backend/evaluate_model.py

import os
import sys
import numpy as np
import pickle
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, accuracy_score, f1_score
from tensorflow.keras.models import load_model
from tensorflow.keras.utils import to_categorical

# Data Paths
BASE_DIR = os.path.dirname(os.path.abspath(__file__)) # This is backend/
ROOT_DIR = os.path.dirname(BASE_DIR) # This is project root
DATA_DIR = os.path.join(ROOT_DIR, "data sets")
MODEL_PATH = os.path.join(ROOT_DIR, "model.h5")
ENCODER_PATH = os.path.join(ROOT_DIR, "encoder.pkl")
FEATURES_PATH = os.path.join(DATA_DIR, "features_cache.npy")
LABELS_PATH = os.path.join(DATA_DIR, "labels_cache.npy")

def evaluate():
    print("Loading data from cache...")
    if not os.path.exists(FEATURES_PATH) or not os.path.exists(LABELS_PATH):
        print("Error: Cached features not found. Please train the model first.")
        return

    X = np.load(FEATURES_PATH)
    y = np.load(LABELS_PATH)
    
    print(f"Loaded {len(X)} samples.")

    print("Loading Label Encoder...")
    with open(ENCODER_PATH, 'rb') as f:
        le = pickle.load(f)
    
    # Encode labels (Same logic as training)
    y_encoded = to_categorical(le.fit_transform(y))
    
    # Split (Same random_state as training to ensure same test set)
    print("Splitting data (random_state=42)...")
    X_train, X_test, y_train, y_test = train_test_split(X, y_encoded, test_size=0.2, random_state=42, stratify=y)
    
    print(f"Test Set Size: {len(X_test)}")

    print("Loading Model...")
    model = load_model(MODEL_PATH)
    
    print("Evaluating...")
    # Predict
    y_pred_prob = model.predict(X_test, verbose=0)
    y_pred = np.argmax(y_pred_prob, axis=1)
    y_true = np.argmax(y_test, axis=1)
    
    # Calculate Metrics
    accuracy = accuracy_score(y_true, y_pred)
    f1 = f1_score(y_true, y_pred, average='weighted')
    
    print("\n" + "="*30)
    print(f"Accuracy: {accuracy:.4f}")
    print(f"F1 Score (Weighted): {f1:.4f}")
    print("="*30 + "\n")
    
    # Detailed Report
    target_names = le.classes_
    print("Classification Report:")
    print(classification_report(y_true, y_pred, target_names=target_names))

if __name__ == "__main__":
    evaluate()