python商品零售购物篮分析

 1 # -*- coding: utf-8 -*-
 2 
 3 # 代码8-1 查看数据特征
 4 
 5 import numpy as np
 6 import pandas as pd
 7 
 8 inputfile = r'C:\Users\86184\Desktop\文件集\data\GoodsOrder.csv'   # 输入的数据文件
 9 data = pd.read_csv(inputfile,encoding = 'gbk')  # 读取数据
10 data .info()  # 查看数据属性
11 
12 data = data['id']
13 description = [data.count(),data.min(), data.max()]  # 依次计算总数、最小值、最大值
14 description = pd.DataFrame(description, index = ['Count','Min', 'Max']).T  # 将结果存入数据框
15 print('描述性统计结果：\n',np.round(description))  # 输出结果

 1 # 代码8-2 分析热销商品
 2 
 3 # 销量排行前10商品的销量及其占比
 4 import pandas as pd
 5 inputfile = r'C:\Users\86184\Desktop\文件集\data\GoodsOrder.csv'  # 输入的数据文件
 6 data = pd.read_csv(inputfile,encoding = 'gbk')  # 读取数据
 7 group = data.groupby(['Goods']).count().reset_index()  # 对商品进行分类汇总
 8 sorted=group.sort_values('id',ascending=False)
 9 print('销量排行前10商品的销量:\n', sorted[:10])  # 排序并查看前10位热销商品
10 
11 # 画条形图展示出销量排行前10商品的销量
12 import matplotlib.pyplot as plt
13 x=sorted[:10]['Goods']
14 y=sorted[:10]['id']
15 plt.figure(figsize = (10, 6))  # 设置画布大小 
16 plt.barh(x,y)
17 plt.rcParams['font.sans-serif'] = 'SimHei'
18 plt.xlabel('销量')  # 设置x轴标题
19 plt.ylabel('商品类别')  # 设置y轴标题
20 plt.title('商品的销量TOP10  num =3013',fontsize = 20)  # 设置标题
21 plt.savefig(r'C:\Users\86184\Desktop\文件集\data\top10.png')  # 把图片以.png格式保存
22 plt.show()  # 展示图片
23 
24 # 销量排行前10商品的销量占比
25 data_nums = data.shape[0]
26 for idnex, row in sorted[:10].iterrows():
27     print(row['Goods'],row['id'],row['id']/data_nums)

 1 # 代码8-3 各类别商品的销量及其占比
 2 
 3 import pandas as pd
 4 inputfile1 = r'C:\Users\86184\Desktop\文件集\data\GoodsOrder.csv'
 5 inputfile2 = r'C:\Users\86184\Desktop\文件集\data\GoodsTypes.csv'
 6 data = pd.read_csv(inputfile1,encoding = 'gbk')
 7 types = pd.read_csv(inputfile2,encoding = 'gbk')  # 读入数据
 8 
 9 group = data.groupby(['Goods']).count().reset_index()
10 sort = group.sort_values('id',ascending = False).reset_index()
11 data_nums = data.shape[0]  # 总量
12 del sort['index']
13 
14 sort_links = pd.merge(sort,types)  # 合并两个datafreame 根据type
15 # 根据类别求和，每个商品类别的总量，并排序
16 sort_link = sort_links.groupby(['Types']).sum().reset_index()
17 sort_link = sort_link.sort_values('id',ascending = False).reset_index()
18 del sort_link['index']  # 删除“index”列
19 
20 # 求百分比，然后更换列名，最后输出到文件
21 sort_link['count'] = sort_link.apply(lambda line: line['id']/data_nums,axis=1)
22 sort_link.rename(columns = {'count':'percent'},inplace = True)
23 print('各类别商品的销量及其占比:\n',sort_link)
24 outfile1 = r'C:\Users\86184\Desktop\文件集\data\percent.csv'
25 sort_link.to_csv(outfile1,index = False,header = True,encoding='gbk')  # 保存结果
26 
27 # 画饼图展示每类商品销量占比
28 import matplotlib.pyplot as plt
29 data = sort_link['percent']
30 labels = sort_link['Types']
31 plt.figure(figsize=(10, 8))  # 设置画布大小   
32 plt.pie(data,labels=labels,autopct='%1.2f%%')
33 plt.rcParams['font.sans-serif'] = 'SimHei'
34 plt.title('每类商品销量占比  num=3013',fontsize = 20)  # 设置标题
35 plt.savefig(r'C:\Users\86184\Desktop\文件集\data\persent.png')  # 把图片以.png格式保存
36 plt.show()

