数据预处理样本选择、交叉验证

# 下采样取样本数据
X &＃61; data.ix[:, data.columns !&＃61; &＃39;Class&＃39;]
y &＃61; data.ix[:, data.columns &＃61;&＃61; &＃39;Class&＃39;]# Number of data points in the minority class
number_records_fraud &＃61; len(data[data.Class &＃61;&＃61; 1])
fraud_indices &＃61; np.array(data[data.Class &＃61;&＃61; 1].index)# Picking the indices of the normal classes
normal_indices &＃61; data[data.Class &＃61;&＃61; 0].index# Out of the indices we picked, randomly select "x" number (number_records_fraud)
random_normal_indices &＃61; np.random.choice(normal_indices, number_records_fraud, replace &＃61; False)
random_normal_indices &＃61; np.array(random_normal_indices)# Appending the 2 indices
under_sample_indices &＃61; np.concatenate([fraud_indices,random_normal_indices])# Under sample dataset
under_sample_data &＃61; data.iloc[under_sample_indices,:]X_undersample &＃61; under_sample_data.ix[:, under_sample_data.columns !&＃61; &＃39;Class&＃39;]
y_undersample &＃61; under_sample_data.ix[:, under_sample_data.columns &＃61;&＃61; &＃39;Class&＃39;]# Showing ratio
print("Percentage of normal transactions: ", len(under_sample_data[under_sample_data.Class &＃61;&＃61; 0])/len(under_sample_data))
print("Percentage of fraud transactions: ", len(under_sample_data[under_sample_data.Class &＃61;&＃61; 1])/len(under_sample_data))
print("Total number of transactions in resampled data: ", len(under_sample_data))# 下采样后的数据进行训练、验证数据集拆分
from sklearn.cross_validation import train_test_split# Whole dataset
X_train, X_test, y_train, y_test &＃61; train_test_split(X,y,test_size &＃61; 0.3, random_state &＃61; 0)print("Number transactions train dataset: ", len(X_train))
print("Number transactions test dataset: ", len(X_test))
print("Total number of transactions: ", len(X_train)&＃43;len(X_test))# Undersampled dataset
X_train_undersample, X_test_undersample, y_train_undersample, y_test_undersample &＃61; train_test_split(X_undersample,y_undersample,test_size &＃61; 0.3,random_state &＃61; 0)
print("")
print("Number transactions train dataset: ", len(X_train_undersample))
print("Number transactions test dataset: ", len(X_test_undersample))
print("Total number of transactions: ", len(X_train_undersample)&＃43;len(X_test_undersample))

#Recall &＃61; TP/(TP&＃43;FN)
from sklearn.linear_model import LogisticRegression
from sklearn.cross_validation import KFold, cross_val_score
from sklearn.metrics import confusion_matrix,recall_score,classification_report
def printing_Kfold_scores(x_train_data,y_train_data):fold &＃61; KFold(len(y_train_data),5,shuffle&＃61;False) # Different C parametersc_param_range &＃61; [0.01,0.1,1,10,100]results_table &＃61; pd.DataFrame(index &＃61; range(len(c_param_range),2), columns &＃61; [&＃39;C_parameter&＃39;,&＃39;Mean recall score&＃39;])results_table[&＃39;C_parameter&＃39;] &＃61; c_param_range# the k-fold will give 2 lists: train_indices &＃61; indices[0], test_indices &＃61; indices[1]j &＃61; 0for c_param in c_param_range:print(&＃39;-------------------------------------------&＃39;)print(&＃39;C parameter: &＃39;, c_param)print(&＃39;-------------------------------------------&＃39;)print(&＃39;&＃39;)recall_accs &＃61; []for iteration, indices in enumerate(fold,start&＃61;1):# Call the logistic regression model with a certain C parameterlr &＃61; LogisticRegression(C &＃61; c_param, penalty &＃61; &＃39;l1&＃39;)# Use the training data to fit the model. In this case, we use the portion of the fold to train the model# with indices[0]. We then predict on the portion assigned as the &＃39;test cross validation&＃39; with indices[1]
lr.fit(x_train_data.iloc[indices[0],:],y_train_data.iloc[indices[0],:].values.ravel())# Predict values using the test indices in the training datay_pred_undersample &＃61; lr.predict(x_train_data.iloc[indices[1],:].values)# Calculate the recall score and append it to a list for recall scores representing the current c_parameterrecall_acc &＃61; recall_score(y_train_data.iloc[indices[1],:].values,y_pred_undersample)recall_accs.append(recall_acc)print(&＃39;Iteration &＃39;, iteration,&＃39;: recall score &＃61; &＃39;, recall_acc)# The mean value of those recall scores is the metric we want to save and get hold of.results_table.ix[j,&＃39;Mean recall score&＃39;] &＃61; np.mean(recall_accs)j &＃43;&＃61; 1print(&＃39;&＃39;)print(&＃39;Mean recall score &＃39;, np.mean(recall_accs))print(&＃39;&＃39;)best_c &＃61; results_table.loc[results_table[&＃39;Mean recall score&＃39;].idxmax()][&＃39;C_parameter&＃39;]# Finally, we can check which C parameter is the best amongst the chosen.print(&＃39;*********************************************************************************&＃39;)print(&＃39;Best model to choose from cross validation is with C parameter &＃61; &＃39;, best_c)print(&＃39;*********************************************************************************&＃39;)return best_cbest_c &＃61; printing_Kfold_scores(X_train_undersample,y_train_undersample)

