第十四章 GBDT模型

# 导入第三方包
import pandas as pd
import matplotlib.pyplot as plt

# 读入数据
default = pd.read_excel(r'C:\Users\Administrator\Desktop\default of credit card.xls')

# 数据集中是否违约的客户比例
# 为确保绘制的饼图为圆形，需执行如下代码
plt.axes(aspect = 'equal')
# 中文乱码和坐标轴负号的处理
plt.rcParams['font.sans-serif'] = ['Microsoft YaHei']
plt.rcParams['axes.unicode_minus'] = False
# 统计客户是否违约的频数
counts = default.y.value_counts()
# 绘制饼图
plt.pie(x = counts, # 绘图数据
        labels=pd.Series(counts.index).map({0:'不违约',1:'违约'}), # 添加文字标签
        autopct='%.1f%%' # 设置百分比的格式，这里保留一位小数
       )
# 显示图形
plt.show()

# 将数据集拆分为训练集和测试集
# 导入第三方包
from sklearn import model_selection
from sklearn import ensemble
from sklearn import metrics

# 排除数据集中的ID变量和因变量，剩余的数据用作自变量X
X = default.drop(['ID','y'], axis = 1)
y = default.y
# 数据拆分
X_train,X_test,y_train,y_test = model_selection.train_test_split(X,y,test_size = 0.25, random_state = 1234)

# 构建AdaBoost算法的类
AdaBoost1 = ensemble.AdaBoostClassifier()
# 算法在训练数据集上的拟合
AdaBoost1.fit(X_train,y_train)
# 算法在测试数据集上的预测
pred1 = AdaBoost1.predict(X_test)

# 返回模型的预测效果
print('模型的准确率为：\n',metrics.accuracy_score(y_test, pred1))
print('模型的评估报告：\n',metrics.classification_report(y_test, pred1))

# 计算客户违约的概率值，用于生成ROC曲线的数据
y_score = AdaBoost1.predict_proba(X_test)[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
# 计算AUC的值
roc_auc = metrics.auc(fpr,tpr)

# 绘制面积图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
# 添加边际线
plt.plot(fpr, tpr, color='black', lw = 1)
# 添加对角线
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
# 添加文本信息
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
# 添加x轴与y轴标签
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
# 显示图形
plt.show()

# 自变量的重要性排序
importance = pd.Series(AdaBoost1.feature_importances_, index = X.columns)
importance.sort_values().plot(kind = 'barh')
plt.show()

# 取出重要性比较高的自变量建模
predictors = list(importance[importance>0.02].index)
predictors

# 通过网格搜索法选择基础模型所对应的合理参数组合
# 导入第三方包
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier

max_depth = [3,4,5,6]
params1 = {'base_estimator__max_depth':max_depth}
base_model = GridSearchCV(estimator = ensemble.AdaBoostClassifier(base_estimator = DecisionTreeClassifier()),
                          param_grid= params1, scoring = 'roc_auc', cv = 5, n_jobs = 4, verbose = 1)
base_model.fit(X_train[predictors],y_train)
# 返回参数的最佳组合和对应AUC值
base_model.best_params_, base_model.best_score_

# 通过网格搜索法选择提升树的合理参数组合
# 导入第三方包
from sklearn.model_selection import GridSearchCV

n_estimators = [100,200,300]
learning_rate = [0.01,0.05,0.1,0.2]
params2 = {'n_estimators':n_estimators,'learning_rate':learning_rate}
adaboost = GridSearchCV(estimator = ensemble.AdaBoostClassifier(base_estimator = DecisionTreeClassifier(max_depth = 3)),
                        param_grid= params2, scoring = 'roc_auc', cv = 5, n_jobs = 4, verbose = 1)
adaboost.fit(X_train[predictors] ,y_train)
# 返回参数的最佳组合和对应AUC值
adaboost.best_params_, adaboost.best_score_

# 使用最佳的参数组合构建AdaBoost模型
AdaBoost2 = ensemble.AdaBoostClassifier(base_estimator = DecisionTreeClassifier(max_depth = 3),
                                       n_estimators = 300, learning_rate = 0.01)
# 算法在训练数据集上的拟合
AdaBoost2.fit(X_train[predictors],y_train)
# 算法在测试数据集上的预测
pred2 = AdaBoost2.predict(X_test[predictors])

# 返回模型的预测效果
print('模型的准确率为：\n',metrics.accuracy_score(y_test, pred2))
print('模型的评估报告：\n',metrics.classification_report(y_test, pred2))

