lightgbm_binary_classification_training_with_early_stopping.py

1import lightgbm as lgb
2import pandas as pd
3from sklearn.metrics import mean_squared_error
4from sklearn.datasets import load_breast_cancer
5from sklearn.model_selection import train_test_split
6
7# Load dataset
8data = load_breast_cancer()
9X_train, X_test, y_train, y_test = train_test_split(data.data, data.target, test_size=0.2, random_state=42)
10
11# Create LightGBM datasets
12train_data = lgb.Dataset(X_train, label=y_train)
13test_data = lgb.Dataset(X_test, label=y_test, reference=train_data)
14
15# Set parameters
16params = {
17    'boosting_type': 'gbdt',
18    'objective': 'binary',
19    'metric': 'binary_logloss',
20    'num_leaves': 31,
21    'learning_rate': 0.05,
22    'feature_fraction': 0.9
23}
24
25# Train the model
26gbm = lgb.train(params,
27                train_data,
28                num_boost_round=100,
29                valid_sets=[test_data],
30                callbacks=[lgb.early_stopping(stopping_rounds=5)])
31
32# Save model
33gbm.save_model('model.txt')
34
35# Predict
36y_pred = gbm.predict(X_test, num_iteration=gbm.best_iteration)
37
38# Evaluate
39mse = mean_squared_error(y_test, y_pred)
40print(f'The mean squared error of prediction is: {mse}')