目录

一、数据探索：

1.1 读取数据

1.2查看数据

1.3 数据预处理

二、字段描述

2.1 非离散型数据

2.2离散数值字段

三、数据建模

四、评估指标：

4.1：混淆矩阵

4.2： 准确率，回归率，F1

五、测试集准确率

六、模型优化

环境：使用python+jupter nodebook

数据：本文数据来源2023年【教学赛】金融数据分析赛题1：银行客户认购产品预测

赛题（数据）网址：【教学赛】金融数据分析赛题1：银行客户认购产品预测-天池大赛-阿里云天池

一、数据探索：

   1.1 读取数据

所需要的库包：

import pandas as pd

import numpy as np

trian=pd.read_csv("train.csv")

test=pd.read_csv("test.csv")

   1.2查看数据

        是否正常，有无异常值：

        查看统计量

print(df.describe().T)

        查看数据分布（散点图）：

# 1 查看统计量

print(df.describe().T)

# 2 duration分箱展示

import matplotlib.pyplot as plt

import seaborn as sns

# 3.查看数据分布

# 分离数值变量与分类变量

Nu_feature = list(df.select_dtypes(exclude=['object']).columns)

Ca_feature = list(df.select_dtypes(include=['object']).columns)

Ca_feature.remove('subscribe')

col1=Ca_feature

plt.figure(figsize=(20,10))

j=1

for col in col1:

ax=plt.subplot(4,5,j)

ax=plt.scatter(x=range(len(df)),y=df[col],color='red')

plt.title(col)

j+=1

k=11

for col in col1:

ax=plt.subplot(4,5,k)

ax=plt.scatter(x=range(len(test)),y=test[col],color='cyan')

plt.title(col)

k+=1

plt.subplots_adjust(wspace=0.4,hspace=0.3)

plt.show()

数据相关图（热力图）

# # 4.数据相关图

from sklearn.preprocessing import LabelEncoder

lb = LabelEncoder()

cols = Ca_feature

for m in cols:

df[m] = lb.fit_transform(df[m])

test[m] = lb.fit_transform(test[m])

#

df['subscribe'] = df['subscribe'].replace(['no', 'yes'], [0, 1])

correlation_matrix = df.corr()

plt.figure(figsize=(12, 10))

sns.heatmap(correlation_matrix, vmax=0.9, linewidths=0.05, cmap="RdGy")

plt.show()

查看数据是否有空值或者unkonw

#数据没有NA值但是有unknow值

train_set.isin(['unknown']).mean()*100

test_set.isin(['unknown']).mean()*100

# 工作，教育和沟通方式用众数填充

1.3 数据预处理

对训练集和测试集数据进行填充：

trian['default'].replace(['unknown'], test['default'].mode(), inplace=True)

trian['job'].replace(['unknown'], trian['job'].mode(), inplace=True)

trian['education'].replace(['unknown'], trian['education'].mode(), inplace=True)

trian['marital'].replace(['unknown'], trian['marital'].mode(), inplace=True)

trian['housing'].replace(['unknown'], trian['housing'].mode(), inplace=True)

trian['loan'].replace(['unknown'], trian['loan'].mode(), inplace=True)

# test.drop(['default'], inplace=True, axis=1)

test['default'].replace(['unknown'], test['default'].mode(), inplace=True)

test['job'].replace(['unknown'], test['job'].mode(), inplace=True)

test['education'].replace(['unknown'], test['education'].mode(), inplace=True)

test['marital'].replace(['unknown'], test['marital'].mode(), inplace=True)

test['housing'].replace(['unknown'], test['housing'].mode(), inplace=True)

test['loan'].replace(['unknown'], test['loan'].mode(), inplace=True)

print(trian["job"].value_counts())

二、字段描述

        2.1 非离散型数据

# #统计图

plt.rcParams['font.sans-serif']=['SimHei'] #用来正常显示中文标签

trian['subscribe'] = trian['subscribe'].replace(['no', 'yes'], [0,1])

plt.figure(figsize = [15,10])#画板大小

sns.barplot(x = "job", y ="subscribe" , data = trian)

x_1=["管理者","蓝领","技术员","服务员","经营者","退役人员","企业家","个体经营者","女佣","失业人员","学生"]

from matplotlib import font_manager

my_font = font_manager.FontProperties(fname='C:\Windows\Fonts\STHUPO.TTF',size=20)

