IDKHowToCodeFr/tinyml-backend

1from typing import Dict, Any2import pandas as pd3import numpy as np4from preprocessing import preprocess_data5 6def evaluate(ensemble_model, data) -> Dict[str, Any]:7    # Domain orchestrator: PatientData -> InferenceResult8    raw_df = pd.DataFrame([data.model_dump(by_alias=True)])9    10    # Mapping properties to expected feature names11    raw_df = raw_df.rename(columns={12        "Heart_Rate": "Heart Rate (bpm)",13        "SpO2_Level": "SpO2 Level (%)",14        "Systolic_BP": "Systolic Blood Pressure (mmHg)",15        "Diastolic_BP": "Diastolic Blood Pressure (mmHg)",16        "Body_Temp": "Body Temperature (°C)",17        "Fall_Detection": "Fall Detection"18    })19 20    # Drop anything the model hasn't seen21    unseen = ["Physical Activity Level", "Age", "Gender", "Stress Level", "Cholesterol Level (mg/dL)", 22              "Physical_Activity_Level", "Stress_Level", "Cholesterol_Level", "Fall_Detection"]23    for f in unseen:24        if f in raw_df.columns:25            raw_df = raw_df.drop(columns=[f])26 27    # Feature engineering28    processed_df, _ = preprocess_data(raw_df, is_training=False)29 30    if ensemble_model is None:31        return {"error": "Ensemble model not loaded"}32 33    # Ensemble Prediction34    disease_label, conf, ind_preds, class_probs, weights, ind_conf = ensemble_model.predict(processed_df)35    is_at_risk = (disease_label.lower() not in ["healthy", "normal", "none"])36 37    return {38        "prediction": is_at_risk,39        "prediction_label": disease_label,40        "probability": float(conf),41        "ensemble_prediction": is_at_risk,42        "model_outputs": ind_preds,43        "model_probs": ind_conf,44        "disease_probs": class_probs,45        "weights": weights46    }47