app.py

import joblib
import pandas as pd
import gradio as gr

# --- Given weights ---
weights = {
    "GradientBoosting": 0.239,
    "RandomForest": 0.573,
    "XGBoost": 0.188
}

# --- Given model paths (place models in the same directory as app.py for HF Spaces) ---
model_paths = {
    "GradientBoosting": "main/GradientBoosting_model.pkl",
    "RandomForest": "main/RandomForest_model.pkl",
    "XGBoost": "main/xgb_model.pkl"
}

# --- Load models ---
models = {name: joblib.load(path) for name, path in model_paths.items()}

# --- Prediction function ---
def predict(
    dep_delay_x_congestion, dep_congestion, block_time_diff, enroute_delay,
    dep_delay_rolling_mean, dep_congestion_roll3h, route_hour_delay_mean,
    taxi_in_ratio, taxi_out_ratio, arr_congestion_roll3h, month_cos,
    pres_delta, wspd_delta, arr_congestion, month_sin, season_wind,
    season, distance_partofday, part_of_day_ord, segment_peak_hours,
    wind_speed_cat_ord, humidity_cat_ord, prcp_delta, pressure_cat_ord,
    distance_bin_ord
):
    # Prepare input DataFrame
    X_input = pd.DataFrame([{
        'dep_delay_x_congestion': dep_delay_x_congestion,
        'dep_congestion': dep_congestion,
        'block_time_diff': block_time_diff,
        'enroute_delay': enroute_delay,
        'dep_delay_rolling_mean': dep_delay_rolling_mean,
        'dep_congestion_roll3h': dep_congestion_roll3h,
        'route_hour_delay_mean': route_hour_delay_mean,
        'taxi_in_ratio': taxi_in_ratio,
        'taxi_out_ratio': taxi_out_ratio,
        'arr_congestion_roll3h': arr_congestion_roll3h,
        'month_cos': month_cos,
        'pres_delta': pres_delta,
        'wspd_delta': wspd_delta,
        'arr_congestion': arr_congestion,
        'month_sin': month_sin,
        'season_wind': season_wind,
        'season': season,
        'distance_partofday': distance_partofday,
        'part_of_day_ord': part_of_day_ord,
        'segment_peak_hours': segment_peak_hours,
        'wind_speed_cat_ord': wind_speed_cat_ord,
        'humidity_cat_ord': humidity_cat_ord,
        'prcp_delta': prcp_delta,
        'pressure_cat_ord': pressure_cat_ord,
        'distance_bin_ord': distance_bin_ord
    }])

    # Get predictions from each model
    preds = {name: models[name].predict(X_input)[0] for name in models}

    # Weighted ensemble prediction
    final_pred = sum(preds[name] * weights[name] for name in preds)

    # Categorize delay
    if 20 <= final_pred < 30:
        delay_category = "Minimal Delay"
    elif 30 <= final_pred <= 60:
        delay_category = "Moderate Delay"
    elif final_pred > 60:
        delay_category = "Excessive Delay"
    else:
        delay_category = "No Significant Delay"

    return preds, final_pred, delay_category

# --- Gradio interface ---
inputs = [
    gr.Number(label="Departure Delay × Congestion Index"),
    gr.Number(label="Departure Congestion"),
    gr.Number(label="Block Time Difference (min)"),
    gr.Number(label="En‑Route Delay (min)"),
    gr.Number(label="Avg Departure Delay (last 3 flights)"),
    gr.Number(label="Departure Congestion (3‑hour rolling)"),
    gr.Number(label="Avg Delay for Route & Hour"),
    gr.Number(label="Taxi‑In Time ÷ Block Time"),
    gr.Number(label="Taxi‑Out Time ÷ Block Time"),
    gr.Number(label="Arrival Congestion (3‑hour rolling)"),
    gr.Number(label="Month (cosine encoding)"),
    gr.Number(label="Pressure Change (Dest − Origin)"),
    gr.Number(label="Wind Speed Change (Dest − Origin)"),
    gr.Number(label="Arrival Congestion"),
    gr.Number(label="Month (sine encoding)"),
    gr.Number(label="Seasonal Wind Category"),
    gr.Number(label="Season (ordinal)"),
    gr.Number(label="Distance × Part of Day"),
    gr.Number(label="Part of Day (ordinal)"),
    gr.Number(label="Peak Hour Segment Flag"),
    gr.Number(label="Wind Speed Category (ordinal)"),
    gr.Number(label="Humidity Category (ordinal)"),
    gr.Number(label="Precipitation Change (Dest − Origin)"),
    gr.Number(label="Pressure Category (ordinal)"),
    gr.Number(label="Distance Bin (ordinal)")
]

outputs = [
    gr.JSON(label="Model Predictions"),
    gr.Number(label="Weighted Ensemble Prediction (minutes)"),
    gr.Textbox(label="Delay Category")
]

demo = gr.Interface(
    fn=predict,
    inputs=inputs,
    outputs=outputs,
    title="Flight Delay Prediction (Weighted Ensemble)",
    description="Enter flight features to get predictions from GradientBoosting, RandomForest, and XGBoost, plus a weighted ensemble result and delay category."
)

if __name__ == "__main__":
    demo.launch()