顾客购物分析

import numpy as np 
import pandas as pd 
import datetime
from matplotlib import pyplot as plt 
import seaborn as sns
plt.style.use(['fivethirtyeight'])

from sklearn.preprocessing import LabelEncoder
from sklearn.preprocessing import StandardScaler
from sklearn import metrics
from sklearn.decomposition import PCA

from sklearn.cluster import KMeans
from sklearn.cluster import AgglomerativeClustering
from yellowbrick.cluster import KElbowVisualizer

import warnings
warnings.filterwarnings('ignore')

data = pd.read_csv('/Users/robert/python/marketing_campaign.csv', sep= '\t')
data.info() 

data.head().T

## 检查缺失值
print("缺失值:\n", data.isnull().sum())    ## 收入数据有缺失值，缺失了24条数据

## 检查是否有重复数据
data.duplicated().sum()

## 采用均值对收入列的24条缺失数据进行填充
data = data.fillna(data['Income'].mean())

## 将注册日期数据格式调整为日期格式
data['Dt_Customer'] = pd.to_datetime(data['Dt_Customer'])
data['Dt_Customer'].apply(['min', 'max'])

print("婚姻状况:\n", data['Marital_Status'].value_counts(), '\n')
print("教育情况:\n", data['Education'].value_counts())

## 不同婚姻状态的顾客所占比例
plt.figure(figsize=(9, 9))
x = data['Marital_Status'].value_counts().sort_values()
plt.pie(x=x, labels=['YoLo', 'Absurd','Alone', 'Window', 'Divorced', 'Single', 'Together','Married'],
        autopct = '%1.1f%%', )
plt.axis('equal')
plt.title('Propotion of Marital_Status')

## 不同受教育水平的顾客所占比例
x = data['Education'].value_counts().sort_values()

plt.figure(figsize=(9,9))
plt.pie(x=x, labels=['2n Cycle', 'Basic', 'Master', 'PhD', 'Graduation'], autopct = '%1.1f%%')
plt.title('Propotion of Education')

data['Age'] = data.apply(lambda data:  2015- data['Year_Birth'], axis=1)
data['Spent'] = data.apply(lambda data: data['MntFishProducts'] + data['MntFruits'] + data['MntGoldProds'] + 
                data['MntMeatProducts'] + data['MntSweetProducts'] + data['MntWines'], axis=1)


data['Children'] = data.apply(lambda data: data['Kidhome'] + data['Teenhome'], axis=1)

data['TotalAcceptedCmp'] = data.apply(lambda data: data['AcceptedCmp1'] + data['AcceptedCmp2'] + 
                                      data['AcceptedCmp3'] + data['AcceptedCmp4'] + 
                                      data['AcceptedCmp5'], axis=1)
data['NumTotalPurchases'] = data.apply(lambda data: data['NumWebPurchases'] + data['NumCatalogPurchases'] + 
                            data['NumStorePurchases'] + data['NumDealsPurchases'], axis=1)

## 删除'Year_Birth', 'Z_CostContact', 'Z_Revenue', 'Kidhome', 'Teenhome'列
data = data.drop(['Year_Birth', 'Z_CostContact', 'Z_Revenue', 'Kidhome', 'Teenhome'], axis=1) 

## 删除数据中收入和年龄的异常值部分
data = data.drop(data.loc[data['Income'] > 600000].index)
data = data.drop(data.loc[data['Age'] > 100].index)

## 创建新的特征”Have_Children“，有小孩值为1，没有小孩值为0
data['Have_Children'] = np.where(data.Children > 0, 1, 0)
data.head()

## 以是否有小孩为分组对数据进行分组绘制散点图矩阵
pairplot = data.loc[:, ['Income', 'Age', 'Spent', 'Recency', 'Have_Children']]

sns.pairplot(pairplot, hue='Have_Children', palette='Set1')

## 对连续型变量Age进行面元划分
bins = [0, 20, 40, 60, 80,100]
aged = pd.cut(data['Age'], bins=bins)
print('number of age_cut:\n', aged.value_counts())