 1 # 代码8-4 非酒精饮料内部商品的销量及其占比
 2 
 3 # 先筛选“非酒精饮料”类型的商品，然后求百分比，然后输出结果到文件。
 4 selected = sort_links.loc[sort_links['Types'] == '非酒精饮料']  # 挑选商品类别为“非酒精饮料”并排序
 5 child_nums = selected['id'].sum()  # 对所有的“非酒精饮料”求和
 6 selected['child_percent'] = selected.apply(lambda line: line['id']/child_nums,axis = 1)  # 求百分比
 7 selected.rename(columns = {'id':'count'},inplace = True)
 8 print('非酒精饮料内部商品的销量及其占比:\n',selected)
 9 outfile2 = r'C:\Users\86184\Desktop\文件集\data\child_percent.csv'
10 sort_link.to_csv(outfile2,index = False,header = True,encoding='gbk')  # 输出结果
11 
12 # 画饼图展示非酒精饮品内部各商品的销量占比
13 import matplotlib.pyplot as plt
14 data = selected['child_percent']
15 labels = selected['Goods']
16 plt.figure(figsize = (10,8))  # 设置画布大小 
17 explode = (0.02,0.03,0.04,0.05,0.06,0.07,0.08,0.08,0.3,0.1,0.3)  # 设置每一块分割出的间隙大小
18 plt.pie(data,explode = explode,labels = labels,autopct = '%1.2f%%',
19         pctdistance = 1.1,labeldistance = 1.2)
20 plt.rcParams['font.sans-serif'] = 'SimHei'
21 plt.title("非酒精饮料内部各商品的销量占比  num=3013",fontsize=20)  # 设置标题
22 plt.axis('equal')
23 plt.savefig(r'C:\Users\86184\Desktop\文件集\data\child_persent.png')  # 保存图形
24 plt.show()  # 展示图形

 1 # -*- coding: utf-8 -*-
 2 
 3 # 代码8-5 数据转换
 4 
 5 import pandas as pd
 6 inputfile=r'C:\Users\86184\Desktop\文件集\data\GoodsOrder.csv'
 7 data = pd.read_csv(inputfile,encoding = 'gbk')
 8 
 9 # 根据id对“Goods”列合并，并使用“，”将各商品隔开
10 data['Goods'] = data['Goods'].apply(lambda x:','+x)
11 data = data.groupby('id').sum().reset_index()
12 
13 # 对合并的商品列转换数据格式
14 data['Goods'] = data['Goods'].apply(lambda x :[x[1:]])
15 data_list = list(data['Goods'])
16 
17 # 分割商品名为每个元素
18 data_translation = []
19 for i in data_list:
20     p = i[0].split(',')
21     data_translation.append(p)
22 print('数据转换结果的前5个元素：\n', data_translation[0:5])