def plot_confusion_matrix(cm, classes,title&＃61;&＃39;Confusion matrix&＃39;,cmap&＃61;plt.cm.Blues):"""This function prints and plots the confusion matrix."""plt.imshow(cm, interpolation&＃61;&＃39;nearest&＃39;, cmap&＃61;cmap)plt.title(title)plt.colorbar()tick_marks &＃61; np.arange(len(classes))plt.xticks(tick_marks, classes, rotation&＃61;0)plt.yticks(tick_marks, classes)thresh &＃61; cm.max() / 2.for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):plt.text(j, i, cm[i, j],horizontalalignment&＃61;"center",color&＃61;"white" if cm[i, j] > thresh else "black")plt.tight_layout()plt.ylabel(&＃39;True label&＃39;)plt.xlabel(&＃39;Predicted label&＃39;)

import itertools
lr &＃61; LogisticRegression(C &＃61; best_c, penalty &＃61; &＃39;l1&＃39;)
lr.fit(X_train_undersample,y_train_undersample.values.ravel())
y_pred_undersample &＃61; lr.predict(X_test_undersample.values)# Compute confusion matrix
cnf_matrix &＃61; confusion_matrix(y_test_undersample,y_pred_undersample)
np.set_printoptions(precision&＃61;2)print("Recall metric in the testing dataset: ", cnf_matrix[1,1]/(cnf_matrix[1,0]&＃43;cnf_matrix[1,1]))# Plot non-normalized confusion matrix
class_names &＃61; [0,1]
plt.figure()
plot_confusion_matrix(cnf_matrix, classes&＃61;class_names, title&＃61;&＃39;Confusion matrix&＃39;)
plt.show()

lr &＃61; LogisticRegression(C &＃61; 0.01, penalty &＃61; &＃39;l1&＃39;)
lr.fit(X_train_undersample,y_train_undersample.values.ravel())
y_pred_undersample_proba &＃61; lr.predict_proba(X_test_undersample.values)thresholds &＃61; [0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9]plt.figure(figsize&＃61;(10,10))j &＃61; 1
for i in thresholds:y_test_predictions_high_recall &＃61; y_pred_undersample_proba[:,1] > iplt.subplot(3,3,j)j &＃43;&＃61; 1# Compute confusion matrixcnf_matrix &＃61; confusion_matrix(y_test_undersample,y_test_predictions_high_recall)np.set_printoptions(precision&＃61;2)print("Recall metric in the testing dataset: ", cnf_matrix[1,1]/(cnf_matrix[1,0]&＃43;cnf_matrix[1,1]))# Plot non-normalized confusion matrixclass_names &＃61; [0,1]plot_confusion_matrix(cnf_matrix, classes&＃61;class_names, title&＃61;&＃39;Threshold >&＃61; %s&＃39;%i)