# 计算正例的预测概率，用于生成ROC曲线的数据
y_score = AdaBoost2.predict_proba(X_test[predictors])[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
# 计算AUC的值
roc_auc = metrics.auc(fpr,tpr)

# 绘制面积图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
# 添加边际线
plt.plot(fpr, tpr, color='black', lw = 1)
# 添加对角线
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
# 添加文本信息
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
# 添加x轴与y轴标签
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
# 显示图形
plt.show()

# 运用网格搜索法选择梯度提升树的合理参数组合
learning_rate = [0.01,0.05,0.1,0.2]
n_estimators = [100,300,500]
max_depth = [3,4,5,6]
params = {'learning_rate':learning_rate,'n_estimators':n_estimators,'max_depth':max_depth}
gbdt_grid = GridSearchCV(estimator = ensemble.GradientBoostingClassifier(),
                         param_grid= params, scoring = 'roc_auc', cv = 5, n_jobs = 4, verbose = 1)
gbdt_grid.fit(X_train[predictors],y_train)
# 返回参数的最佳组合和对应AUC值
gbdt_grid.best_params_, gbdt_grid.best_score_

# 基于最佳参数组合的GBDT模型，对测试数据集进行预测
pred = gbdt_grid.predict(X_test[predictors])
# 返回模型的预测效果
print('模型的准确率为：\n',metrics.accuracy_score(y_test, pred))
print('模型的评估报告：\n',metrics.classification_report(y_test, pred))

# 计算违约客户的概率值，用于生成ROC曲线的数据
y_score = gbdt_grid.predict_proba(X_test[predictors])[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
# 计算AUC的值
roc_auc = metrics.auc(fpr,tpr)

# 绘制面积图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
# 添加边际线
plt.plot(fpr, tpr, color='black', lw = 1)
# 添加对角线
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
# 添加文本信息
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
# 添加x轴与y轴标签
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
# 显示图形
plt.show()

# 读入数据
import pandas as pd
creditcard = pd.read_csv(r'./creditcard.csv')

import matplotlib.pyplot as plt
# 为确保绘制的饼图为圆形，需执行如下代码
plt.axes(aspect = 'equal')
# 统计交易是否为欺诈的频数
counts = creditcard.Class.value_counts()
# 绘制饼图
plt.pie(x = counts, # 绘图数据
        labels=pd.Series(counts.index).map({0:'正常',1:'欺诈'}), # 添加文字标签
        autopct='%.2f%%' # 设置百分比的格式，这里保留一位小数
       )
# 显示图形
plt.show()

<Figure size 640x480 with 1 Axes>

from sklearn import model_selection
# 将数据拆分为训练集和测试集
# 删除自变量中的Time变量
X = creditcard.drop(['Time','Class'], axis = 1)
y = creditcard.Class
# 数据拆分
X_train,X_test,y_train,y_test = model_selection.train_test_split(X,y,test_size = 0.3, random_state = 1234)

# 导入第三方包
from imblearn.over_sampling import SMOTE

# 运用SMOTE算法实现训练数据集的平衡
over_samples = SMOTE(random_state=1234) 
over_samples_X,over_samples_y = over_samples.fit_sample(X_train, y_train)
#over_samples_X, over_samples_y = over_samples.fit_sample(X_train.values,y_train.values.ravel())
# 重抽样前的类别比例
print(y_train.value_counts()/len(y_train))
# 重抽样后的类别比例
print(pd.Series(over_samples_y).value_counts()/len(over_samples_y))

0    0.998239
1    0.001761
Name: Class, dtype: float64
1    0.5
0    0.5
Name: Class, dtype: float64

# 导入第三方包
import xgboost
import numpy as np
# 构建XGBoost分类器
xgboost = xgboost.XGBClassifier()
# 使用重抽样后的数据，对其建模
xgboost.fit(over_samples_X,over_samples_y)
# 将模型运用到测试数据集中
resample_pred = xgboost.predict(np.array(X_test))

# 返回模型的预测效果
print('模型的准确率为：\n',metrics.accuracy_score(y_test, resample_pred))
print('模型的评估报告：\n',metrics.classification_report(y_test, resample_pred))

C:\Users\12048\Anaconda3\lib\site-packages\dask\dataframe\utils.py:13: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.
  import pandas.util.testing as tm



---------------------------------------------------------------------------

ValueError                                Traceback (most recent call last)

