itemCF 基於物品的協同過濾

itemCF 基於物品的協同過濾

代碼來自csdn分享的開源。
我加入一些輸出。得到矩陣的內容。

# -*- coding=utf-8 -*-
import math
import sys
import pandas
import numpy as np
from texttable import Texttable
from collections import defaultdict
# from Wtemp import *
from operator import itemgetter


# 讀取文件
def readFile(fileData):
    data = []
    rates = []
    f = open(fileData, "r")
    data = f.readlines()
    f.close()
    for line in data:
        dataLine = line.split("\t")
        rates.append([int(dataLine[0]), int(dataLine[1]), int(dataLine[2])])
    return rates


# 創建字典，生成用戶評分的數據結構
#   輸入：數據集合，格式：用戶id\t硬盤id\t用戶評分
#   輸出:1.用戶字典：dic[用戶id]=[(電影id,電影評分)...]
#        2.電影字典：dic[電影id]=[用戶id1,用戶id2...]
def createDict(rates):
    user_dict = {}
    movie_dict = {}
    for i in rates:
        if i[0] in user_dict:
            user_dict[i[0]].append((i[1], i[2]))
        else:
            user_dict[i[0]] = [(i[1], i[2])]
        if i[1] in movie_dict:
            movie_dict[i[1]].append(i[0])
        else:
            movie_dict[i[1]] = [i[0]]
    user_txt=pandas.Series(user_dict)
    user_txt.to_csv('user_txt.csv')
    item_txt = pandas.Series(movie_dict)
    item_txt.to_csv('movie_txt.csv')
    return user_dict, movie_dict


# 建立物品倒排表,計算物品相似度
def itemCF(user_dict):
    N = dict()
    C = defaultdict(defaultdict)
    W = defaultdict(defaultdict)
    for key in user_dict:
        for i in user_dict[key]:
            if i[0] not in N.keys():  # i[0]表示movie_id
                N[i[0]] = 0
            N[i[0]] += 1  # N[i[0]]表示評論過某電影的用戶數
            for j in user_dict[key]:
                if i == j:
                    continue
                if j not in C[i[0]].keys():
                    C[i[0]][j[0]] = 0
                C[i[0]][j[0]] += 1  # C[i[0]][j[0]]表示電影兩兩之間的相似度，eg：同時評論過電影1和電影2的用戶數
    for i, related_item in C.items():
        for j, cij in related_item.items():
            W[i][j] = cij / math.sqrt(N[i] * N[j])
            data = [i,j,W[i][j]]
            print >>f,(data)
    return W


# 結合用戶喜好對物品排序
def recommondation(user_id, user_dict, K):
    rank = defaultdict(int)
    l = list()
    W = itemCF(user_dict)
    for i, score in user_dict[user_id]:  # i爲特定用戶的電影id，score爲其相應評分
        for j, wj in sorted(W[i].items(), key=itemgetter(1), reverse=True)[0:K]:  # sorted()的返回值爲list,list的元素爲元組
            if j in user_dict[user_id]:
                continue
            rank[j] += score * wj  # 先找出用戶評論過的電影集合，對每一部電影id，假設其中一部電影id1,找出與該電影最相似的K部電影，計算出在id1下用戶對每部電影的興趣度，接着迭代整個用戶評論過的電影集合，求加權和，再排序，可推薦出前n部電影，我這裏取10部。
    l = sorted(rank.items(), key=itemgetter(1), reverse=True)[0:10]
    return l


# 獲取電影列表
def getMovieList(item):
    items = {}
    f = open(item, "r")
    movie_content = f.readlines()
    f.close()
    for movie in movie_content:
        movieLine = movie.split("\t")
        items[int(movieLine[0])] = movieLine[1:]
    item_txt = pandas.Series(items)
    item_txt.to_csv('item_txt.csv')
    return items


# 主程序
if __name__ == '__main__':
    itemTemp = getMovieList("D:/PycharmProjects/reMov/artists.item")  # 獲取電影列表
    fileTemp = readFile("D:/PycharmProjects/reMov/user_artists.data")  # 讀取文件
    user_dic, movie_dic = createDict(fileTemp)  # 創建字典
    #numpy.savetxt('user_dict.txt', user_dic)
    print("創建字典")
    user_id = 66
    movieTemp = recommondation(user_id, user_dic, 80)  # 對電影tuijian排序
    movieTemp_txt = pandas.Series(movieTemp)
    movieTemp_txt.to_csv('movieTemp_txt.csv')
    print("創建字典2")
    rows = []
    table = Texttable()  # 創建表格並顯示
    table.set_deco(Texttable.HEADER)
    table.set_cols_dtype(['t', 'f', 'a'])
    table.set_cols_align(["l", "l", "l"])
    rows.append(["user name", "recommondation_movie", "from userid"])
    for i in movieTemp:
        rows.append([user_id, itemTemp[i[0]][0], ""])
    table.add_rows(rows)
    print(table.draw())