简单的推荐系统搭建--Python（pearson,欧几里得相似度）

#coding=utf-8
from math import sqrt
##--------------以下使用欧几里得计算用户之间的相似度--------------------##
def sim_distance(prefer, person1, person2):
    sim = {}
    for item in prefer[person1]:
        if item in prefer[person2]:
            sim[item] = 1  #添加共同项到字典中

    #无共同项，返回0
    if len(sim)==0:
        return 0

    #计算所有共有项目的差值的平方和
    sum_all = sum([pow(prefer[person1][item]-prefer[person2][item], 2) for item in sim])
    #返回改进的相似度函数
    return 1/(1+sqrt(sum_all))

##-----------------------pearson相关度系数------------------------##
def sim_pearson(prefer, person1, person2):
    sim = {}
    #查找双方都评价过的项,并将相同的项添加到字典sim中
    for item in prefer[person1]:
        if item in prefer[person2]:
            sim[item] = 1

    #元素的个数
    n = len(sim)

    if len(sim)==0:
        return -1

    #所有偏好之和
    sum1 = sum([prefer[person1][item] for item in sim])
    sum2 = sum([prefer[person2][item] for item in sim])

    #求平方和
    sum1Sq = sum( [pow(prefer[person1][item] ,2) for item in sim] )
    sum2Sq = sum( [pow(prefer[person2][item] ,2) for item in sim] )

    #求乘积之和 ∑XY
    sumMulti = sum([prefer[person1][item]*prefer[person2][item] for item in sim])

    num1 = sumMulti - (sum1*sum2/n) #计算∑XY-∑X∑Y/n
    num2 = sqrt( (sum1Sq-pow(sum1,2)/n)*(sum2Sq-pow(sum2,2)/n)) #计算sqrt((∑x*x -(∑x*x/n))(∑y*y -(∑y*y/n)))

    if num2==0:
        return 0
    return num1/num2

##-----------------获取用户相似度（从用户评价字典中返回Top-K匹配者，K,相似度函数 为可选参数）--------------------------##
def topMatches(prefer, person, n=1, similarity=sim_pearson):
    scores=[ (similarity(prefer,person,other),other) for other in prefer if other!=person ]
    #对scores列表排序,从高到底
    scores.sort()    
    scores.reverse()

    #返回排序列表， 注意[0:n],仅返回前n项；
    return scores[0:n]               

##--------------------获取推荐（提供推荐，利用所有人评价的加权均值。相似度高，影响因子越大。）---------------------------##
def getRecommendations(prefer, person, similarity=sim_pearson):
    totals = {}
    simSums = {}
    for other in prefer:
        if other == person:
            continue
        else:
            sim = similarity(prefer, person, other)    #计算比较其他用户的相似度

        #相似度>0
        if sim<=0: continue
        for item in prefer[other]:
            if item not in prefer[person]:
            #加权评价值：相似度*评价值
                totals.setdefault(item,0)  #每轮循环开始时初始化为0
                totals[item] += prefer[other][item]*sim
                #相似度之和
                simSums.setdefault(item,0)
                simSums[item] += sim

    #建立归一化列表,注意此处归一化并没有将最后的结果限制在0-1之间， sigmoid函数可以将最后的结果设置为0-1之间，
    #还有一种最简单的方法进行归一化，将分母取为所有totals[item]的最大值即可
    ##使用ranks=[]表明最后ranks返回的是一个列表，列表的每一项为(total/simSums[item],item)
    ##其中total/simSums[item] = totals[item]/simSums[item] 分子是加权相似度，分母只是相似度之和
    ##这是分子和分母的唯一区别
    ranks = [ (total/simSums[item],item) for item,total in totals.items() ] 

    #返回经排序后的列表
    ranks.sort()
    ranks.reverse()
    return ranks

if __name__ == "__main__":
    print("\n测试计算欧几里得距离的方法sim_distance()....")

    #数据集格式
    Prefer = {"tommy":{'War':2.3,'The lord of wings':3.0,'Kongfu':5.0}, 
    "lily":{'War':2.0,'The lord of wings':3.6,'Kongfu':4.1},
    "jim":{'War':1.9,'The lord of wings':4.0,'Beautiful America':4.7,'the big bang':1.0},
    "jack":{'War':2.8,'The lord of wings':3.5,'Kongfu':5.5}}

    print("sim_distance(dic,'lily','jim') = ",sim_distance(Prefer, 'lily', 'jack'))
    print("sim_distance(dic,'tommy','jim') = ",sim_distance(Prefer, 'tommy', 'jim'))
    print("sim_distance(dic,'tommy','lily') = ",sim_distance(Prefer, 'tommy', 'lily'))
    print("sim_distance(dic,'tommy','jack') = ",sim_distance(Prefer, 'tommy', 'jack'))

    print("\n测试计算Pearson系数的方法sim_pearson()....")
    print("sim_pearson(dic,'lily','jim') = ",sim_pearson(Prefer, 'lily', 'jim'))
    print("sim_pearson(dic,'tommy','jim') = ",sim_pearson(Prefer, 'tommy', 'jim'))
    print("sim_pearson(dic,'tommy','lily') = ",sim_pearson(Prefer, 'tommy', 'lily'))
    print("sim_pearson(dic,'tommy','jack') = ",sim_pearson(Prefer, 'tommy', 'jack'))

    print("\n测试topMatches()方法......")
    print(topMatches(Prefer, 'tommy'))

    print("\n测试推荐方法getRecommendations(prefer, person, similarity=sim_pearson)......")
    print(getRecommendations(Prefer, 'tommy'))