProjection
 from sklearn.preprocessing import StandardScaler
 from sklearn.impute import SimpleImputer
-# import umap
+
 import warnings
 
 
@@ -124,6 +124,7 @@ def dimensionality_reduction(df, method, explained_variance=0.95, n_components=N
         "truncated_svd": TruncatedSVD,
         "factor_analysis": FactorAnalysis,
         "gaussian_random_projection": GaussianRandomProjection,
+        
     }
     
     if n_components is not None: