pv_similarity.py

# -*- coding: utf-8 -*-
# @Author: xuyannis
# @Date:   2017-02-25 12:35:15
# @Last Modified by:   xuyannis
# @Last Modified time: 2017-02-25 13:22:54
import numpy as np
from numpy import linalg as la
import pandas as pd

def cosineSimilarity (vec1, vec2):    
    inA = np.mat(vec1) 
    inB = np.mat(vec2)
    num = float(inA * inB.T) #若为行向量: A * B.T
    donom = la.norm(inA) * la.norm(inB) ##余弦值 
    return 0.5+ 0.5*(num / donom) # 归一化

def pearsonSimilar(inA,inB):  
    if len(inA)<3:  
        return 1.0  
    return 0.5+0.5*np.corrcoef(inA,inB,rowvar=0)[0][1]

df = pd.read_csv('30.csv')

vec_pv = df.pv  
vec_pv_real = df.pv_real
vec_cost = df.cost
vec_pv_ad = df.pv_ad

print "PV与无爬虫PV相似度"
print pearsonSimilar(vec_pv, vec_pv_real)

print "PV与消耗相似度"
print pearsonSimilar(vec_pv, vec_cost)

print "无爬虫PV与消耗相似度"
print pearsonSimilar(vec_pv_real, vec_cost)

print "PV与广告PV相似度"
print pearsonSimilar(vec_pv, vec_pv_ad)

print "无爬虫PV与广告PV相似度"
print pearsonSimilar(vec_pv_real, vec_pv_ad)

print "广告PV与广告消耗相似度"
print pearsonSimilar(vec_pv, vec_cost)