fix the activation 'relu', it performs better

credit_detection.py

@@ -0,0 +1,262 @@
+#coding:utf-8
+# 信用卡数据异常检测
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from scipy import stats
+import seaborn as sns
+
+from sklearn.model_selection import train_test_split
+
+LABELS=['Normal','Fraud']
+
+# 加载数据
+df=pd.read_csv('data/creditcard.csv')
+print(df.shape)
+
+print(df.isnull().values.any())  # False 没有缺失值
+
+# 欺诈数据与正常数据的比例
+count_classes=pd.value_counts(df['Class'],sort=True)
+
+count_classes.plot(kind='bar',rot=0)
+plt.title('Transaction class distribution')
+plt.xticks(range(2),LABELS)
+plt.xlabel('Class')
+plt.ylabel('Frequency')
+plt.show()
+
+frauds=df[df.Class==1]
+normal=df[df.Class==0]
+
+# (492,31) (284315,31)
+print(frauds.shape,normal.shape)
+
+# 不同交易中使用的货币数量
+print(frauds.Amount.describe())
+print(normal.Amount.describe())
+
+# 两行一列 2子图  共享x轴
+f,(ax1,ax2)=plt.subplots(2,1,sharex=True)
+f.suptitle('Amount per transaction by class')
+bins=50
+
+# 柱形图
+ax1.hist(frauds.Amount,bins=bins)
+ax1.set_title('Fraud')
+
+ax2.hist(normal.Amount,bins=bins)
+ax2.set_title('Normal')
+
+plt.xlabel('Amount($)')
+plt.ylabel('Number of Transactions')
+plt.xlim(0,20000)
+# log(y)
+plt.yscale('log')
+plt.show()
+
+# 不同时间中使用的数量 // 时间特征并不重要
+f,(ax1,ax2)=plt.subplots(2,1,sharex=True)
+f.suptitle('Time of transaction vs Amount by class')
+
+ax1.scatter(frauds.Time,frauds.Amount)
+ax1.set_title('Fraud')
+
+ax2.scatter(normal.Time,normal.Amount)
+ax2.set_title('Normal')
+
+plt.xlabel('Time (s)')
+plt.ylabel('Amount')
+
+plt.show()
+
+# 建立自编码器
+from sklearn.preprocessing import StandardScaler
+data=df.drop(['Time'],axis=1)
+
+print(data['Amount'].values.shape)
+
+# (x-mean)/std
+data['Amount']=StandardScaler().fit_transform(data['Amount'].values.reshape(-1,1))
+
+print(data['Amount'].shape)
+# print(data['Amount'])
+
+# 划分0.2为测试集
+X_train,X_test=train_test_split(data,test_size=0.2,random_state=42)
+X_train=X_train[X_train.Class==0]
+X_train=X_train.drop(['Class'],axis=1)
+
+y_test=X_test['Class']
+X_test=X_test.drop(['Class'],axis=1)
+X_train=X_train.values
+X_test=X_test.values
+print(X_train.shape,X_test.shape,y_test.shape)
+
+from keras.models import Model,load_model
+from keras.layers import Input,Dense
+from keras.callbacks import ModelCheckpoint,TensorBoard
+from keras import regularizers
+
+# create model
+input_dim=X_train.shape[1]
+encoding_dim=14
+
+input_layer=Input(shape=(input_dim,))
+encoder=Dense(encoding_dim,activation='tanh',\
+		activity_regularizer=regularizers.l1(10e-5))(input_layer)
+encoder=Dense(int(encoding_dim/2),activation='relu')(encoder)
+decoder=Dense(encoding_dim,activation='tanh')(encoder)
+decoder=Dense(input_dim,activation='linear')(decoder)
+
+auto_encoder=Model(inputs=input_layer,outputs=decoder)
+
+print(auto_encoder.summary())
+
+nb_epoch=150
+batch_size=256
+
+auto_encoder.compile(optimizer='adam',loss='mse',metrics=['accuracy'])
+
+checkpointer=ModelCheckpoint(filepath='model_me.h5',verbose=1,save_best_only=True)
+
+tensorboard=TensorBoard(log_dir='./logs',histogram_freq=0,\
+		write_graph=True,write_images=True)
+
+
+# 训练
+hist=auto_encoder.fit(X_train,X_train,epochs=nb_epoch,\
+	batch_size=batch_size,shuffle=True,validation_data=(X_test,X_test),\
+	verbose=1,callbacks=[checkpointer,tensorboard]).history
+
+plt.plot(hist['loss'])
+plt.plot(hist['val_loss'])
+plt.title('model loss')
+plt.ylabel('loss')
+plt.xlabel('epoch')
+plt.legend(['train','test'],loc='upper right')
