python寻找相似用户_python手撸代码计算用户相似度

一、代码实例：

import math

class UserPerference:

def __init__(self,username):

self.sku_perf_vector={}

self.username=username

def addSKUPerf(self,sku_no,perf_val):

self.sku_perf_vector.setdefault(sku_no,perf_val)

def print_info(self):

print("username:%s" %(self.username));

print(self.sku_perf_vector)

def get_all_sku_nos(self):

#return dictionary type

return self.sku_perf_vector.keys();

def calcUserBasedSimilarity(up1:UserPerference,up2:UserPerference):

'''

****使用余弦距离****

Notice:如果两个对象中维度差别很大(差集较多)，计算出来结果就比较小，所以可以选取少量的top指标进行差集计算。

推荐步骤：step1：计算用户相似度(可以按照SKU或者分类或者标签来计算)， step2：进行相互推荐

***下表是用户和商品喜好程度的一个矩阵表格***

---------------------------------------------------

|User/SKU |- SKU1 -|-- SKu2 -|--Sku3-|-- Sku4--|

---------------------------------------------------

| B | 3 | 32 | 32 | 64 |

---------------------------------------------------

| A | 2 | 3 | 32 | 43 |

---------------------------------------------------

基于用户推荐算法计算相似度，计算公司为：similarity=(A+B)/|A|*|B|

'''

p1_skus=up1.get_all_sku_nos();

p2_skus=up2.get_all_sku_nos();

#求sku的差集，然后补齐数据

p2_diff_coll=p1_skus-p2_skus;

p1_diff_coll=p2_skus-p1_skus;

for sku_no in p1_diff_coll:

up1.addSKUPerf(sku_no,0)

for sku_no in p2_diff_coll:

up2.addSKUPerf(sku_no, 0)

molecules=0;

denominators=0;

sku_nos=up1.get_all_sku_nos()

for sku_no in sku_nos:

molecules=molecules+up1.sku_perf_vector.__getitem__(sku_no)*up2.sku_perf_vector.__getitem__(sku_no);

denominators_up1=0;

denominators_up2=0;

for sku_no in sku_nos:

denominators_up1=denominators_up1+math.pow(up1.sku_perf_vector.__getitem__(sku_no),2)

for sku_no in sku_nos:

denominators_up2 = denominators_up2 + math.pow(up2.sku_perf_vector.__getitem__(sku_no), 2)

denominators=math.sqrt(denominators_up1)*math.sqrt(denominators_up2)

print("分之为:{}，分母为:{}".format(molecules,denominators))

return molecules/denominators;

def calcItemBasedSimilarity():

'''

推荐步骤：step1：计算物品相似度， step2：根据用户偏好进行推算

数据矩阵：

基于商品的推荐算法计算相似度，计算公式为：similarity=N(A与B)/√￣N(A)*N(B)

'''

None

user_p1=UserPerference('liuyi');

user_p1.addSKUPerf(2,22)

user_p1.addSKUPerf(3,33)

user_p1.addSKUPerf(4,30)

'''user_p1.addSKUPerf(5,30)

user_p1.addSKUPerf(6,30)

user_p1.addSKUPerf(7,30)

user_p1.addSKUPerf(8,30)

user_p1.addSKUPerf(9,30)

'''

#user_p1.addSKUPerf(11,30)

user_p2=UserPerference('liuchen');

user_p2.addSKUPerf(2,20)

user_p2.addSKUPerf(3,30)

user_p2.addSKUPerf(4,3)

'''user_p2.addSKUPerf(5,30)

user_p2.addSKUPerf(6,30)

user_p2.addSKUPerf(7,30)

user_p2.addSKUPerf(8,30)

user_p2.addSKUPerf(9,30)

'''

#user_p2.addSKUPerf(10,30)

print(set(user_p1.sku_perf_vector));

print(set(user_p2.sku_perf_vector));

###有一种组合计算出来居然值>1,为1.00000000002

print("similarity is:{}".format(calcUserBasedSimilarity(user_p1,user_p2)));

二、运行结果：

分之为:1520，分母为:1799.2101044625108

similarity is:0.8448151754094773