Python-玩转数据-数据分析之电商用户行为

0、代码

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# author : sunbo time : 2022/2/13 0013
# import matplotlib.pyplot
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from datetime import  datetime

# 文件解压
import os
import zipfile
# import shutil

def unzipDir(path):
    """解压"""
    zip_file = zipfile.ZipFile(path)
    try:
        zip_file.extractall(os.path.split(path)[0]) #解压到当前目录
    except Exception as e:
        raise e
path = r'D:\data_test\tianchi_mobile_recommend_train_user.zip'
unzipDir(path)

# 读取数据
user_action = pd.read_csv(r'd:\data_test\tianchi_mobile_recommend_train_user.csv')
print(user_action.shape)  #显示表数据的行列数

#random select 5% from dataset
sample = user_action.sample(frac=0.05,random_state=5,axis=0)
# export to csv file
sample.to_csv(r'D:\data_test\user_result.csv',encoding='utf_8_sig')
user_data = pd.read_csv(r'd:\data_test\user_result.csv')
print(user_data.shape)
print(user_data.head().append(user_data.tail()))

user_data.info()

del user_data['user_geohash']
del user_data['Unnamed: 0']

print(user_data.columns)

time = pd.to_datetime(user_data['time'])
user_data['hour'] = time.dt.hour
user_data['weekday'] = time.dt.weekday
user_data['date'] = time.dt.date

# 把用户行为对应成相对应的pv collect buy
behavior = {
   1:'pv',2:'collect',3:'cart',4:'buy'}
user_data['behavior_type'] = user_data['behavior_type'].apply(lambda x: behavior[x])
print(user_data.head())

#总体访问量，收藏，加购，下单
total_liu = user_data.groupby(['behavior_type']).size()
pv = total_liu[3]
#总访客量
uv = user_data['user_id'].nunique()
#消费用户数
user_pay = user_data[user_data['behavior_type'] == 'buy']['user_id'].unique()
#日均访问量
pv_per_day = pv/user_data['date'].nunique()
#人均访问量
pv_per_user = pv/uv
#消费用户访问量
pv_pay = user_data[user_data['user_id'].isin(user_pay)]['behavior_type'].value_counts().pv
#消费用户占比
user_pay_rate = len(user_pay)/uv
#消费用户访问量占比
pv_pay_rate = pv_pay/pv
#消费用户人均访问量
pv_per_buy_user = pv_pay/len(user_pay)

print('总访问量为',total_liu)
print('总访客数为%i'%uv)
print('消费用户数为 %i'%len(user_pay))
print('消费用户访问量为 %i'%pv_pay)
print('日均访问量为 %.3f'%pv_per_day)
print('人均访问量为 %.3f'%pv_per_user)
print('消费用户人均访问量为 %.3f'%pv_per_buy_user)
print('消费用户数占比为 %.3f%%'%(user_pay_rate * 100))
print('消费用户访问量占比为 %.3f%%'%(pv_pay_rate * 100))

#每日流量
#每日pv,收藏，加购，下单
#unstack() 这是一个根据索引行列转换的函数
day_liu = user_data.groupby(['date','behavior_type']).size().unstack()
print(day_liu)
#每日uv
dayily_uv = user_data.groupby('date')['user_id'].\
    apply(lambda x:x.drop_duplicates().count()).reset_index().rename(columns={
   'user_id':'uv'})
print(dayily_uv)

#可是化展示每日流量分布
dayily_uv.plot(figsize=(12,3))
#由于购买，收藏，加入购入车无法看清，所以还要进行单独plot
for i in day_liu.columns:
    plt.figure()
    ax = day_liu[i].plot(figsize=(12,3))
    ax.set_title(i+""+"day of analysis")
# plt.show()

#每时PV,收藏，加购，下单
hour_liu = user_data.groupby(['hour','behavior_type']).size().unstack()
#每时UV
hour_uv = user_data.groupby('hour')['user_id'].apply(lambda x: x.drop_duplicates().count()).reset_index().rename(columns={
   'user_id':'uv'})

hour_uv.plot(figsize = (12,3))

#由于购买,收藏，加入购物车无法看清,所以还要进行单独plot
for i in hour_liu.columns:
    plt.figure()
    ax = hour_liu[i].plot(figsize = (12,3))
    ax.set_title(i+" "+"hour of analysis")
# plt.show()

pv_to_car_collect = (total_liu.cart + total_liu.collect)/total_liu.pv*100
car_collect_to_buy = total_liu.buy/(total_liu.cart + total_liu.collect)*100
pv_to_buy = (total_liu.buy/total_liu.pv)*100
print("浏览到收藏加入购物车的转化率为 %.3f%%"%pv_to_car_collect)
print("购物车收藏到购买的转化率为 %.3f%%"%car_collect_to_buy)
print("浏览到购买转化率 %.3f%%"%pv_to_buy)

