train_model.py

import pandas as pd
import numpy as np
from sklearn.model_selection import LeaveOneOut
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_absolute_error, mean_squared_error
import matplotlib.pyplot as plt
import os

# Updated paths
DATA_PATH = r"C:\Users\AUSTIN PAUL\OneDrive\Desktop\glucose_training_data.csv"

OUTPUT_DIR = r"C:\Users\AUSTIN PAUL\OneDrive\Desktop\glucose_training_data\model_output"

os.makedirs(OUTPUT_DIR, exist_ok=True)

# 1. Load data
if not os.path.exists(DATA_PATH):
    raise FileNotFoundError(f"Could not find CSV at {DATA_PATH}")

data = pd.read_csv(DATA_PATH)

# Clean up column names (handles both ir_red_ratio and ratio)
data.columns = data.columns.str.strip().str.lower()
if "ir_red_ratio" in data.columns:
    data = data.rename(columns={"ir_red_ratio": "ratio"})
if "reference_glucose" in data.columns:
    data = data.rename(columns={"reference_glucose": "glucose"})

# Filter required columns and remove outliers
data = data[["ratio", "variability", "slope", "glucose"]]
data = data.dropna()
data = data[(data["glucose"] >= 40) & (data["glucose"] <= 400)]

X = data[["ratio", "variability", "slope"]].values
y = data["glucose"].values

# 2. Leave-One-Out Cross-Validation
loo = LeaveOneOut()
scaler = StandardScaler()

y_true = []
y_pred = []

print("Running Cross-Validation...")
for train_idx, test_idx in loo.split(X):
    X_train, X_test = X[train_idx], X[test_idx]
    y_train, y_test = y[train_idx], y[test_idx]

    # Standardize based on training set
    X_train_scaled = scaler.fit_transform(X_train)
    X_test_scaled = scaler.transform(X_test)

    model = LinearRegression()
    model.fit(X_train_scaled, y_train)

    y_pred.append(model.predict(X_test_scaled)[0])
    y_true.append(y_test[0])

# 3. Calculate Performance Metrics
mae = mean_absolute_error(y_true, y_pred)
rmse = np.sqrt(mean_squared_error(y_true, y_pred))
mard = np.mean(np.abs((np.array(y_true) - np.array(y_pred)) / np.array(y_true))) * 100

print("\n===== MODEL PERFORMANCE =====")
print(f"MAE   : {mae:.2f} mg/dL")
print(f"RMSE  : {rmse:.2f} mg/dL")
print(f"MARD  : {mard:.2f} %")
print(f"Within ±20 mg/dL : {np.mean(np.abs(np.array(y_true) - np.array(y_pred)) < 20)*100:.1f}%")

# 4. Train Final Model and Save Scaling Parameters
# IMPORTANT: Arduino needs the Mean and Scale to normalize real-time data
scaler_final = StandardScaler()
X_scaled = scaler_final.fit_transform(X)

final_model = LinearRegression()
final_model.fit(X_scaled, y)

# 5. Export to Arduino Header (.h)
with open(os.path.join(OUTPUT_DIR, "model_coefficients.h"), "w") as f:
    f.write("#ifndef MODEL_COEFFICIENTS_H\n#define MODEL_COEFFICIENTS_H\n\n")
    
    # Regression Coefficients
    f.write(f"const float intercept = {final_model.intercept_:.6f};\n")
    f.write(f"const float coeff_ratio = {final_model.coef_[0]:.6f};\n")
    f.write(f"const float coeff_variability = {final_model.coef_[1]:.6f};\n")
    f.write(f"const float coeff_slope = {final_model.coef_[2]:.6f};\n\n")
    
    # Scaling Parameters (Essential for real-time prediction)
    f.write(f"const float mean_ratio = {scaler_final.mean_[0]:.6f};\n")
    f.write(f"const float mean_variability = {scaler_final.mean_[1]:.6f};\n")
    f.write(f"const float mean_slope = {scaler_final.mean_[2]:.6f};\n\n")
    
    f.write(f"const float std_ratio = {scaler_final.scale_[0]:.6f};\n")
    f.write(f"const float std_variability = {scaler_final.scale_[1]:.6f};\n")
    f.write(f"const float std_slope = {scaler_final.scale_[2]:.6f};\n")
    
    f.write("\n#endif\n")

# 6. Plotting
plt.figure(figsize=(8,6))
plt.scatter(y_true, y_pred, alpha=0.6, edgecolors='k')
plt.plot([min(y_true), max(y_true)], [min(y_true), max(y_true)], 'r--', label="Perfect Prediction")
plt.xlabel("Reference Glucose (mg/dL)")
plt.ylabel("Predicted Glucose (mg/dL)")
plt.title("Leave-One-Out Validation Results")
plt.legend()
plt.grid(True)
plt.savefig(os.path.join(OUTPUT_DIR, "prediction_plot.png"))
plt.close()

print(f"\nTraining complete. Header and Plot saved to: {OUTPUT_DIR}")