scripts/train.py - master - Dinky6/Assessment-Question-Difficulty-Classifier

scripts/train.py
91 строка3 KB
Dinky6
create scripts/train.py
28 дек 2025, 17:15
28 дек 2025, 17:1530ee71c
#!/usr/bin/env python3
"""
Скрипт для обучения модели классификации сложности вопросов
"""

import sys
import os
sys.path.append(os.path.join(os.path.dirname(__file__), '..'))

from src.data_preprocessing import load_data, prepare_features, split_data
from src.feature_extraction import TextFeatureExtractor
from src.model_training import DifficultyClassifier
from src.evaluation import ModelEvaluator
import pandas as pd
import numpy as np
import argparse

def main():
    parser = argparse.ArgumentParser(description='Train question difficulty classifier')
    parser.add_argument('--data', default='data/questions.csv', help='Path to data file')
    parser.add_argument('--model', choices=['random_forest', 'svm', 'naive_bayes'], 
                       default='random_forest', help='Model type')
    parser.add_argument('--tune', action='store_true', help='Perform hyperparameter tuning')
    parser.add_argument('--output', default='models/question_classifier.pkl', 
                       help='Output model path')
    
    args = parser.parse_args()
    
    print("=" * 60)
    print("Training Question Difficulty Classifier")
    print("=" * 60)
    
    # 1. Загрузка данных
    print("\n1. Loading data...")
    df, encoder = load_data(args.data)
    print(f"   Loaded {len(df)} samples")
    print(f"   Classes: {encoder.classes_}")
    
    # 2. Подготовка признаков
    print("\n2. Preparing features...")
    # Текстовые признаки
    feature_extractor = TextFeatureExtractor()
    text_features = feature_extractor.extract_features(df['question_text'].tolist())
    
    # Простые признаки
    simple_features = prepare_features(df)
    
    # Объединение признаков
    X = np.hstack([text_features, simple_features.values])
    y = df['difficulty_encoded'].values
    
    # 3. Разделение данных
    print("\n3. Splitting data...")
    X_train, X_test, y_train, y_test, scaler = split_data(
        pd.DataFrame(X), y, test_size=0.2, random_state=42
    )
    print(f"   Train set: {X_train.shape[0]} samples")
    print(f"   Test set: {X_test.shape[0]} samples")
    
    # 4. Обучение модели
    print(f"\n4. Training {args.model} model...")
    classifier = DifficultyClassifier(model_type=args.model)
    classifier.feature_extractor = feature_extractor
    classifier.scaler = scaler
    classifier.encoder = encoder
    
    if args.tune:
        print("   Performing hyperparameter tuning...")
        classifier.hyperparameter_tuning(X_train, y_train)
    
    classifier.train(X_train, y_train)
    
    # 5. Оценка модели
    print("\n5. Evaluating model...")
    evaluator = ModelEvaluator(classifier.model, encoder)
    metrics, report, cm = evaluator.evaluate(X_test, y_test, save_path='models/metrics.json')
    
    print("\nModel Performance:")
    for metric, value in metrics.items():
        if value is not None:
            print(f"   {metric}: {value:.4f}")
    
    # 6. Сохранение модели
    print(f"\n6. Saving model to {args.output}...")
    classifier.save_model(args.output)
    
    print("\nTraining completed successfully!")
    return metrics

if __name__ == '__main__':
    main()