+plt.show()
+
+
+# 导入model
+auto_encoder=load_model('model_me.h5')
+
+predictions=auto_encoder.predict(X_test)
+mse=np.mean(np.power(X_test-predictions,2),axis=1)
+
+error_df=pd.DataFrame({'reconstruction_error':mse,'true_class':y_test})
+print(error_df.describe())
+
+# 无欺诈数据的重构误差
+fig=plt.figure()
+ax=fig.add_subplot(111)
+normal_error_df=error_df[(error_df['true_class']==0) & (error_df['reconstruction_error']<10)]
+_=ax.hist(normal_error_df.reconstruction_error.values,bins=10)
+plt.show()
+
+# 有欺诈数据的重构误差
+fig=plt.figure()
+ax=fig.add_subplot(111)
+fraud_error_df=error_df[error_df['true_class']==1]
+_=ax.hist(fraud_error_df.reconstruction_error.values,bins=10)
+plt.show()
+
+from sklearn.metrics import (confusion_matrix,precision_recall_curve,auc,\
+	roc_curve,recall_score,classification_report,f1_score,precision_recall_fscore_support)
+
+# ROC曲线
+fpr,tpr,thresholds=roc_curve(error_df.true_class,error_df.reconstruction_error)
+print(fpr.shape,tpr.shape)
+# print(thresholds)
+
+roc_auc=auc(fpr,tpr)  # 面积
+print('roc_auc:',roc_auc)
+
+plt.title('Receiver Operating Characteristic')
+plt.plot(fpr,tpr,label='AUC=%.4F' %(roc_auc))
+plt.legend(loc='lower right')
+
+plt.plot([0,1],[0,1],'r--')
+plt.xlim([0,1])
+plt.ylim([0,1])
+
+plt.ylabel('True Positive Rate')
+plt.xlabel('False Positive Rate')
+plt.show()
+
+# precision VS recall
+precision,recall,th=precision_recall_curve(error_df.true_class,error_df.reconstruction_error)
+plt.plot(recall,precision,'b',label='Precision-Recall curve')
+plt.title('Recall VS Precision')
+plt.xlabel('Recall')
+plt.ylabel('Precision')
+
+plt.show()
+
+print(precision.shape,recall.shape,th.shape)
+
+# precision by threshold
+plt.plot(th,precision[1:],'b',label='Threshold-Precision curve')
+plt.title('Precision for different threshold values')
+plt.xlabel('threshold')
+plt.ylabel('Precision')
+plt.show()
+
+# recall by threshold
+plt.plot(th,recall[1:],'r',label='Threshold-Recall curve')
+plt.title('Recall for different threshold values')
+plt.xlabel('threshold')
+plt.ylabel('Recall')
+plt.show()
+
+'''
+我们需要从交易数据本身计算重构误差。
+如果误差大于预先定义的阈值，我们就把它标记为欺诈
+（因为我们的模型在正常交易上应该有一个低的重构误差）
+'''
+# TODO
+threshold=2.5
+
+groups=error_df.groupby('true_class')
+fig,ax=plt.subplots()
+
+for name,group in groups:
+	ax.plot(group.index,group.reconstruction_error,marker='o',\
+		ms=3.5,linestyle='',label='Fraud' if name==1 else 'Normal')
+
+ax.hlines(threshold,ax.get_xlim()[0],ax.get_xlim()[1],color='r',\
+		zorder=100,label='Threshold')
+ax.legend()
+
+plt.title('Reconstruction error for different classes')
+plt.ylabel('Reconstruction error')
+plt.xlabel('Data point index')
+
+plt.show()
+
+# 预测
+y_pred=[1 if e>threshold else 0 for e in error_df.reconstruction_error.values]
+conf_matrix=confusion_matrix(error_df.true_class,y_pred)
+
+print(conf_matrix)
+p_normal=conf_matrix[0][0]*1.0/(conf_matrix[0][0]+conf_matrix[1][0])
+r_normal=conf_matrix[0][0]*1.0/(conf_matrix[0][0]+conf_matrix[0][1])
+
+p_fraud=conf_matrix[1][1]*1.0/(conf_matrix[0][1]+conf_matrix[1][1])
+r_fraud=conf_matrix[1][1]*1.0/(conf_matrix[1][0]+conf_matrix[1][1])
+
+print('p_normal:%f  r_normal:%f' %(p_normal,r_normal))
+print('p_fraud:%f  r_fraud:%f' %(p_fraud,r_fraud))
+
+
+plt.figure(figsize=(12,12))
+sns.heatmap(conf_matrix,xticklabels=LABELS,yticklabels=LABELS,\
+							annot=True,fmt='d')
+plt.title('Confusion matrix')
+plt.ylabel('True class')
+plt.xlabel('Predicted class')
+plt.show()
+
+
+
+