src/features/price_pressure.py

"""
src/features/price_pressure.py
================================
Computes the Fuel Price Pressure Indicator — a 0–100 score that
captures the lagged and rolling impact of oil prices on aviation costs.

Also computes rolling oil price features used as regression inputs.
"""

import pandas as pd
import numpy as np
from src.utils.logger import get_logger
from config.settings import PROCESSED_DIR, OIL_ROLLING_DAYS

logger = get_logger(__name__)


def compute_fuel_pressure(df: pd.DataFrame) -> pd.DataFrame:
    """
    Compute fuel_pressure_indicator for a DataFrame that contains oil_price
    and oil_price_change_pct columns.

    Formula:
        FPI = clip(
            (pct_change_norm * 60) + (disruption_impact * 20) + (rolling_trend * 20),
            0, 100
        )
    """
    df = df.copy()

    pct_chg = pd.to_numeric(df.get("oil_price_change_pct", 0), errors="coerce").fillna(0)
    disruption = pd.to_numeric(df.get("disruption_index", 0), errors="coerce").fillna(0)

    # Normalise pct change: 10% change → 60 pressure units
    pct_norm = (pct_chg / 10).clip(-1, 3) * 60

    # Disruption contribution (0-20)
    disruption_contrib = (disruption / 100) * 20

    df["fuel_pressure_indicator"] = (pct_norm + disruption_contrib).clip(0, 100).round(2)

    return df


def enrich_oil_features(oil_df: pd.DataFrame) -> pd.DataFrame:
    """
    Add rolling average columns to an oil price DataFrame.
    Expects 'brent_usd' and 'date' columns.
    """
    df = oil_df.copy().sort_values("date")

    for w in OIL_ROLLING_DAYS:
        df[f"brent_rolling_{w}d"] = (
            df["brent_usd"].rolling(w, min_periods=1).mean().round(2)
        )

    df["oil_volatility_14d"] = (
        df["brent_usd"].rolling(14, min_periods=2).std().round(2)
    )

    df["brent_above_90"] = (df["brent_usd"] > 90).astype(int)
    df["brent_above_100"] = (df["brent_usd"] > 100).astype(int)

    return df


def run(df: pd.DataFrame = None) -> pd.DataFrame:
    if df is None:
        df = pd.read_csv(PROCESSED_DIR / "flight_disruptions.csv", low_memory=False)
    return compute_fuel_pressure(df)


if __name__ == "__main__":
    oil_df = pd.read_csv(PROCESSED_DIR / "oil_prices.csv")
    print(enrich_oil_features(oil_df).tail(10).to_string(index=False))