银行客户违约信息分析（数据挖掘）

#-*- coding: utf-8 -*- # 对数据进行基本的探索 import pandas as pd datafile= 'data.csv' # 读取原始数据，指定UTF-8编码（需要用文本编辑器将数据装换为UTF-8编码） data = pd.read_csv(datafile) #了解数据表的基本情况，包括行数、列数、每列的数据类型、数据完整度 #print(data.info()) #data1=data.describe() #data1.to_excel('data_dexcribe.xls') #了解数据表的统计情况；统计违约与不违约客户数量并绘制柱状图； # 提取违约和不违约的人数 import matplotlib.pyplot as plt '''weiyue = pd.value_counts(data['违约'])[1] not_weiyue = pd.value_counts(data['违约'])[0] # 绘制条形图 fig = plt.figure(figsize = (8 ,5)) # 设置画布大小 #设置字体为楷体 plt.rcParams['font.sans-serif'] = ['KaiTi'] plt.bar(x=range(2),height=[weiyue,not_weiyue,], width=0.4, alpha=0.8, color='skyblue') plt.xticks([index for index in range(2)], ['1','0']) #plt.xlabel('是否违约') #plt.ylabel('客户人数') plt.title('客户是否违约柱状图') #plt.show() #plt.close()''' import seaborn as sns # 提取会员年龄 '''age = data['年龄'].dropna() age = age.astype('int64') # 绘制客户年龄分布箱型图 #fig = plt.figure() fig,axes=plt.subplots(1,3,figsize=(10,5)) sns.distplot(data["年龄"],ax = axes[0],axlabel='所有客户年龄分布') sns.distplot(data.loc[data["违约"] == 0]["年龄"],ax = axes[1],axlabel='非违约客户年龄分布') sns.distplot(data.loc[data["违约"] == 1]["年龄"],ax = axes[2],axlabel='违约客户年龄分布') #plt.show() #plt.close()''' data['年龄'].fillna(data['年龄'].mean(), inplace=True) #删除年龄大于80的数据 # 均值填充年龄 data.drop(data[(data['年龄'] > 80)].index.tolist(),inplace = True) #print(data["年龄"].describe()) data.loc[data["性别"] == "male","性别"] = 0 data.loc[data["性别"] == "female","性别"] = 1 '''# 计算相关性矩阵 data_corr=data dt_corr = data_corr.corr(method = 'pearson') print('相关性矩阵为：\n',dt_corr) # 绘制热力图 import seaborn as sns #设置字体为楷体 plt.rcParams['font.sans-serif'] = ['KaiTi'] plt.subplots(figsize=(10, 10)) # 设置画面大小 sns.heatmap(dt_corr, annot=True, vmax=1, square=True, cmap='Blues') plt.show() plt.close ''' from sklearn.feature_extraction import DictVectorizer #通过分析选择合适的特征用于预测 features=['年龄','工龄','负债率','信用卡负债','其他负债','收入'] features_train=data[features] train_label=data['违约'] dev=DictVectorizer(sparse=False) #转化成字典型 features_train=dev.fit_transform(features_train.to_dict(orient='record')) from sklearn.tree import DecisionTreeClassifier import numpy as np from sklearn.model_selection import cross_val_score dt_stump =DecisionTreeClassifier(max_depth=1,min_samples_leaf=1) dt_stump.fit(features_train, train_label) print(u'决策树弱分类器准确率为 %.4lf' % np.mean(cross_val_score(dt_stump, features_train,train_label, cv=10))) # 决策树分类器 dt = DecisionTreeClassifier() dt.fit(features_train, train_label) print(u'决策树分类器准确率为 %.4lf' %np.mean(cross_val_score(dt, features_train, train_label, cv=10))) #adaboost from sklearn.ensemble import AdaBoostClassifier ada=AdaBoostClassifier(n_estimators=120,random_state=7,learning_rate=1.2) ada.fit(features_train,train_label) print(u'adaBoost分类器准确率为 %.4lf' %np.mean(cross_val_score(ada, features_train, train_label,cv=10))) data.to_csv('clean_data.csv') # 数据标准化 from sklearn.preprocessing import StandardScaler features_train = StandardScaler().fit_transform(features_train) #K——means聚类 from sklearn.cluster import KMeans kmeans = KMeans(n_clusters=4,n_jobs=-1,random_state=1234) # 模型训练 kmeans_fit = kmeans.fit(features_train) # 聚类中心 kmeans_cluster = kmeans_fit.cluster_centers_ print('聚类中心为\n',kmeans_fit.cluster_centers_) # 聚类后样本的类别标签 kmeans_label = kmeans_fit.labels_ print('聚类后样本标签为\n',kmeans_fit.labels_) # 聚类后各个类别数目 r1 = pd.Series(kmeans_label).value_counts() print('聚类后各个类别数目\n',r1) # 输出聚类分群结果 cluster_center = pd.DataFrame(kmeans_cluster,columns=['年龄','工龄','负债率','信用卡负债','其他负债','收入']) cluster_center.index = pd.DataFrame(kmeans_label).drop_duplicates().iloc[:,0] cluster = pd.concat([r1,cluster_center],axis=1) # 修改第一列列名 list_column = list(cluster.columns) list_column[0] = '类别数目' cluster.columns = list_column