#可视化漏斗

df_con = user_data[['user_id','item_id','time','behavior_type']]
#计算每个行为的人数
loudou = df_con['behavior_type'].value_counts().reset_index()
loudou.loc[1,:]='collect+cart',loudou.loc[1,'behavior_type']+loudou.loc[2,'behavior_type']
loudou.drop(index = 2,inplace = True)
orders = [0,1,2]
loudou.index = orders
print(loudou)

# 流量转化漏斗
# 增加列 每个阶段对应上个阶段的人数
jieduan0=list(loudou['behavior_type'])
print(jieduan0)
jieduan1=[]
for j in range(len(jieduan0)):
    if j==0:
        jieduan1.append(jieduan0[j])
    else:
        jieduan1.append(jieduan0[j-1])
loudou['行为阶段1']=jieduan1
# 增加列，每个阶段的转化率
loudou['转化率']=loudou['behavior_type']/loudou['行为阶段1']
print(loudou)

#绘制漏斗转化图
from pyecharts import options as opts
from pyecharts.charts import Funnel
from pyecharts.globals import CurrentConfig, NotebookType
#CurrentConfig.NOTEBOOK_TYPE = NotebookType.JUPYTER_LAB
CurrentConfig.NOTEBOOK_TYPE = NotebookType.JUPYTER_NOTEBOOK
CurrentConfig.ONLINE_HOST='https://assets.pyecharts.org/assets/'


#添加百分比显示
xinwei=list(loudou['index'])
zhuanhua=list(loudou['转化率'])
label=[]
for i in range(len(xinwei)):
    l=xinwei[i]+str(round(zhuanhua[i]*100,2))+'%'
    label.append(l)
c=(
    Funnel()
    .add('行为',[list(z) for z in zip(label,
                                   list(loudou['behavior_type']))],
        label_opts=opts.LabelOpts(is_show=True,position='inside'),
         tooltip_opts=opts.TooltipOpts(trigger="item", formatter="{b} : {c}%")
        )
     .set_global_opts(title_opts=opts.TitleOpts(
         title="流量转化漏斗图"
     )
                     )
)
c.render_notebook()

### 复购率
cos_rebuy = user_data[user_data.behavior_type=='buy'].groupby('user_id')['date'].apply(lambda x:len(x.unique())).rename('rebuy_count')
print('复购率为 :',round(cos_rebuy[cos_rebuy>=2].count()/cos_rebuy.count(),4))

#跳失率
attrition_rates=sum(user_data.groupby('user_id')['behavior_type'].count()==1)/(user_data['user_id'].nunique())
print('跳失率为 %.3f%%'%(attrition_rates * 100) )

#用户行为统计
user_liu = user_data.groupby(['user_id','behavior_type']).size().unstack()
user_liu.fillna(0)

user_liu=pd.crosstab(user_data["user_id"],user_data["behavior_type"])
print(user_liu)

# 单个用户消费总次数
total_buy_count=user_liu.buy.to_frame()
# 消费次数前10客户
topbuyer10=total_buy_count.sort_values(by='buy',ascending=False)[:10]
print(topbuyer10)

topbuyer10.reset_index().style.bar(color='skyblue',subset=['buy'])


tbc_box = total_buy_count.reset_index()
fig, ax = plt.subplots(figsize=[16,6])
ax.set_yscale("log")
sns.countplot(x=tbc_box['buy'],data=tbc_box,palette='Set1')
for p in ax.patches:
    ax.annotate('{:.2f}%'.format(100*p.get_height()/len(tbc_box['buy'])), (p.get_x() - 0.1, p.get_height()))
plt.title('用户消费总次数')
plt.show()

# 日活跃人数（有一次操作即视为活跃）
daily_active_user = user_data.groupby('date')['user_id'].nunique()

# 日消费人数
daily_buy_user = user_data[user_data['behavior_type'] == 'buy'].groupby('date')['user_id'].nunique()
# 日消费人数占比
proportion_of_buyer = daily_buy_user / daily_active_user
# 日消费总次数
daily_buy_count = user_data[user_data['behavior_type'] == 'buy'].groupby('date')['behavior_type'].count()
# 消费用户日人均消费次数
consumption_per_buyer = daily_buy_count / daily_buy_user
# 日消费人数占比可视化

