|
| 1 | +# coding:UTF-8 |
| 2 | +''' |
| 3 | +Date:20160928 |
| 4 | +@author: zhaozhiyong |
| 5 | +''' |
| 6 | + |
| 7 | +import numpy as np |
| 8 | +from user_based_recommend import load_data, similarity |
| 9 | + |
| 10 | +def item_based_recommend(data, w, user): |
| 11 | + '''基于商品相似度为用户user推荐商品 |
| 12 | + input: data(mat):商品用户矩阵 |
| 13 | + w(mat):商品与商品之间的相似性 |
| 14 | + user(int):用户的编号 |
| 15 | + output: predict(list):推荐列表 |
| 16 | + ''' |
| 17 | + m, n = np.shape(data) # m:商品数量 n:用户数量 |
| 18 | + interaction = data[:,user].T # 用户user的互动商品信息 |
| 19 | + |
| 20 | + # 1、找到用户user没有互动的商品 |
| 21 | + not_inter = [] |
| 22 | + for i in xrange(n): |
| 23 | + if interaction[0, i] == 0: # 用户user未打分项 |
| 24 | + not_inter.append(i) |
| 25 | + |
| 26 | + # 2、对没有互动过的商品进行预测 |
| 27 | + predict = {} |
| 28 | + for x in not_inter: |
| 29 | + item = np.copy(interaction) # 获取用户user对商品的互动信息 |
| 30 | + for j in xrange(m): # 对每一个商品 |
| 31 | + if item[0, j] != 0: # 利用互动过的商品预测 |
| 32 | + if x not in predict: |
| 33 | + predict[x] = w[x, j] * item[0, j] |
| 34 | + else: |
| 35 | + predict[x] = predict[x] + w[x, j] * item[0, j] |
| 36 | + # 按照预测的大小从大到小排序 |
| 37 | + return sorted(predict.items(), key=lambda d:d[1], reverse=True) |
| 38 | + |
| 39 | +def top_k(predict, k): |
| 40 | + '''为用户推荐前k个商品 |
| 41 | + input: predict(list):排好序的商品列表 |
| 42 | + k(int):推荐的商品个数 |
| 43 | + output: top_recom(list):top_k个商品 |
| 44 | + ''' |
| 45 | + top_recom = [] |
| 46 | + len_result = len(predict) |
| 47 | + if k >= len_result: |
| 48 | + top_recom = predict |
| 49 | + else: |
| 50 | + for i in xrange(k): |
| 51 | + top_recom.append(predict[i]) |
| 52 | + return top_recom |
| 53 | + |
| 54 | +if __name__ == "__main__": |
| 55 | + # 1、导入用户商品数据 |
| 56 | + print "------------ 1. load data ------------" |
| 57 | + data = load_data("data.txt") |
| 58 | + # 将用户商品矩阵转置成商品用户矩阵 |
| 59 | + data = data.T |
| 60 | + # 2、计算商品之间的相似性 |
| 61 | + print "------------ 2. calculate similarity between items -------------" |
| 62 | + w = similarity(data) |
| 63 | + # 3、利用用户之间的相似性进行预测评分 |
| 64 | + print "------------ 3. predict ------------" |
| 65 | + predict = item_based_recommend(data, w, 0) |
| 66 | + # 4、进行Top-K推荐 |
| 67 | + print "------------ 4. top_k recommendation ------------" |
| 68 | + top_recom = top_k(predict, 2) |
| 69 | + print top_recom |
0 commit comments