## 不同年龄段顾客分类汇总
plt.figure(figsize=(12,6))
sns.countplot(y=aged, palette='Set2')
plt.title('countplot of Age')

## 不同年龄区间在购物花费上的区别
plt.figure(figsize=(12,6))
sns.boxplot(x=aged, y=data['Spent'], palette='Set2')

## 受教育水平和购物花费间的关系
## 不同婚姻状态下顾客的花费比较
to_boxplot = ['Education', 'Marital_Status']
fig, axes = plt.subplots(2, 1, figsize=(12, 12))
axes = axes.flatten()

for col, ax in zip(to_boxplot, axes):
    ax = sns.boxplot(data=data, x=col, y='Spent', ax=ax, palette='Set2')
    ax.set_title(f'boxplot of spent by {col}')

## 以教育水平的不同对不同婚姻状态的顾客进行分组，并对比其在收入和购物花费上的差异对比
fig, axes = plt.subplots(2, 1, figsize=(14,10))
sns.barplot(x= 'Marital_Status',y='Income',hue='Education',data=data, ci=0,palette='Set2', ax=axes[0])
sns.barplot(x= 'Marital_Status',y='Spent',hue='Education',data=data, ci=0, palette='Set2', ax=axes[1])

## 收入和支出呈线性增长关系
fig, axes = plt.subplots(1, 2, figsize=(14,6))
sns.scatterplot(y=data['Spent'], x=data['Income'], ax=axes[0])
sns.regplot(y='Spent', x='Income', data=data, ax=axes[1])

## 网站顾客整体花费直方图
plt.figure(figsize=(12,6))
sns.histplot(data['Spent'], bins=50, kde=True)
plt.title('histogram of Spent')

## 自用户上次购买以来的天数的频数分布
plt.figure(figsize=(12,6))
sns.histplot(data['Recency'], bins=60)

## 网站用户月度活跃频数统计
plt.figure(figsize=(12,6))
sns.histplot(data['NumWebVisitsMonth'], bins=50, kde=True)
plt.title('Monthly activities (5-7)/month')

## 网站投诉情况
x = data['Complain'].value_counts().sort_values()

plt.figure(figsize=(8,8))
plt.pie(x=x, labels=['YES', 'NO'], autopct = '%1.1f%%')
plt.title('Complain of customer')

## 酒、鱼、水果、黄金、肉、糖果各自销量频数统计
to_histplot = ['MntWines', 'MntFishProducts', 'MntFruits', 'MntGoldProds', 
               'MntMeatProducts', 'MntSweetProducts']

fig, axes = plt.subplots(3, 2, sharex=True, sharey=True, figsize=(14, 10))
axes = axes.flatten()

for col, ax in zip(to_histplot, axes):
    ax = sns.histplot(data=data, x=col, ax=ax)
    ax.set_title(f'histogram of {col}')

## 通过目录、商店和网站三个渠道购买的频数统计
to_histplot = ['NumCatalogPurchases', 'NumStorePurchases', 'NumWebPurchases']
fig, axes = plt.subplots(1, 3, sharey=True, figsize=(14, 9))
axes = axes.flatten()

for col, ax in zip(to_histplot, axes):
    ax = sns.histplot(data=data, x=col, ax=ax)
    plt.subplots_adjust(hspace = 0.5, wspace = 0.3)

## 创建总活动接受报价次数指标
acceptedConcat = data[['AcceptedCmp1', 'AcceptedCmp2', 'AcceptedCmp3', 'AcceptedCmp4', 'AcceptedCmp5']]
acceptedConcat = acceptedConcat.apply(pd.DataFrame.sum)

print('acceptedConcat:\n', acceptedConcat)

## 每次接受活动报价的用户的频数统计
plt.figure(figsize=(12,6))
plt.title('accepted the campaings in every attempt')
sns.barplot(x=acceptedConcat.index, y=acceptedConcat, palette='Set2')

## 用户上次购买以来的天数和在上次活动是否接受报价间的关系
plt.figure(figsize=(12,6))
plt.title('Recency Vs Acceptance of an offer')
sns.lineplot(x='Recency', y='Response', data=data)

## 网站每月活跃用户折线图
groupedDate = data.set_index('Dt_Customer')
groupedDate = groupedDate.resample('M').count()

plt.figure(figsize=(14,4))
plt.title('Customer NumWebVisitsMonth by mounths')
groupedDate.NumWebVisitsMonth.plot(kind='line')

## 每月总体消费金额折线图
groupedDate = data.set_index('Dt_Customer')
groupedDate = groupedDate.resample('M').count()

plt.figure(figsize=(14,4))
plt.title('Customer Spent by mounths')
groupedDate.Spent.plot(kind='line')

