prediction_baseline.py

import pandas as pd

df = pd.read_csv("data/tester.csv")

# naive baseline:
# predict unstable if boundary is very small or drift is strongly positive
df["prediction"] = (
    (df["boundary_distance"] <= 0.08) |
    (df["drift_gradient"] >= 0.07)
).astype(int)

# optional intervention direction baseline
def predict_direction(row):
    if pd.isna(row["boundary_distance_before"]) or pd.isna(row["boundary_distance_after"]):
        return None
    if row["boundary_distance_after"] > row["boundary_distance_before"]:
        return 1
    if row["boundary_distance_after"] < row["boundary_distance_before"]:
        return -1
    return 0

df["intervention_effect_direction"] = df.apply(predict_direction, axis=1)

out = df[["scenario_id", "prediction", "intervention_effect_direction"]]
out.to_csv("predictions.csv", index=False)

print("Baseline predictions written to predictions.csv")
print(f"rows: {len(out)}")