feat : Add TARDIS weather model training notebook with data loading, …

scripts/model/train_meteo_aux.py

@@ -0,0 +1,141 @@
+import json
+import os
+from pathlib import Path
+
+import joblib
+import pandas as pd
+from sklearn.compose import ColumnTransformer
+from sklearn.ensemble import HistGradientBoostingRegressor
+from sklearn.impute import SimpleImputer
+from sklearn.metrics import mean_absolute_error, r2_score, root_mean_squared_error
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import Pipeline
+from sklearn.preprocessing import OrdinalEncoder, StandardScaler
+
+
+RANDOM_STATE = 42
+
+NUMERIC_FEATURES = ["LAT", "LON", "ALTI", "year", "month", "day_of_week", "day_of_year"]
+CATEGORICAL_FEATURES = ["NUM_POSTE", "NOM_USUEL", "season"]
+ALL_FEATURES = NUMERIC_FEATURES + CATEGORICAL_FEATURES
+
+
+def compute_metrics(y_true, y_pred) -> dict:
+    return {
+        "RMSE": root_mean_squared_error(y_true, y_pred),
+        "MAE": mean_absolute_error(y_true, y_pred),
+        "R2": r2_score(y_true, y_pred),
+    }
+
+
+def build_preprocessor() -> ColumnTransformer:
+    return ColumnTransformer(
+        transformers=[
+            (
+                "num",
+                Pipeline(
+                    [
+                        ("imputer", SimpleImputer(strategy="median")),
+                        ("scaler", StandardScaler()),
+                    ]
+                ),
+                NUMERIC_FEATURES,
+            ),
+            (
+                "cat",
+                Pipeline(
+                    [
+                        ("imputer", SimpleImputer(strategy="most_frequent")),
+                        (
+                            "ordinal",
+                            OrdinalEncoder(
+                                handle_unknown="use_encoded_value",
+                                unknown_value=-1,
+                            ),
+                        ),
+                    ]
+                ),
+                CATEGORICAL_FEATURES,
+            ),
+        ]
+    )
+
+
+def train_one(df: pd.DataFrame, target: str, out_path: Path, max_rows: int) -> dict:
+    df_model = df.dropna(subset=[target]).copy()
+    if len(df_model) > max_rows:
+        df_model = df_model.sample(max_rows, random_state=RANDOM_STATE)
+
+    X = df_model[ALL_FEATURES]
+    y = df_model[target]
+    X_train, X_test, y_train, y_test = train_test_split(
+        X, y, test_size=0.2, random_state=RANDOM_STATE
+    )
+
+    model = HistGradientBoostingRegressor(
+        max_iter=220,
+        learning_rate=0.06,
+        max_leaf_nodes=31,
+        l2_regularization=0.02,
+        random_state=RANDOM_STATE,
+    )
+    pipeline = Pipeline([("prep", build_preprocessor()), ("model", model)])
+    pipeline.fit(X_train, y_train)
+    metrics = compute_metrics(y_test, pipeline.predict(X_test))
+
+    joblib.dump(
+        {
+            "pipeline": pipeline,
+            "target": target,
+            "features": ALL_FEATURES,
+            "metrics": metrics,
+        },
+        out_path,
+    )
+    return metrics
+
+
+def run_training(
+    dataset_path: str = "data/processed/meteo/cleaned_meteo_dataset.csv",
+    out_dir: str = "models/meteo_aux",
+    max_rows: int = 250_000,
+) -> dict:
+    usecols = [
+        "NUM_POSTE",
+        "NOM_USUEL",
+        "LAT",
+        "LON",
+        "ALTI",
+        "RR",
+        "FFM",
+        "date",
+        "year",
+        "month",
+        "season",
+    ]
+    df = pd.read_csv(dataset_path, usecols=usecols, parse_dates=["date"], low_memory=False)
+    df["day_of_week"] = df["date"].dt.dayofweek
+    df["day_of_year"] = df["date"].dt.dayofyear
+
+    out = Path(out_dir)
+    out.mkdir(parents=True, exist_ok=True)
+
+    results = {
+        "RR": train_one(df, "RR", out / "rr_model.joblib", max_rows=max_rows),
+        "FFM": train_one(df, "FFM", out / "ffm_model.joblib", max_rows=max_rows),
+    }
+    with open(out / "metadata.json", "w") as f:
+        json.dump(results, f, indent=2)
+
+    for target, metrics in results.items():
+        print(
+            f"{target:3s} | RMSE: {metrics['RMSE']:.3f} "
+            f"| MAE: {metrics['MAE']:.3f} | R2: {metrics['R2']:.4f}"
+        )
+    print(f"Saved auxiliary weather models in {out_dir}/")
+    return results
+
+
+if __name__ == "__main__":
+    max_rows = int(os.environ.get("METEO_AUX_MAX_ROWS", "250000"))
+    run_training(max_rows=max_rows)