Apriori算法在购物篮分析中的运用

　　购物篮分析是一个很经典的数据挖掘案例,运用到了Apriori算法。下面从网上下载的一超市某月份的数据库,利用Apriori算法进行管理分析。例子使用Python+MongoDB

　　处理过程1 数据建模(将Excel中的数据写入到MongoDB数据库), 2 从数据库中读取数据进行分析。

　　Excel文件http://download.csdn.net/detail/artscrafts/6805689

　　案例配置文件 setting.py

data_source = 'supermarket.xls'
host = 'localhost'
port = 27017
db_name = 'shopping_basket'
items_name = 'goods_items'
record_name = 'transaction_record'

　　读取Excel数据到MongoDB中 load_basket.py

from xlrd import open_workbook
from pymongo import MongoClient
import setting

wb = open_workbook(setting.data_source, encoding_override='utf-8')
client = MongoClient(setting.host, setting.port)
db = client[setting.db_name]
items = []

#read xls
def read_one_line(workbook, sheet_index=0, row_index=0, start_col_index=0):
    sheet = workbook.sheets()[0]
    max_row = sheet.nrows
    max_col = sheet.ncols
    start_col_index = (start_col_index if (start_col_index > 0 and start_col_index <= max_col) else max_col)
    if row_index < 0 or row_index >= max_row:
        raise IndexError()
    for col_index in xrange(start_col_index, max_col):
        yield sheet.cell(row_index, col_index).value

#read xls
def readlines(workbook, sheet_index=0, start_row_index=0, end_row_index=None, start_col_index=0, end_col_index=None):
    sheet = workbook.sheets()[sheet_index]
    max_row = sheet.nrows
    max_col = sheet.ncols
    end_row_index = (end_row_index if end_row_index  else max_row)
    end_col_index = (end_col_index if end_col_index  else max_col)
    for row_index in xrange(start_row_index, end_row_index):
        yield [sheet.cell(row_index, col_index).value for col_index in xrange(start_col_index, end_col_index)]

#from xls to mongodb
def load_items():
    collection = db[setting.items_name]
    items_line = read_one_line(wb, row_index=1, start_col_index=1)
    id = 1
    tmp = []
    for item in items_line:
        if id % 100 == 0:
            collection.insert(tmp)
            tmp = []
        tmp.append({'id':id, 'name':item})
        items.append(item)
        id += 1

# from xls to mongodb
def load_record():
    collection = db[setting.record_name]
    lines = readlines(wb,start_row_index=2, start_col_index = 1)
    tmp = []
    id = 1
    for line in lines:
        if id % 100 == 0:
            collection.insert(tmp)
            tmp = []
        tmp.append({'id':id, 'items':[items[i] for i in xrange(len(line)) if line[i] == 'T']})
        id += 1


def main():
    print '........start loading........'
    load_items()
    load_record()
    client.close()
    print '.........end loading.........'

if __name__ == '__main__':
    main()

　　进行数据分析 analysis_basket.py

#Apriori
from pymongo import MongoClient
import setting

client = MongoClient(setting.host, setting.port)
db = client[setting.db_name]
data = []

#from mongodb to items
def filldata():
    collection = db[setting.record_name]
    cur = collection.find()
    for row in cur:
        data.append(row['items'])

def connect(items):
    result = {}
    keys = items.keys()
    length = len(keys)
    for i in range(length):
        prev = keys[i][:len(keys[i]) - 1]
        for j in range(i + 1, length):
            tmp = keys[j][:len(keys[j]) - 1]
            if prev == tmp:
                key = keys[i] + (keys[j][len(keys[i]) - 1],)
                result[key] = getsupp(key)
            else:
                break
    return result


def pruning(items, minsupp):
    result = {}
    for key in items.keys():
        if items[key] >= minsupp:
            result[key] = items[key]
    return result

def contain(par, sub):
    for v in sub:
        if not v in par:
            return False
    return True


def getsupp(item):
    supp = 0
    for row in data:
        if contain(row, item):
            supp+=1
    return supp

def apriori(data, minsupp, k):
    candidate_set = {}
    for row in data:
        for i in row:
            key = (i,)
            candidate_set[key] = candidate_set.get(key, 0) + 1
    frequently_set = pruning(candidate_set, minsupp)
    result = {}
    result['k=1'] = frequently_set
    for n in range(2, k):
        candidate_set = connect(frequently_set)
        frequently_set = pruning(candidate_set, minsupp)
        if len(frequently_set) <= 1:
            return result
        result['K=' + str(n)] = frequently_set
    return result

def main():
    filldata()
    client.close()
    res = apriori(data, 30, 8)


if __name__ == '__main__':
    main()