LightGBM
2値分類　

---

2値分類
（Jupyter Lab などで、順にコードを実行することで動作）

初期準備　 データ準備　 モデル作成〜訓練　 評価　

---

・初期準備

import lightgbm as lgb
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.datasets import load_breast_cancer
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split

plt.style.use('dark_background')Copy

2値分類に戻る

---

・データ準備

data = load_breast_cancer()
print(data.keys())
print(data.data.shape, data.target.shape)

X = data.data
y = data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y)
X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, stratify=y_train)Copy

2値分類に戻る

---

・モデル作成〜訓練

Parameters（外部リンク）：https://lightgbm.readthedocs.io/en/latest/Parameters.html

params = {
    # 目的関数(binary, multiclass, regression, cross_entropy, etc)
    'objective': 'binary',
    # 各特徴量の最大分割数(default=255, 大->精度UP & 過学習)
    'max_bin': 255,
    # 各決定木の分岐最大数(default=31, 大->精度UP & 過学習)
    'num_leaves': 31,
    # 学習率(default=0.1, 小->緩やかに精度UP & 時間長)
    'learning_rate': 0.1,
    # 木の深さ(-1:制限なし)
    'max_depth': -1,
    # 葉の最少データ数(default=20)
    "min_data_in_leaf": 20,
    "early_stopping_rounds": 50,
    "verbosity": -1,
    "random_state": 0,
}

train_data = lgb.Dataset(X_train, label=y_train) 
val_data = lgb.Dataset(X_val, label=y_val) 

model = lgb.train(
    params,
    train_data, 
    num_boost_round=200, # default=100
    valid_names=['train','valid'],
    valid_sets=[train_data,val_data],
    
)

pred = model.predict(X_val, num_iteration=model.best_iteration)
pred = np.where(pred >= 0.5, 1, 0)
print('\naccuracy：', accuracy_score(y_val, pred).round(3), '\n')

# パラメータ確認
# print('params:\n', model.params, '\n')

print('Done!')Copy

2値分類に戻る

---

・評価

# テストデータ
pred = model.predict(X_test, num_iteration=model.best_iteration)
pred = np.where(pred >= 0.5, 1, 0)
print('accuracy：', accuracy_score(y_test, pred).round(3), '\n')

# 特徴量重要度
importances = pd.DataFrame({'feature_name': data.feature_names, 
                            'feature_importance': model.feature_importance()})
importances.sort_values('feature_importance', ascending=False, inplace=True)
print(importances, '\n')

# 特徴量重要度上位をグラフ化
fig, ax = plt.subplots(figsize=(12,5))
ax.bar(importances.feature_name, importances.feature_importance)
fig.autofmt_xdate(rotation=45)
ax.set_title('Feature Importance');Copy

2値分類に戻る