# plt.xticks(range(len(x_1)),x_1,fontproperties = my_font)

plt.xticks(range(len(x_1)),x_1,fontsize=20,rotation=45)

plt.yticks(fontsize=15)

my= font_manager.FontProperties(size=20)

plt.xlabel("“客户身份”",fontproperties = my)

plt.ylabel("产品购买数量指数",fontproperties = my)

plt.title("客户购买银行产品意向图",fontdict={"size": 25})

plt.tight_layout()

plt.show()

import seaborn as sns

object_columns = ['job', 'marital', 'education', 'default', 'housing','loan', 'contact','month','day_of_week','poutcome']

#连续变量列名

num_columns = ['age', 'duration', 'campaign', 'pdays','previous', "cons_conf_index",'emp_var_rate',"cons_price_index","lending_rate3m","nr_employed"]

# # 统计图

# #统计图

plt.rcParams['font.sans-serif']=['SimHei'] #用来正常显示中文标签

plt.figure(figsize = [10,10])#画板大小

sns.barplot(x = "marital", y ="subscribe" , data = trian)

x_1=["结婚"," 已婚","单身"]

from matplotlib import font_manager

my_font = font_manager.FontProperties(fname='C:\Windows\Fonts\STHUPO.TTF',size=20)

# plt.xticks(range(len(x_1)),x_1,fontproperties = my_font)

plt.xticks(range(len(x_1)),x_1,fontsize=20)

plt.yticks(fontsize=15)

my= font_manager.FontProperties(size=20)

plt.xlabel("客户婚姻状态",fontproperties = my)

plt.ylabel("产品购买数量指数",fontproperties = my)

plt.title("不同婚姻状态的客户购买银行产品意向图",fontdict={"size": 25})

plt.tight_layout()

plt.show()

# #统计图

plt.rcParams['font.sans-serif']=['SimHei'] #用来正常显示中文标签

plt.figure(figsize = [10,8])#画板大小

sns.barplot(x = "education", y ="subscribe" , data = trian)

x_1=["大学学历"," 高中","基本9年","教授","基本4年","基本6年","文盲"]

from matplotlib import font_manager

my_font = font_manager.FontProperties(fname='C:\Windows\Fonts\STHUPO.TTF',size=20)

# plt.xticks(range(len(x_1)),x_1,fontproperties = my_font)

plt.xticks(range(len(x_1)),x_1,fontsize=25)

plt.yticks(fontsize=15)

my= font_manager.FontProperties(size=20)

plt.xlabel("客户教育程度",fontproperties = my)

plt.ylabel("产品购买数量指数",fontproperties = my)

plt.title("不同教育程度的客户购买银行产品意向图",fontdict={"size": 25})

plt.tight_layout()

plt.show()

print(trian["education"].value_counts())

object_columns = ['job', 'marital', 'education', 'default', 'housing','loan', 'contact','month','day_of_week','poutcome']

# #统计图

plt.rcParams['font.sans-serif']=['SimHei'] #用来正常显示中文标签

plt.figure(figsize = [8,8])#画板大小

sns.barplot(x = "month", y ="subscribe" , data = trian)

# x_1=["大学学历"," 高中","基本9年","教授","基本4年","基本6年","文盲"]

from matplotlib import font_manager

# my_font = font_manager.FontProperties(fname='C:\Windows\Fonts\STHUPO.TTF',size=20)

# plt.xticks(range(len(x_1)),x_1,fontproperties = my_font)

plt.xticks(fontsize=25)

plt.yticks(fontsize=15)

my= font_manager.FontProperties(size=20)

plt.xlabel("月份",fontproperties = my)

plt.ylabel("产品购买数量指数",fontproperties = my)

plt.title("不同月份最后联系客户购买银行产品意向图",fontdict={"size": 25})

plt.tight_layout()

plt.show()

 其余字段大同小异

下面结合结婚状态字段对产品进行分析：

# print(trian["marital"].value_counts())

marital_colum=["married" ,"single" ,"divorced"]

