记录第一次数据预处理过程

记录第一次数据预处理过程

数据描述

item_idx和user_idx是商品和用户名分别对应的id。train_pair.json 是一些二元组[u,i]的列表，表示用户u喜欢商品i，是我们的训练数据（这里你需要自己构造负例）。valid_pair和test_pair里面的格式一样，也是一些二元组，是一个用户对应一个正例商品和100个负例商品，我们任务的目的是给这个用户的这101个商品排序，然后看那一个整理商品的排序位置。

最终数据目标

类似于gowalla中 train.txt 是

每行第一个是user 后面是正例item

test.txt 中: user - 要测试的items

从list到txt的remap:

import torch
import numpy as np
from torch.utils.data import Dataset
import json
import string

org_path = "D:\GraphNetworkCF-main\Data\men\\user_idx.json"
des_path = "D:\GraphNetworkCF-main\Data\men\\user_list.txt"

with open(des_path, "a") as fl:
    fl.write("org_id remap_id\n")

with open(org_path,'r',encoding='utf8')as fp:
    json_datas = json.load(fp)
    #print('这是文件中的json数据：',json_datas)
    #print('这是读取到文件数据的数据类型：', type(json_datas))
    length = len(json_datas)
    count = 1
    for json_data in json_datas:
        with open(des_path, "a") as fl:
            fl.write(str(json_data)+" "+str(count)+"\n")
        count = count + 1

映射，转化为索引应该比字符串好处理了

一共4807个user,43832个item

train_pair 到 user-items 初步

import torch
import numpy as np
from torch.utils.data import Dataset
import json
import string
import textwrap
import fileinput
import os

org_path_user = "D:\GraphNetworkCF-main\Data\men\\user_list.txt"
org_path_item = "D:\GraphNetworkCF-main\Data\men\\item_list.txt"
org_path_train_pari = "D:\GraphNetworkCF-main\Data\men\\train_pair.json"
des_path = "D:\GraphNetworkCF-main\Data\men\\train.txt"


#coding:utf-8


with open(org_path_user,'r',encoding='utf-8') as f:
        user_dic=[]
        for line in f.readlines():
            line=line.strip('\n') #去掉换行符\n
            b=line.split(' ') #将每一行以空格为分隔符转换成列表
            user_dic.append(b)

user_dic=dict(user_dic)

with open(org_path_item, 'r', encoding='utf-8') as f:
    item_dic = []
    for line in f.readlines():
        line = line.strip('\n')  # 去掉换行符\n
        b = line.split(' ')  # 将每一行以空格为分隔符转换成列表
        item_dic.append(b)

item_dic=dict(item_dic)

with open(org_path_train_pari,'r',encoding='utf8')as fp:
    json_datas = json.load(fp)
    #print('这是文件中的json数据：',json_datas)
    #print('这是读取到文件数据的数据类型：', type(json_datas))
    length = len(json_datas)
    #print(length)
    # 0 是 user  1是item
    #print(json_datas[2][0])
    now_user_id = -1
    for json_data in json_datas:
        nxt_user = json_data[0]
        nxt_item = json_data[1]
        nxt_user_id = user_dic[nxt_user]
        item_id = item_dic[nxt_item]
        #print(nxt_user_id)
        #print(item_id)
        if now_user_id != nxt_user_id:
            # 第一次
            now_user_id = nxt_user_id
            with open(des_path, "a") as fl:
                fl.write("\n")
                fl.write(str(now_user_id) + " " + str(item_id))
        else:
            with open(des_path, "a") as fl:
                fl.write(" " + str(item_id))

进一步处理train.txt

import torch
import numpy as np
from torch.utils.data import Dataset
import json
import string
import textwrap
import fileinput
import os

org_path_train = "D:\GraphNetworkCF-main\Data\men\\test.txt"
des_path = "D:\GraphNetworkCF-main\Data\men\\test_1.txt"

trains = {}
with open(org_path_train,'r',encoding='utf-8') as f:
        line_dic=[]
        for line in f.readlines():
            if len(line) == 0:break
            line=line.strip('\n') #去掉换行符\n
            items = [int(i) for i in line.split(' ')]  # 物品
            uid, train_items = items[0], items[1:]
            train_items.sort()
            trains[uid] = train_items

        #print(trains.items())
        train = sorted(trains.items(), key=lambda d: d[0])
        # print(train)
        # print(type(train))
        #count = 4807; # 0 - 4806
        #print(max(train))
        #print(len(train))
        # print(type(train[count]))
        # print(train[count][1])
        length = len(train) # 1 - 4807

        with open(des_path, "a") as fl:
            for i in range(0,length):
                fl.write("\n")
                uid = i # 0 - 4806
                u_items = train[i][1]
                #print(u_items)
                fl.write(str(uid)+" ")
                for item in u_items:
                    fl.write(str(item)+" ")

最终效果