八大直播作业-第三讲-Aringrhus

# coding: utf-8



# In[243]:





import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

import seaborn as sns

from datetime import timedelta

get_ipython().magic('matplotlib inline')





# In[3]:





import os

os.getcwd()





# In[86]:





from pylab import mpl

mpl.rcParams["font.sans-serif"]=["SimHei"]  #设定默认字体

mpl.rcParams["axes.unicode_minus"]=False   #解决保存图像是负号“-”显示为方块的问题





# In[8]:





auto_ins = pd.read_csv(r'./auto_ins.csv',encoding='gbk')

auto_ins.info()

# auto_ins.head()





# # homework04



# ## 首先对loss重新编码为1/0，有数值为1，命名为loss_flag



# In[50]:





auto_ins['loss_flag'] = auto_ins.Loss.map(bool).map(int)

auto_ins.tail(10)





# ## 对loss_flag分布情况进行描述分析



# In[35]:





loss_desc = pd.DataFrame(auto_ins.loss_flag.value_counts())

loss_desc['row_sum'] = loss_desc.apply(sum,axis=1).apply(lambda x:round(x/4233,2))

loss_desc.loc['col_sum']=loss_desc.apply(lambda x:x.sum())

loss_desc





# In[45]:





auto_ins.loss_flag.value_counts().plot(kind='bar')

plt.figure(figsize=(4,4))

paches,texts,autotexts=plt.pie([loss_desc.loc[0,'row_sum'],loss_desc.loc[1,'row_sum']],labels=["0","1"],autopct="%1.1f%%",colors=["#FA8072","#BBFFFF"])



for text in texts+autotexts:

    text.set_fontsize(12)

for text in autotexts:

    text.set_color("black")

# ## 分析是否出险和年龄、驾龄、性别、婚姻状态等变量之间的关系  



# In[62]:





cols =list(auto_ins.columns)

cols.remove('Loss')

cols.remove('loss_flag')

cols





# ### Gender



# In[182]:





gender_desc = pd.crosstab(auto_ins['Gender'],auto_ins['loss_flag'])

gender_desc.plot(kind='bar',stacked=True)

# gender_desc

# In[180]:





def bar_line(data,x_label,y_label,title):

    gender_desc = pd.crosstab(data[x_label],data[y_label])

    rate_list = list(round(gender_desc[1]/sum(data[y_label]),2))



    x = list(gender_desc.index)

    y_bottom = list(gender_desc[0])

    y_top = list(gender_desc[1])

    fig = plt.figure()



    ax1 = fig.add_subplot(111)

    ax1.grid(True)

    # ax1.grid(True,color='b',linewidth='0.3',linestyle='--')  #语句不对

    ax1.bar(range(len(y_bottom)), y_bottom, label='not default',fc = '#FA8072')  

    ax1.bar(range(len(y_top)), y_top, bottom=y_bottom, label='default loss',tick_label = x,fc = '#BBFFFF')  

    ax1.set_ylabel(x_label)

    ax1.set_title(title)

    ax1.legend() 

    # ax1.set_xlim()

    # ax1.grid(False)





    ax2 = ax1.twinx()

    ax2.plot(x,rate_list,'')

    ax2.set_ylabel(y_label)

    ax2.set_xlabel(x)

    ax2.grid(False)#, color = "r") 

    for x, y in zip(x, rate_list):

        plt.text(x, y+0.04, str(y), ha='center', va='bottom', fontsize=10.5)





# In[183]:





bar_line(auto_ins,'Gender','loss_flag','性别与逾期关系')

# ### Marital



# In[181]:





bar_line(auto_ins,'Marital','loss_flag','婚姻状态与逾期关系')

# ### Owner



# In[185]:





bar_line(auto_ins,'Owner','loss_flag','归属主体与逾期关系')

# ### Age



# In[225]:





sns.boxplot(x = 'loss_flag', y = 'Age', data = auto_ins)

# ### vAge



# In[226]:





sns.boxplot(x = 'loss_flag', y = 'vAge', data = auto_ins)

# HWK5

# # homework05



# In[186]:





os.getcwd()





# In[189]:





clients = pd.read_csv(r"../HW5/clients.csv",encoding='gbk')

print(clients.head())

clients.info()





# In[190]:





card = pd.read_csv(r"../HW5/card.csv",encoding='gbk')

print(card.head())

card.info()





# In[191]:





accounts = pd.read_csv(r"../HW5/accounts.csv",encoding='gbk')

print(accounts.head())

accounts.info()





