forward_test_nn.py

from pathlib import Path
import json
import joblib
import pandas as pd
import numpy as np

from tensorflow.keras.models import load_model

BASE_DIR = Path(__file__).resolve().parent
OUT_DIR = BASE_DIR / "outputs"

PREPROCESSOR_PATH = OUT_DIR / "nn_preprocessor_label_1to1.joblib"
MODEL_PATH = OUT_DIR / "nn_label_1to1.keras"
DATA_PATH = OUT_DIR / "ml_dataset_exact_all_v2_2026-03-02_to_2026-03-06.csv"

OUT_SCORED_PATH = OUT_DIR / "forward_scored_nn_2026-03-02_to_2026-03-06.csv"
OUT_SUMMARY_PATH = OUT_DIR / "forward_summary_nn_2026-03-02_to_2026-03-06.json"

THRESHOLDS = [0.50, 0.55]

DROP_COLS_ALWAYS = [
    "trade_key",
    "label_1to1",
    "label_1to2",
    "bt_buy_signal_time",
    "bt_sell_signal_time",
    "bt_buy_time",
    "bt_buy_price",
    "bt_stop_loss",
    "bt_target_1",
    "bt_target_2",
    "bt_qty_per_lot",
    "bt_capital_per_lot",
    "bt_stop_loss_amt_per_lot",
    "signal_time",
    "confirmation_time",
    "indication_time",
    "buy_time",
]

OPTIONAL_DROP_COLS = [
    "exit_status",
    "option_symbol",
    "trade_side",
]


def build_feature_matrix(df: pd.DataFrame):
    drop_cols = [c for c in DROP_COLS_ALWAYS if c in df.columns]
    drop_cols += [c for c in OPTIONAL_DROP_COLS if c in df.columns]

    X = df.drop(columns=drop_cols, errors="ignore").copy()

    if "sector" in X.columns:
        X["sector"] = X["sector"].fillna("UNKNOWN").replace("", "UNKNOWN")

    return X


def summarize_for_threshold(df: pd.DataFrame, threshold: float):
    kept = df[df["y_prob"] >= threshold].copy()
    total = len(df)
    kept_n = len(kept)

    baseline_hit_rate = float(df["label_1to1"].mean()) if total else 0.0

    if kept_n == 0:
        return {
            "threshold": threshold,
            "total_trades": total,
            "kept_trades": 0,
            "coverage_pct": 0.0,
            "baseline_hit_rate_1to1": round(baseline_hit_rate, 4),
            "kept_hit_rate_1to1": None,
            "gross_pnl_1to1_per_lot_sum": 0.0,
            "avg_pnl_1to1_per_lot": None,
            "daily_breakdown": [],
        }

    # For 1:1, profit magnitude equals stop-loss magnitude
    kept["realized_pnl_1to1_per_lot"] = np.where(
        kept["label_1to1"] == 1,
        kept["bt_stop_loss_amt_per_lot"],
        -kept["bt_stop_loss_amt_per_lot"],
    )

    hit_rate_1to1 = float(kept["label_1to1"].mean())
    gross_pnl = float(kept["realized_pnl_1to1_per_lot"].sum())
    avg_pnl = float(kept["realized_pnl_1to1_per_lot"].mean())

    by_day = (
        kept.groupby("trade_date")
        .agg(
            trades=("label_1to1", "size"),
            wins=("label_1to1", "sum"),
            pnl_1to1_per_lot=("realized_pnl_1to1_per_lot", "sum"),
        )
        .reset_index()
    )

    by_day["trade_date"] = pd.to_datetime(by_day["trade_date"], errors="coerce").dt.strftime("%Y-%m-%d")

    return {
        "threshold": threshold,
        "total_trades": total,
        "kept_trades": kept_n,
        "coverage_pct": round((kept_n / total) * 100, 2),
        "baseline_hit_rate_1to1": round(baseline_hit_rate, 4),
        "kept_hit_rate_1to1": round(hit_rate_1to1, 4),
        "gross_pnl_1to1_per_lot_sum": round(gross_pnl, 2),
        "avg_pnl_1to1_per_lot": round(avg_pnl, 2),
        "daily_breakdown": by_day.to_dict(orient="records"),
    }


def main():
    preprocessor = joblib.load(PREPROCESSOR_PATH)
    model = load_model(MODEL_PATH)

    df = pd.read_csv(DATA_PATH)
    df["trade_date"] = pd.to_datetime(df["trade_date"], errors="coerce")

    X_raw = build_feature_matrix(df)
    X = preprocessor.transform(X_raw)

    if hasattr(X, "toarray"):
        X = X.toarray()

    df["y_prob"] = model.predict(X).ravel()
    df.to_csv(OUT_SCORED_PATH, index=False)

    summaries = [summarize_for_threshold(df, th) for th in THRESHOLDS]

    with open(OUT_SUMMARY_PATH, "w") as f:
        json.dump(summaries, f, indent=2)

    print(f"Saved scored trades to: {OUT_SCORED_PATH}")
    print(f"Saved summary to: {OUT_SUMMARY_PATH}")

    for s in summaries:
        print("\n", s)


if __name__ == "__main__":
    main()