Python機械学習(GBDT)：XGBoost

[In]
import sklearn.datasets as datasets
import pandas as pd
from sklearn.model_selection import train_test_split

cancer = datasets.load_breast_cancer() #分類データ(Classification)

def make_data(data, dftype = True):
    X, y = data.data, data.target
    columns = data.feature_names
    if dftype:
        X = pd.DataFrame(X, columns=columns)
        y = pd.DataFrame(y, columns=['target'])
    return X, y

def split_data(X: pd.DataFrame, y: pd.DataFrame, test_size: float = 0.3, random_state: int = 0):
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size, random_state=random_state)
    return X_train, X_test, y_train, y_test


X, y = make_data(cancer)
X_train, X_test, y_train, y_test = split_data(X, y)
print('Cancer data：','X_train.shape', X_train.shape, 'X_test.shape', X_test.shape, 'y_train.shape', y_train.shape, 'y_test.shape', y_test.shape)

[Out]
Cancer data： X_train.shape (398, 30) X_test.shape (171, 30) y_train.shape (398, 1) y_test.shape (171, 1)

[In]
import xgboost as xgb
from sklearn.metrics import confusion_matrix
import seaborn as sns
import matplotlib.pyplot as plt
import japanize_matplotlib


#モデルの作成・学習・評価
model = xgb.XGBClassifier(random_state=0) #XGBoostの分類モデルを作成
model.fit(X_train, y_train) #学習
y_pred = model.predict(X_test) #予測
score = model.score(X_test, y_test) #正解率
print(f'score{score:.2f}')

#混合行列(Confusion Matrix)
matrix = confusion_matrix(y_test, y_pred) 
matrix = pd.DataFrame(matrix, columns=['0:悪性', '1:良性'], index=['0:悪性', '1:良性'])
plt.figure(figsize=(6,6))
sns.heatmap(matrix, annot=True, fmt='d', cmap='Blues', cbar=False)

#重要度分析
importance = model.feature_importances_
df_imp = pd.Series(importance, index=X_train.columns).sort_values(ascending=True)
plt.figure(figsize=(12,6))
plt.barh(df_imp.index, df_imp)

[Out]
score0.96

[In]
from sklearn.metrics import r2_score

diabetes = datasets.load_diabetes() #回帰データ(Regression)
X, y = make_data(diabetes)
X_train, X_test, y_train, y_test = split_data(X, y)
print('Cancer data：','X_train.shape', X_train.shape, 'X_test.shape', X_test.shape, 'y_train.shape', y_train.shape, 'y_test.shape', y_test.shape)

#モデルの作成・学習・評価
model = xgb.XGBRegressor(objective = 'reg:squarederror', random_state=0) #XGBoostの分類モデルを作成
model.fit(X_train, y_train) #学習
y_pred = model.predict(X_test) #予測
score = model.score(X_test, y_test) #正解率
print(f'score(MSE){score:.3f}', f'r2_score{r2_score(y_test, y_pred):.3f}') #SCOREはR2値(決定係数)と同じであることを確認

#予測精度の可視化
plt.figure(figsize=(6,6))
plt.scatter(y_test, y_pred)
plt.plot([0, y_test.max()], [0, y_test.max()], c='k')
plt.xlabel('y_test(正解値)'); plt.ylabel('y_pred(予測値)')
plt.grid()

#重要度分析
importance = model.feature_importances_
df_imp = pd.Series(importance, index=X_train.columns).sort_values(ascending=True)
plt.figure(figsize=(12,6))
plt.barh(df_imp.index, df_imp)

[Out]
Cancer data： X_train.shape (309, 10) X_test.shape (133, 10) y_train.shape (309, 1) y_test.shape (133, 1)
score(MSE)0.246 r2_score0.246

[In1]
from sklearn.model_selection import GridSearchCV

params ={'max_depth':[3,5,7],
            'min_child_weight':[1.0, 2.0, 4.0],
            }

#モデルの作成・学習・評価
model = xgb.XGBRegressor(objective = 'reg:squarederror', tree_method='gpu_hist' , random_state=0) #XGBoostの分類モデルを作成

model_grid = GridSearchCV(model, params, cv=5)

model_grid.fit(X_train,
                y_train,
                early_stopping_rounds=50,
                eval_set=[(X_train, y_train)],
                verbose=0)

[In2]
model_best = model_grid.best_estimator_
y_pred = model_best.predict(X_test) #予測
score = model_best.score(X_test, y_test) #正解率
print(f'score(MSE){score:.2f}')

#予測精度の可視化
plt.figure(figsize=(6,6))
plt.scatter(y_test, y_pred)
plt.plot([0, y_test.max()], [0, y_test.max()], c='k')
plt.xlabel('y_test(正解値)'); plt.ylabel('y_pred(予測値)')
plt.grid()

#重要度分析
importance = model_best.feature_importances_
df_imp = pd.Series(importance, index=X_train.columns).sort_values(ascending=True)
plt.figure(figsize=(12,6))
plt.barh(df_imp.index, df_imp)

[Out]
score(MSE)0.14