# In[192]:





disp = pd.read_csv(r"../HW5/disp.csv",encoding='gbk')

print(disp.head())

disp.info()





# ## 不同类型卡的持卡人的性别对比



# In[203]:





card_disp = pd.merge(card,disp[['disp_id','client_id']],how='left',on='disp_id')

card_disp.head(10)

client_card = pd.merge(card_disp,clients,how='left',on='client_id')

client_card.head(10)





# In[209]:





card_sex = pd.crosstab(client_card['type'],client_card['sex'])

# card_sex['sum1'] = card_sex.sum(1)

card_sex = card_sex.div(card_sex.sum(1),axis = 0)

card_sex .plot(kind='bar',stacked=True)

# ## 不同类型卡的持卡人在办卡时的平均年龄对比



# In[213]:





import datetime

def convert_date(date_str):

    date = datetime.datetime.strptime(date_str,'%Y-%m-%d')

    return date

client_card['issued_date'] = client_card['issued'].map(convert_date)

client_card['birth_date_t'] = client_card['birth_date'].map(convert_date)





# In[222]:





client_card['issued_age_dt'] = client_card['issued_date']-client_card['birth_date_t']

client_card['issued_age'] = client_card['issued_age_dt'].map(lambda x:round(x.days/365))





# In[223]:





client_card.head()





# In[224]:





client_card['issued_age'].groupby(client_card['type']).mean().plot(kind="bar")

# ## 不同类型卡的持卡人在办卡前一年内的平均帐户余额对比



# In[236]:





trans = pd.read_csv(r"../HW5/trans.csv",encoding='gbk')

# trans.head()

trans.drop(['trans_id','k_symbol','bank','account'],axis=1,inplace = True)

trans.columns=['account_id', 'trans_date', 'type', 'operation', 'amount', 'balance']

trans.head()





# In[246]:





# trans['trans_date_dt'] = trans['trans_date'].map(convert_date)

trans['trans_pre_year'] = trans['trans_date_dt'].apply(lambda x:x- timedelta(days=365))

trans.head()





# In[249]:





# client_card_account = pd.merge(client_card,disp[['account_id','client_id']],how='left',on='client_id')

client_card_account['issued_pre_year'] = client_card_account['issued_date'].apply(lambda x:x- timedelta(days=365))

client_card_account.head()





# In[250]:





client_card_account_trans = pd.merge(client_card_account,trans[['account_id','type','amount','balance','trans_date_dt']],how='left',on='account_id')

client_card_account_trans.head()





# In[260]:





all_trans_pre_year = client_card_account_trans[client_card_account_trans['trans_date_dt']<=client_card_account_trans['issued_pre_year']]

all_trans_pre_year['balance'] = all_trans_pre_year['balance'].apply(lambda x:int(x.replace('$','').replace(',','')))

all_trans_pre_year['amount'] = all_trans_pre_year['amount'].apply(lambda x:int(x.replace('$','').replace(',','')))

all_trans_pre_year.head()





# In[289]:





client_balance_mean = all_trans_pre_year[['account_id','balance']].groupby(all_trans_pre_year['account_id']).mean()

del client_balance_mean['account_id']

client_balance_mean=client_balance_mean.reset_index()

client_balance_mean = pd.merge(client_balance_mean,all_trans_pre_year[['account_id','type_x']],how='left',on='account_id')

client_balance_mean.drop_duplicates(inplace=True)

client_balance_mean.head()





# In[290]:





sns.boxplot(x='type_x',y='balance',data=client_balance_mean)

# ## 不同类型卡的持卡人在办卡前一年内的平均收入对比



# In[292]:





all_trans_pre_year_income = all_trans_pre_year[all_trans_pre_year['type_y']=='贷']

all_trans_pre_year_income.head()





# In[295]:





# all_trans_pre_year_income_mean = all_trans_pre_year_income[['account_id','amount']].groupby(all_trans_pre_year_income['account_id']).mean()

# del all_trans_pre_year_income_mean['account_id']

# all_trans_pre_year_income_mean=all_trans_pre_year_income_mean.reset_index()

# all_trans_pre_year_income_mean.head()



all_trans_pre_year_income_mean = pd.merge(all_trans_pre_year_income_mean,all_trans_pre_year[['account_id','type_x']],how='left',on='account_id')

all_trans_pre_year_income_mean.drop_duplicates(inplace=True)

all_trans_pre_year_income_mean.head()





# In[296]:





sns.boxplot(x='type_x',y='amount',data=all_trans_pre_year_income_mean)

• HW45.py