ItemPop.py

import numpy as np
import pandas as pd
from time import time
from evaluation import eval_model_pro
import sys
import codecs


def item_freq(train_ratings):
    pos_ratings = train_ratings[train_ratings[:,2]>0]
    item_dict = dict()

    for rate in pos_ratings:
        uid = int(rate[0])
        iid = int(rate[1])

        if iid in item_dict.keys():
            item_dict[iid] += 1
        else:
            item_dict[iid] = 1

    # item_id = sorted(item_dict, key=item_dict.__getitem__, reverse=True)
    return item_dict


def evaluate(data, item_freq_dict):
    sample_num = data.shape[0]
    print("test/valid data number:" + str(sample_num))

    y_gnd = data[:, 2:]

    y_pred = np.zeros(y_gnd.shape)

    for j in range(sample_num):
        i_id = int(data[j, 1])
        if i_id in item_freq_dict.keys():
            i_freq = item_freq_dict[i_id]
        else:
            i_freq = 1
        y_pred[j, 0] = i_freq

    hit_scores = []
    ndcg_scores = []
    for k in range(1, 16):
        hits, ndcgs = eval_model_pro(y_gnd, y_pred, K=k, row_len=100+1)
        hit_scores.append(hits)
        ndcg_scores.append(ndcgs)
    return hit_scores, ndcg_scores


if __name__ == '__main__':
    t1 = time()

    path = '../Data/'
    test_file_path = "baseModel_test.txt"
    train_ratings = np.loadtxt(path + 'baseModel_train.txt')
    test_ratings = np.loadtxt(path + test_file_path)

    print('already load train and test data')

    item_freq_dict = item_freq(train_ratings)
    item_num = len(item_freq_dict.keys())

    test_hits, test_ndcgs = evaluate(test_ratings, item_freq_dict)

    print("cost time:", time()-t1)
    print("hit scores:", test_hits)
    print("ndcg scores:", test_ndcgs)

    result_writer = codecs.open("ItemPop.res", mode="w", encoding="utf-8")
    result_writer.write("hit scores:")
    for hit_score in test_hits:
        result_writer.write("%.5f\t" % hit_score)
    result_writer.write("\n")

    result_writer.write("ndcg scores:")
    for ndcg_score in test_ndcgs:
        result_writer.write("%.5f\t" % ndcg_score)
    result_writer.write("\n")