04_construct_evaluate_vec.py

import pandas as pd
import numpy as np
import networkx as nx
from collections import Counter
import sys
"""
generate the test evaluation data 
Output file:
uninteract_vec_element

"""


def prepare_vector_element_test(G, relation, filename):
    f = open('./Data/' + filename + '.txt', 'wb')
    for r in relation.values:
        line = ''
        user, item = r
        line = line + user.split('_')[1] + ','
        line = line + item.split('_')[1] + '|'
        try:
            user2item_neighbor = nx.neighbors(G, item)
            np.random.shuffle(user2item_neighbor)
            if len(user2item_neighbor) > Max_Num_Neighbor:
                user2item_neighbor = user2item_neighbor[:Max_Num_Neighbor]
        except:
            user2item_neighbor = []

        user2item_neighbor = ','.join([str(x.split('_')[1]) for x in user2item_neighbor])
        line = line + user2item_neighbor + '|'

        try:
            item2user_neighbor = nx.neighbors(G, user)

            np.random.shuffle(item2user_neighbor)
            if len(item2user_neighbor) > Max_Num_Neighbor:
                item2user_neighbor = item2user_neighbor[:Max_Num_Neighbor]
        except:
            item2user_neighbor = []

        item2user_neighbor = ','.join([str(x.split('_')[1]) for x in item2user_neighbor])
        line = line + item2user_neighbor

        f.writelines(line)
        f.writelines('\n')
    f.close()


if __name__=="__main__":
    print '=================04: Generating test data ==================='
    # N: the number of negative samples in the testing ranking list
    #N = 99
    #Max_Num_Neighbor =50
    N =int(sys.argv[1])
    Max_Num_Neighbor =int(sys.argv[2])
    print 'the number of negative samples in the testing ranking list: ',N
    print 'max_neighbors: ', Max_Num_Neighbor
    relation_train_positive = pd.read_csv('./Data/relation_train_positive.csv')
    relation_test = pd.read_csv('./Data/relation_test.csv')
    all_item = pd.read_csv('./Data/all_item.csv')
    all_item = all_item.item.values
    G_all = nx.Graph()
    G_all.add_edges_from(relation_train_positive.values)
    G_all.add_edges_from(relation_test.values)
    test_user = relation_test.user.values
    uninteract_user_list = []
    uninteract_item_list = []
    for user in test_user:
        interact_item = nx.neighbors(G_all,user)
        un_interact_item = list(set(all_item) - set(interact_item))
        np.random.shuffle(un_interact_item)
        if len(un_interact_item) < N:
            un_interact_item = un_interact_item * (int(N / len(un_interact_item))+1)
        un_interact_item_choose = un_interact_item[:N]
        uninteract_user_list += [user] * N
        uninteract_item_list += un_interact_item_choose
    uninteract = pd.DataFrame(np.array([uninteract_user_list,uninteract_item_list]).T, columns=['user', 'item'])
    G = nx.Graph()
    G.add_edges_from(relation_train_positive.values)
    prepare_vector_element_test(G, relation_test, 'relation_test_vec_element')
    prepare_vector_element_test(G, uninteract, 'uninteract_vec_element')
    print 'test samples:', len(relation_test)
    print 'test uninteract samples:', len(uninteract)
    print 'Generate test data successfully!'