# # 选取某列含有特定“marital”的行

trian1 = trian[trian['marital'].isin([marital_colum[0]])]

trian1.dropna(axis=0, how='any', thresh=None, subset=None, inplace=False)

print(trian1["marital"].value_counts())

plt.figure(figsize=[10, 10])

sns.barplot(x="default", y="subscribe", hue="education", data=trian1, palette="muted")

x_1=["yes"," no"]

from matplotlib import font_manager

plt.xticks(range(len(x_1)),x_1,fontsize=25)

plt.yticks(fontsize=15)

my= font_manager.FontProperties(size=20)

plt.xlabel("有无违约记录",fontproperties = my)

plt.ylabel("产品购买数量指数",fontproperties = my)

plt.title("已婚",fontdict={"size": 25})

plt.legend(prop = {'size':18})

plt.tight_layout()

plt.show()

# print(trian["marital"].value_counts())

marital_colum=["married" ,"single" ,"divorced"]

# # 选取某列含有特定“marital”的行

trian1 = trian[trian['marital'].isin([marital_colum[1]])]

trian1.dropna(axis=0, how='any', thresh=None, subset=None, inplace=False)

print(trian1["marital"].value_counts())

plt.figure(figsize=[10, 10])

sns.barplot(x="default", y="subscribe", hue="education", data=trian1, palette="muted")

x_1=["no"," yes"]

from matplotlib import font_manager

plt.xticks(range(len(x_1)),x_1,fontsize=25)

plt.yticks(fontsize=15)

my= font_manager.FontProperties(size=20)

plt.xlabel("有无违约记录",fontproperties = my)

plt.ylabel("产品购买数量指数",fontproperties = my)

plt.title("单身",fontdict={"size": 25})

plt.legend(prop = {'size':18})

plt.tight_layout()

plt.show()

2.2离散数值字段

plt.rcParams['font.sans-serif'] = ['Microsoft YaHei']

z=0

while(z<=9):

trian1 = trian.loc[:,[num_columns[z],'subscribe']]

# ax = plt.subplot(3, 3, z + 1)

f = pd.melt(trian1, value_vars=num_columns[z], id_vars='subscribe')

g = sns.FacetGrid(f,col='variable', hue='subscribe')

z = z + 1

g = g.map(sns.distplot,"value",bins=20)

plt.show()

三、数据建模

from lightgbm.sklearn import LGBMClassifier

from sklearn.model_selection import train_test_split

from sklearn.model_selection import KFold

from sklearn.metrics import accuracy_score, auc, roc_auc_score

X = df.drop(columns=['id', 'subscribe'])

Y = df['subscribe']

testA = test.drop(columns='id')

# 划分训练及测试集

x_train, x_test, y_train, y_test = train_test_split(X, Y, test_size=0.3, random_state=1)

from xgboost import XGBClassifier

import xgboost as xgb

from sklearn.metrics import precision_score, recall_score, f1_score

model = xgb.XGBClassifier()

# 交叉验证

result1 = []

mean_score1 = 0

n_folds = 10

import time

start =time.time()

kf = KFold(n_splits=n_folds, shuffle=True, random_state=2022)

for train_index, test_index in kf.split(X):

x_train = X.iloc[train_index]

y_train = Y.iloc[train_index]

x_test = X.iloc[test_index]

y_test = Y.iloc[test_index]

model.fit(x_train, y_train)

y_pred1 = model.predict_proba((x_test))[:, 1]

print('验证集AUC:{}'.format(roc_auc_score(y_test, y_pred1)))

mean_score1 += roc_auc_score(y_test, y_pred1) / n_folds

y_pred_final1 = model.predict_proba((testA))[:, 1]

y_pred_test1 = y_pred_final1

result1.append(y_pred_test1)

end =time.time()

print('程序运行时间为: %s Seconds'%(end-start))

使用验证集AUC模型评估模型：

 ROC曲线：

AUC值： 

四、评估指标：

4.1：混淆矩阵

4.2： 准确率，回归率，F1

五、测试集准确率

输出文件为：

cat_pre1 = sum(result1) / n_folds

ret1 = pd.DataFrame(cat_pre1, columns=['subscribe'])

ret1['subscribe'] = np.where(ret1['subscribe'] > 0.5, 'yes', 'no').astype('str')

ret1.to_csv('./XGB预测.csv', index=False)

最终提交结果为：

六、模型优化

 本文章未调参，如果进行网格优化调参可以让模型进一步变好