<ipython-input-8-b8af401ceafd> in <module>()
      7 xgboost.fit(over_samples_X,over_samples_y)
      8 # 将模型运用到测试数据集中
----> 9 resample_pred = xgboost.predict(np.array(X_test))
     10 
     11 # 返回模型的预测效果


~\Anaconda3\lib\site-packages\xgboost\sklearn.py in predict(self, data, output_margin, ntree_limit, validate_features, base_margin)
    886             output_margin=output_margin,
    887             ntree_limit=ntree_limit,
--> 888             validate_features=validate_features)
    889         if output_margin:
    890             # If output_margin is active, simply return the scores


~\Anaconda3\lib\site-packages\xgboost\core.py in predict(self, data, output_margin, ntree_limit, pred_leaf, pred_contribs, approx_contribs, pred_interactions, validate_features, training)
   1569 
   1570         if validate_features:
-> 1571             self._validate_features(data)
   1572 
   1573         length = c_bst_ulong()


~\Anaconda3\lib\site-packages\xgboost\core.py in _validate_features(self, data)
   2129 
   2130                 raise ValueError(msg.format(self.feature_names,
-> 2131                                             data.feature_names))
   2132 
   2133     def get_split_value_histogram(self, feature, fmap='', bins=None,


ValueError: feature_names mismatch: ['V1', 'V2', 'V3', 'V4', 'V5', 'V6', 'V7', 'V8', 'V9', 'V10', 'V11', 'V12', 'V13', 'V14', 'V15', 'V16', 'V17', 'V18', 'V19', 'V20', 'V21', 'V22', 'V23', 'V24', 'V25', 'V26', 'V27', 'V28', 'Amount'] ['f0', 'f1', 'f2', 'f3', 'f4', 'f5', 'f6', 'f7', 'f8', 'f9', 'f10', 'f11', 'f12', 'f13', 'f14', 'f15', 'f16', 'f17', 'f18', 'f19', 'f20', 'f21', 'f22', 'f23', 'f24', 'f25', 'f26', 'f27', 'f28']
expected V2, V12, V18, V16, V19, V22, V15, V7, Amount, V21, V5, V3, V6, V14, V13, V8, V20, V26, V10, V25, V11, V9, V17, V27, V1, V4, V24, V23, V28 in input data
training data did not have the following fields: f28, f14, f24, f22, f19, f21, f4, f18, f1, f3, f15, f5, f10, f25, f11, f16, f2, f7, f0, f27, f6, f20, f26, f17, f12, f9, f8, f13, f23

# 计算欺诈交易的概率值，用于生成ROC曲线的数据
y_score = xgboost.predict_proba(np.array(X_test))[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
# 计算AUC的值
roc_auc = metrics.auc(fpr,tpr)

# 绘制面积图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
# 添加边际线
plt.plot(fpr, tpr, color='black', lw = 1)
# 添加对角线
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
# 添加文本信息
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
# 添加x轴与y轴标签
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
# 显示图形
plt.show()

# 构建XGBoost分类器
xgboost2 = xgboost.XGBClassifier()
# 使用非平衡的训练数据集拟合模型
xgboost2.fit(X_train,y_train)
# 基于拟合的模型对测试数据集进行预测
pred2 = xgboost2.predict(X_test)
# 混淆矩阵
pd.crosstab(pred2,y_test)

---------------------------------------------------------------------------

AttributeError                            Traceback (most recent call last)

<ipython-input-9-e26ec1b51ca2> in <module>()
      1 # 构建XGBoost分类器
----> 2 xgboost2 = xgboost.XGBClassifier()
      3 # 使用非平衡的训练数据集拟合模型
      4 xgboost2.fit(X_train,y_train)
      5 # 基于拟合的模型对测试数据集进行预测


AttributeError: 'XGBClassifier' object has no attribute 'XGBClassifier'

# 返回模型的预测效果
print('模型的准确率为：\n',metrics.accuracy_score(y_test, pred2))
print('模型的评估报告：\n',metrics.classification_report(y_test, pred2))

# 计算欺诈交易的概率值，用于生成ROC曲线的数据
y_score = xgboost2.predict_proba(X_test)[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
# 计算AUC的值
roc_auc = metrics.auc(fpr,tpr)

# 绘制面积图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
# 添加边际线
plt.plot(fpr, tpr, color='black', lw = 1)
# 添加对角线
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
# 添加文本信息
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
# 添加x轴与y轴标签
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
# 显示图形
plt.show()