  1 # 代码8-6 构建关联规则模型
  2 
  3 from numpy import *
  4  
  5 def loadDataSet():
  6     return [['a', 'c', 'e'], ['b', 'd'], ['b', 'c'], ['a', 'b', 'c', 'd'], ['a', 'b'], ['b', 'c'], ['a', 'b'],
  7             ['a', 'b', 'c', 'e'], ['a', 'b', 'c'], ['a', 'c', 'e']]
  8  
  9 def createC1(dataSet):
 10     C1 = []
 11     for transaction in dataSet:
 12         for item in transaction:
 13             if not [item] in C1:
 14                 C1.append([item])
 15     C1.sort()
 16     # 映射为frozenset唯一性的，可使用其构造字典
 17     return list(map(frozenset, C1))     
 18     
 19 # 从候选K项集到频繁K项集（支持度计算）
 20 def scanD(D, Ck, minSupport):
 21     ssCnt = {}
 22     for tid in D:   # 遍历数据集
 23         for can in Ck:  # 遍历候选项
 24             if can.issubset(tid):  # 判断候选项中是否含数据集的各项
 25                 if not can in ssCnt:
 26                     ssCnt[can] = 1  # 不含设为1
 27                 else:
 28                     ssCnt[can] += 1  # 有则计数加1
 29     numItems = float(len(D))  # 数据集大小
 30     retList = []  # L1初始化
 31     supportData = {}  # 记录候选项中各个数据的支持度
 32     for key in ssCnt:
 33         support = ssCnt[key] / numItems  # 计算支持度
 34         if support >= minSupport:
 35             retList.insert(0, key)  # 满足条件加入L1中
 36             supportData[key] = support  
 37     return retList, supportData
 38  
 39 def calSupport(D, Ck, min_support):
 40     dict_sup = {}
 41     for i in D:
 42         for j in Ck:
 43             if j.issubset(i):
 44                 if not j in dict_sup:
 45                     dict_sup[j] = 1
 46                 else:
 47                     dict_sup[j] += 1
 48     sumCount = float(len(D))
 49     supportData = {}
 50     relist = []
 51     for i in dict_sup:
 52         temp_sup = dict_sup[i] / sumCount
 53         if temp_sup >= min_support:
 54             relist.append(i)
 55 # 此处可设置返回全部的支持度数据（或者频繁项集的支持度数据）
 56             supportData[i] = temp_sup
 57     return relist, supportData
 58  
 59 # 改进剪枝算法
 60 def aprioriGen(Lk, k):
 61     retList = []
 62     lenLk = len(Lk)
 63     for i in range(lenLk):
 64         for j in range(i + 1, lenLk):  # 两两组合遍历
 65             L1 = list(Lk[i])[:k - 2]
 66             L2 = list(Lk[j])[:k - 2]
 67             L1.sort()
 68             L2.sort()
 69             if L1 == L2:  # 前k-1项相等，则可相乘，这样可防止重复项出现
 70                 # 进行剪枝（a1为k项集中的一个元素，b为它的所有k-1项子集）
 71                 a = Lk[i] | Lk[j]  # a为frozenset()集合
 72                 a1 = list(a)
 73                 b = []
 74                 # 遍历取出每一个元素，转换为set，依次从a1中剔除该元素，并加入到b中
 75                 for q in range(len(a1)):
 76                     t = [a1[q]]
 77                     tt = frozenset(set(a1) - set(t))
 78                     b.append(tt)
 79                 t = 0
 80                 for w in b:
 81                     # 当b（即所有k-1项子集）都是Lk（频繁的）的子集，则保留，否则删除。
 82                     if w in Lk:
 83                         t += 1
 84                 if t == len(b):
 85                     retList.append(b[0] | b[1])
 86     return retList
 87 
 88 def apriori(dataSet, minSupport=0.2):
 89 # 前3条语句是对计算查找单个元素中的频繁项集
 90     C1 = createC1(dataSet)
 91     D = list(map(set, dataSet))  # 使用list()转换为列表
 92     L1, supportData = calSupport(D, C1, minSupport)
 93     L = [L1]  # 加列表框，使得1项集为一个单独元素
 94     k = 2
 95     while (len(L[k - 2]) > 0):  # 是否还有候选集
 96         Ck = aprioriGen(L[k - 2], k)
 97         Lk, supK = scanD(D, Ck, minSupport)  # scan DB to get Lk
 98         supportData.update(supK)  # 把supk的键值对添加到supportData里
 99         L.append(Lk)  # L最后一个值为空集
100         k += 1
101     del L[-1]  # 删除最后一个空集
102     return L, supportData  # L为频繁项集，为一个列表，1，2，3项集分别为一个元素
103 
104 # 生成集合的所有子集
105 def getSubset(fromList, toList):
106     for i in range(len(fromList)):
107         t = [fromList[i]]
108         tt = frozenset(set(fromList) - set(t))
109         if not tt in toList:
110             toList.append(tt)
111             tt = list(tt)
112             if len(tt) > 1:
113                 getSubset(tt, toList)
114  
115 def calcConf(freqSet, H, supportData, ruleList, minConf=0.7):
116     for conseq in H:  #遍历H中的所有项集并计算它们的可信度值
117         conf = supportData[freqSet] / supportData[freqSet - conseq]  # 可信度计算，结合支持度数据
118         # 提升度lift计算lift = p(a & b) / p(a)*p(b)
119         lift = supportData[freqSet] / (supportData[conseq] * supportData[freqSet - conseq])
120  
121         if conf >= minConf and lift > 1:
122             print(freqSet - conseq, '-->', conseq, '支持度', round(supportData[freqSet], 6), '置信度：', round(conf, 6),
123                   'lift值为：', round(lift, 6))
124             ruleList.append((freqSet - conseq, conseq, conf))
125  
126 # 生成规则
127 def gen_rule(L, supportData, minConf = 0.7):
128     bigRuleList = []
129     for i in range(1, len(L)):  # 从二项集开始计算
130         for freqSet in L[i]:  # freqSet为所有的k项集
131             # 求该三项集的所有非空子集，1项集，2项集，直到k-1项集，用H1表示，为list类型,里面为frozenset类型，
132             H1 = list(freqSet)
133             all_subset = []
134             getSubset(H1, all_subset)  # 生成所有的子集
135             calcConf(freqSet, all_subset, supportData, bigRuleList, minConf)
136     return bigRuleList
137  
138 if __name__ == '__main__':
139     dataSet = data_translation
140     L, supportData = apriori(dataSet, minSupport = 0.02)
141     rule = gen_rule(L, supportData, minConf = 0.35)