# 柱状图数据
pob_bar = (pd.merge(daily_active_user, daily_buy_user, on='date').reset_index()
           .rename(columns={
   'user_id_x': '日活跃人数', 'user_id_y': '日消费人数'})
           .set_index('date').stack().reset_index().rename(columns={
   'level_1': 'Variable', 0: 'Value'}))
# 线图数据
pob_line = proportion_of_buyer.reset_index().rename(columns={
   'user_id': 'Rate'})

fig1 = plt.figure(figsize=[25, 6])
ax1 = fig1.add_subplot(111)
ax2 = ax1.twinx()

sns.barplot(x='date', y='Value', hue='Variable', data=pob_bar, ax=ax1, alpha=0.8, palette='husl')

ax1.legend().set_title('')
ax1.legend().remove()

sns.pointplot(pob_line['date'], pob_line['Rate'], ax=ax2, markers='D', linestyles='--', color='teal')
x = list(range(0, 16))
for a, b in zip(x, pob_line['Rate']):
    plt.text(a + 0.1, b + 0.001, '%.2f%%' % (b * 100), ha='center', va='bottom', fontsize=12)

fig1.legend(loc='upper center', ncol=2)
plt.title('日消费人数占比')

plt.show()


# 消费用户日人均消费次数可视化

# 柱状图数据
cpb_bar = (daily_buy_count.reset_index().rename(columns={
   'behavior_type': 'Num'}))
# 线图数据
cpb_line = (consumption_per_buyer.reset_index().rename(columns={
   0: 'Frequency'}))

fig2 = plt.figure(figsize=[16, 6])
ax3 = fig2.add_subplot(111)
ax4 = ax3.twinx()

sns.barplot(x='date', y='Num', data=cpb_bar, ax=ax3, alpha=0.8, palette='pastel')
sns.pointplot(cpb_line['date'], cpb_line['Frequency'], ax=ax4, markers='D', linestyles='--', color='teal')

x = list(range(0, 16))
for a, b in zip(x, cpb_line['Frequency']):
    plt.text(a + 0.1, b + 0.001, '%.2f' % b, ha='center', va='bottom', fontsize=12)
plt.title('消费用户日人均消费次数')
plt.show()


dau3_df = user_data.groupby(['date', 'user_id'])['behavior_type'].count().reset_index()
dau3_df = dau3_df[dau3_df['behavior_type'] >= 3]

# 每日高活跃用户数（每日操作数大于3次）
dau3_num = dau3_df.groupby('date')['user_id'].nunique()

fig, ax = plt.subplots(figsize=[16, 6])
sns.pointplot(dau3_num.index, dau3_num.values, markers='D', linestyles='--', color='teal')
x = list(range(0, 16))
for a, b in zip(x, dau3_num.values):
    plt.text(a + 0.1, b + 300, '%i' % b, ha='center', va='bottom', fontsize=14)
plt.title('每日高活跃用户数')
plt.show()

# 高活跃用户累计活跃天数分布
dau3_cumsum = dau3_df.groupby('user_id')['date'].count()
fig, ax = plt.subplots(figsize=[16, 6])
ax.set_yscale("log")
sns.countplot(dau3_cumsum.values, palette='Set1')
for p in ax.patches:
    ax.annotate('{:.2f}%'.format(100 * p.get_height() / len(dau3_cumsum.values)),
                (p.get_x() + 0.2, p.get_height() + 100))
plt.title('高活跃用户累计活跃天数分布')
plt.show()

# 导入相关包
from pyecharts import options as opts
from pyecharts.charts import Funnel  # 构建漏斗图
# import math

user_data['time'] = pd.to_datetime(user_data['time'], format='%Y-%m-%d %H:%M:%S')

df_con = user_data[['user_id', 'item_id', 'time', 'behavior_type']]

df_pv = df_con[df_con['behavior_type'] == 'pv']
df_collect = df_con[df_con['behavior_type'] == 'collect']
df_cart = df_con[df_con['behavior_type'] == 'cart']
df_buy = df_con[df_con['behavior_type'] == 'buy']
# 各种行为用户数
df_pv_uid = df_con[df_con['behavior_type'] == 'pv']['user_id'].unique()
df_collect_uid = df_con[df_con['behavior_type'] == 'collect']['user_id'].unique()
df_cart_uid = df_con[df_con['behavior_type'] == 'cart']['user_id'].unique()
df_buy_uid = df_con[df_con['behavior_type'] == 'buy']['user_id'].unique()

collect_cart_list = set(df_collect_uid) | set(df_cart_uid)
# 浏览用户行为与消费用户行为交集
pv_buy_df = pd.merge(left=df_pv, right=df_buy, how='inner', on=['user_id', 'item_id'], suffixes=('_pv', '_pay'))
# print(pv_pay_df)

# 浏览用户不在收藏、加入购物车用户且浏览时间小于消费时间即为经过浏览就消费的用户数
pv_buy_df = pv_buy_df[(~pv_buy_df['user_id'].isin(collect_cart_list)) & (pv_buy_df['time_pv'] < pv_buy_df['time_pay'])]

uv = user_data['user_id'].nunique()
pv_buy_num = pv_buy_df['user_id'].nunique()
pv_buy_data = pd.DataFrame({
   'type': ['浏览', '付款'], 'num': [uv, pv_buy_num]})
pv_buy_data['conversion_rates'] = (round((pv_buy_data['num'] / pv_buy_data['num']