from fastapi import FastAPI from pydantic import BaseModel import joblib import pandas as pd import numpy as np import json scaler = joblib.load("scaler.joblib") model = joblib.load("fatigue_model.joblib") with open("model_metadata.json") as f: metadata = json.load(f) feature_cols = metadata["feature_cols"] base_feature_cols = [ "AVRR", "SDNN", "RMSSD", "PNN50", "Coefficient_of_Variation", "Age", "Weight", "Height" ] engineered_cols = [ "RMSSD_SDNN_ratio", "HRV_index", "Stress_index", "Parasympathetic", "AVRR_PNN50", "CV_SDNN", "RMSSD_sq", "SDNN_sq" ] class FatigueInput(BaseModel): AVRR: float SDNN: float RMSSD: float PNN50: float Coefficient_of_Variation: float Age: float Weight: float Height: float def build_features(data: dict) -> pd.DataFrame: """Compute engineered features from the 8 base inputs the backend sends.""" row = dict(data) row["RMSSD_SDNN_ratio"] = row["RMSSD"] / (abs(row["SDNN"]) + 0.001) row["HRV_index"] = (row["SDNN"] + row["RMSSD"]) / 2 row["Stress_index"] = row["AVRR"] / (abs(row["SDNN"]) + 0.001) row["Parasympathetic"] = row["RMSSD"] * row["PNN50"] row["AVRR_PNN50"] = row["AVRR"] * row["PNN50"] row["CV_SDNN"] = row["Coefficient_of_Variation"] * row["SDNN"] row["RMSSD_sq"] = row["RMSSD"] ** 2 row["SDNN_sq"] = row["SDNN"] ** 2 return pd.DataFrame([row])[feature_cols] app = FastAPI() @app.post("/predict") def predict(input_data: FatigueInput): try: input_df = build_features(input_data.dict()) scaled_input = scaler.transform(input_df) prediction = int(model.predict(scaled_input)[0]) prob_dict = {} if hasattr(model, "predict_proba"): proba = model.predict_proba(scaled_input)[0] classes = model.classes_ prob_dict = {int(cls): float(prob) for cls, prob in zip(classes, proba)} return { "fatigue_level": prediction, "fatigueLevel": prediction, "cluster_probabilities": list(prob_dict.values()), "rounded_level": prediction, "confidence": prob_dict, } except Exception as e: return {"error": str(e)} @app.post("/predictGMM") def predict_gmm(input_data: FatigueInput): """Legacy endpoint -- redirects to the supervised classifier.""" return predict(input_data)