## 每月通过网站购买的用户折线图
groupedDate = data.set_index('Dt_Customer')
groupedDate = groupedDate.resample('M').count()

plt.figure(figsize=(14,4))
plt.title('Customer NumWebPurchases by mounths')
groupedDate.NumWebPurchases.plot()

## 每月通过目录购买的用户折线图
groupedDate = data.set_index('Dt_Customer')
groupedDate = groupedDate.resample('M').count()

plt.figure(figsize=(14,4))
plt.title('Customer NumCatalogPurchases by mounths')
groupedDate.NumCatalogPurchases.plot(kind='line')

##每月通过商店购买的用户折现图
groupedDate = data.set_index('Dt_Customer')

groupedDate = groupedDate.resample('M').count()

plt.figure(figsize=(14,4))
plt.title('Customer NumStorePurchases by mounths')
groupedDate.NumStorePurchases.plot(kind='line')

s = (data.dtypes == 'object')
object_cols = list(s[s].index)

print("Categorical variables in the dataset:", object_cols)

LE=LabelEncoder()
for i in object_cols:
    data[i]=data[[i]].apply(LE.fit_transform)

## 相关性热力图
plt.figure( figsize = (14, 14))
data1 = data.iloc[:, 1:-1]
sns.heatmap(data1.corr(), annot=True, cmap='Set2')

ds = data.copy()

cols_del = ['ID', 'AcceptedCmp3', 'AcceptedCmp4', 'AcceptedCmp5', 'AcceptedCmp1','AcceptedCmp2', 
            'Complain', 'Response', 'Have_Children', 'Dt_Customer',  'Children']
ds = ds.drop(cols_del, axis=1)

## 数据标准化
scaler = StandardScaler()
scaler.fit(ds)
scaled_ds = pd.DataFrame(scaler.transform(ds),columns= ds.columns )

pca = PCA(n_components=3)
pca.fit(scaled_ds)
PCA_ds = pd.DataFrame(pca.transform(scaled_ds), columns=(["col1","col2", "col3"]))
PCA_ds.describe().T

x =PCA_ds["col1"]
y =PCA_ds["col2"]
z =PCA_ds["col3"]

## 降维后数据可视化
fig = plt.figure(figsize=(10,8))
ax = fig.add_subplot(111, projection="3d")
ax.scatter(x,y,z, marker="o" )
ax.set_title("Data Visualization In The Reduced Dimension")
plt.show()

Elbow_M = KElbowVisualizer(KMeans(), k=10)
Elbow_M.fit(PCA_ds)
Elbow_M.show()

AC = AgglomerativeClustering(n_clusters=4)

## 拟合模型并预测聚类
yhat_AC = AC.fit_predict(PCA_ds)
PCA_ds["Clusters"] = yhat_AC

## 将聚类特征添加到原始数据
data["Clusters"]= yhat_AC

## 聚类结果可视化
fig = plt.figure(figsize=(10,8))
ax = plt.subplot(111, projection='3d', label="bla")
ax.scatter(x, y, z, s=50, c=PCA_ds["Clusters"], marker='o', cmap = 'Set2' )
ax.set_title("The Plot Of The Clusters")
plt.show()

## 聚类结果分布图
plt.figure(figsize=(12, 6))
pl = sns.countplot(x=data["Clusters"], palette= 'Set2')
pl.set_title("Distribution Of The Clusters")
plt.show()

## 按照收入和支出对聚类结果进行可视化
plt.figure(figsize=(12, 6))
pl = sns.scatterplot(data = data,x=data["Spent"], y=data["Income"],hue=data["Clusters"], palette= 'Set2')
pl.set_title("Cluster's Profile Based On Income And Spent")
plt.legend()
plt.show()

plt.figure(figsize=(12,6))
sns.swarmplot(x=data["Clusters"], y=data["Spent"], alpha=0.9, palette= 'Set2' )

## 按照聚类结果对活动总接受情况进行可视化
plt.figure(figsize=(12,6))
pl = sns.countplot(x=data["TotalAcceptedCmp"],hue=data["Clusters"], palette= 'Set2')
pl.set_title("Count Of Promotion Accepted")
pl.set_xlabel("Number Of Total Accepted Promotions")
plt.show()