背景描述:
以数据bankloan.xls,前8列作为x,最后一列为y,建立银行风控模型。采用三种算法模型分别得到训练的结果,训练的误差以及混淆矩阵。
一、BP神经网络
混淆矩阵可视化函数cm_plot:
def cm_plot(y, yp): from sklearn.metrics import confusion_matrix cm = confusion_matrix(y, yp) import matplotlib.pyplot as plt plt.matshow(cm, cmap=plt.cm.Greens) plt.colorbar() for x in range(len(cm)): for y in range(len(cm)): plt.annotate(cm[x,y], xy=(x, y), horizontalalignment='center', verticalalignment='center') plt.ylabel('True label') plt.xlabel('Predicted label') return plt
分类预测训练结果:
import pandas as pd inputfile = r'C:\Users\86158\Desktop\python数据分析\data\sales_data.xls' data = pd.read_excel(inputfile, index_col = '序号') data[data == '好'] = 1 data[data == '是'] = 1 data[data == '高'] = 1 data[data != 1] = 0 x = data.iloc[:,:3].astype(int) y = data.iloc[:,3].astype(int) from keras.models import Sequential from keras.layers.core import Dense, Activation model = Sequential() # 建立模型 model.add(Dense(input_dim = 3, units = 10)) model.add(Activation('relu')) # 用relu函数作为激活函数,能够大幅提供准确度 model.add(Dense(input_dim = 10, units = 1)) model.add(Activation('sigmoid')) # 由于是0-1输出,用sigmoid函数作为激活函数 model.compile(loss = 'binary_crossentropy', optimizer = 'adam') model.fit(x, y, epochs = 1000, batch_size = 10) yp = model.predict_classes(x).reshape(len(y)) from cm_plot import * cm_plot(y,yp).show()
混淆矩阵如下:
二、支持向量机(SVM)和决策树
import pandas as pd import time import numpy as np import seaborn as sns import matplotlib.pyplot as plt from sklearn.model_selection import train_test_split from sklearn.tree import DecisionTreeClassifier as DTC from sklearn import svm from sklearn import tree from sklearn.metrics import confusion_matrix from sklearn.metrics import accuracy_score from sklearn.metrics import roc_curve, auc from sklearn.neighbors import KNeighborsClassifier as KNN from sklearn.metrics import plot_roc_curve,roc_curve,auc,roc_auc_score filePath = r'C:\Users\86158\Desktop\python数据分析\data\bankloan.xls' data = pd.read_excel(filePath) x = data.iloc[:,:8] y = data.iloc[:,8] x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=100) svm_clf = svm.SVC()#支持向量机 dtc_clf = DTC(criterion='entropy')#决策树 #训练 dtc_clf.fit(x_train,y_train) svm_clf.fit(x_train, y_train) #ROC曲线比较 fig,ax = plt.subplots(figsize=(12,10)) svm_roc = plot_roc_curve(estimator=svm_clf, X=x, y=y, ax=ax, linewidth=1) dtc_roc = plot_roc_curve(estimator=dtc_clf, X=x, y=y, ax=ax, linewidth=1) ax.legend(fontsize=12) plt.show() #模型评价 svm_yp = svm_clf.predict(x) svm_score = accuracy_score(y, svm_yp) dtc_yp = dtc_clf.predict(x) dtc_score = accuracy_score(y, dtc_yp) score = {"支持向量机得分":svm_score,"决策树得分":dtc_score} score = sorted(score.items(),key = lambda score:score[0],reverse=True) print(pd.DataFrame(score)) plt.rcParams['font.sans-serif'] = ['SimHei'] plt.rcParams['axes.unicode_minus'] = False #绘制混淆矩阵 figure = plt.subplots(figsize=(12,10)) plt.subplot(1,2,1) plt.title('支持向量机') svm_cm = confusion_matrix(y, svm_yp) heatmap = sns.heatmap(svm_cm, annot=True, fmt='d') heatmap.yaxis.set_ticklabels(heatmap.yaxis.get_ticklabels(), rotation=0, ha='right') heatmap.xaxis.set_ticklabels(heatmap.xaxis.get_ticklabels(), rotation=45, ha='right') plt.ylabel("true label") plt.xlabel("predict label") plt.subplot(1,2,2) plt.title('决策树') dtc_cm = confusion_matrix(y, dtc_yp) heatmap = sns.heatmap(dtc_cm, annot=True, fmt='d') heatmap.yaxis.set_ticklabels(heatmap.yaxis.get_ticklabels(), rotation=0, ha='right') heatmap.xaxis.set_ticklabels(heatmap.xaxis.get_ticklabels(), rotation=45, ha='right') plt.ylabel("true label") plt.xlabel("predict label") plt.show() #画出决策树 import pandas as pd import os os.environ["PATH"] += os.pathsep + 'D:/软件下载安装/Graphviz/bin' from sklearn.tree import export_graphviz x = pd.DataFrame(x) with open( r'C:\Users\86158\Desktop\python数据分析\data\banklodan_tree.dot', 'w') as f: export_graphviz(dtc_clf, feature_names = x.columns, out_file = f) f.close() from IPython.display import Image from sklearn import tree import pydotplus dot_data = tree.export_graphviz(dtc_clf, out_file=None, #regr_1 是对应分类器 feature_names=x.columns, #对应特征的名字 class_names= ['不违约','违约'], #对应类别的名字 filled=True, rounded=True, special_characters=True) graph = pydotplus.graph_from_dot_data(dot_data.replace('helvetica',"MicrosoftYaHei")) graph.write_png(r'C:\Users\86158\Desktop\python数据分析\banklodan_tree.png') #保存图像 Image(graph.create_png())
得到ROC曲线:
混淆矩阵:
显然,三种训练模型的对比下,决策树和支持向量机的效果比较好,决策树的效果最好。