{"cells":[{"cell_type":"markdown","metadata":{},"source":["# 基于用户的协同过滤算法"]},{"cell_type":"code","execution_count":1,"metadata":{},"outputs":[],"source":["# 导入包\n","import random\n","import math\n","import time\n","from tqdm import tqdm"]},{"cell_type":"markdown","metadata":{},"source":["## 一. 通用函数定义"]},{"cell_type":"code","execution_count":2,"metadata":{},"outputs":[],"source":["# 定义装饰器，监控运行时间\n","def timmer(func):\n"," def wrapper(*args, **kwargs):\n"," start_time = time.time()\n"," res = func(*args, **kwargs)\n"," stop_time = time.time()\n"," print('Func %s, run time: %s' % (func.__name__, stop_time - start_time))\n"," return res\n"," return wrapper"]},{"cell_type":"markdown","metadata":{},"source":["### 1. 数据处理相关\n","1. load data\n","2. split data"]},{"cell_type":"code","execution_count":3,"metadata":{},"outputs":[],"source":["class Dataset():\n"," \n"," def __init__(self, fp):\n"," # fp: data file path\n"," self.data = self.loadData(fp)\n"," \n"," @timmer\n"," def loadData(self, fp):\n"," data = []\n"," for l in open(fp):\n"," data.append(tuple(map(int, l.strip().split('::')[:2])))\n"," return data\n"," \n"," @timmer\n"," def splitData(self, M, k, seed=1):\n"," '''\n"," :params: data, 加载的所有(user, item)数据条目\n"," :params: M, 划分的数目，最后需要取M折的平均\n"," :params: k, 本次是第几次划分，k~[0, M)\n"," :params: seed, random的种子数，对于不同的k应设置成一样的\n"," :return: train, test\n"," '''\n"," train, test = [], []\n"," random.seed(seed)\n"," for user, item in self.data:\n"," # 这里与书中的不一致，本人认为取M-1较为合理，因randint是左右都覆盖的\n"," if random.randint(0, M-1) == k: \n"," test.append((user, item))\n"," else:\n"," train.append((user, item))\n","\n"," # 处理成字典的形式，user->set(items)\n"," def convert_dict(data):\n"," data_dict = {}\n"," for user, item in data:\n"," if user not in data_dict:\n"," data_dict[user] = set()\n"," data_dict[user].add(item)\n"," data_dict = {k: list(data_dict[k]) for k in data_dict}\n"," return data_dict\n","\n"," return convert_dict(train), convert_dict(test)"]},{"cell_type":"markdown","metadata":{},"source":["### 2. 评价指标\n","1. Precision\n","2. Recall\n","3. Coverage\n","4. Popularity(Novelty)"]},{"cell_type":"code","execution_count":4,"metadata":{},"outputs":[],"source":["class Metric():\n"," \n"," def __init__(self, train, test, GetRecommendation):\n"," '''\n"," :params: train, 训练数据\n"," :params: test, 测试数据\n"," :params: GetRecommendation, 为某个用户获取推荐物品的接口函数\n"," '''\n"," self.train = train\n"," self.test = test\n"," self.GetRecommendation = GetRecommendation\n"," self.recs = self.getRec()\n"," \n"," # 为test中的每个用户进行推荐\n"," def getRec(self):\n"," recs = {}\n"," for user in self.test:\n"," rank = self.GetRecommendation(user)\n"," recs[user] = rank\n"," return recs\n"," \n"," # 定义精确率指标计算方式\n"," def precision(self):\n"," all, hit = 0, 0\n"," for user in self.test:\n"," test_items = set(self.test[user])\n"," rank = self.recs[user]\n"," for item, score in rank:\n"," if item in test_items:\n"," hit += 1\n"," all += len(rank)\n"," return round(hit / all * 100, 2)\n"," \n"," # 定义召回率指标计算方式\n"," def recall(self):\n"," all, hit = 0, 0\n"," for user in self.test:\n"," test_items = set(self.test[user])\n"," rank = self.recs[user]\n"," for item, score in rank:\n"," if item in test_items:\n"," hit += 1\n"," all += len(test_items)\n"," return round(hit / all * 100, 2)\n"," \n"," # 定义覆盖率指标计算方式\n"," def coverage(self):\n"," all_item, recom_item = set(), set()\n"," for user in self.test:\n"," for item in self.train[user]:\n"," all_item.add(item)\n"," rank = self.recs[user]\n"," for item, score in rank:\n"," recom_item.add(item)\n"," return round(len(recom_item) / len(all_item) * 100, 2)\n"," \n"," # 定义新颖度指标计算方式\n"," def popularity(self):\n"," # 计算物品的流行度\n"," item_pop = {}\n"," for user in self.train:\n"," for item in self.train[user]:\n"," if item not in item_pop:\n"," item_pop[item] = 0\n"," item_pop[item] += 1\n","\n"," num, pop = 0, 0\n"," for user in self.test:\n"," rank = self.recs[user]\n"," for item, score in rank:\n"," # 取对数，防止因长尾问题带来的被流行物品所主导\n"," pop += math.log(1 + item_pop[item])\n"," num += 1\n"," return round(pop / num, 6)\n"," \n"," def eval(self):\n"," metric = {'Precision': self.precision(),\n"," 'Recall': self.recall(),\n"," 'Coverage': self.coverage(),\n"," 'Popularity': self.popularity()}\n"," print('Metric:', metric)\n"," return metric"]},{"cell_type":"markdown","metadata":{},"source":["## 二. 算法实现\n","1. Random\n","2. MostPopular\n","3. UserCF\n","4. UserIIF"]},{"cell_type":"code","execution_count":5,"metadata":{},"outputs":[],"source":["# 1. 随机推荐\n","def Random(train, K, N):\n"," '''\n"," :params: train, 训练数据集\n"," :params: K, 可忽略\n"," :params: N, 超参数，设置取TopN推荐物品数目\n"," :return: GetRecommendation，推荐接口函数\n"," '''\n"," items = {}\n"," for user in train:\n"," for item in train[user]:\n"," items[item] = 1\n"," \n"," def GetRecommendation(user):\n"," # 随机推荐N个未见过的\n"," user_items = set(train[user])\n"," rec_items = {k: items[k] for k in items if k not in user_items}\n"," rec_items = list(rec_items.items())\n"," random.shuffle(rec_items)\n"," return rec_items[:N]\n"," \n"," return GetRecommendation"]},{"cell_type":"code","execution_count":6,"metadata":{},"outputs":[],"source":["# 2. 热门推荐\n","def MostPopular(train, K, N):\n"," '''\n"," :params: train, 训练数据集\n"," :params: K, 可忽略\n"," :params: N, 超参数，设置取TopN推荐物品数目\n"," :return: GetRecommendation, 推荐接口函数\n"," '''\n"," items = {}\n"," for user in train:\n"," for item in train[user]:\n"," if item not in items:\n"," items[item] = 0\n"," items[item] += 1\n"," \n"," def GetRecommendation(user):\n"," # 随机推荐N个没见过的最热门的\n"," user_items = set(train[user])\n"," rec_items = {k: items[k] for k in items if k not in user_items}\n"," rec_items = list(sorted(rec_items.items(), key=lambda x: x[1], reverse=True))\n"," return rec_items[:N]\n"," \n"," return GetRecommendation"]},{"cell_type":"code","execution_count":7,"metadata":{},"outputs":[],"source":["# 3. 基于用户余弦相似度的推荐\n","def UserCF(train, K, N):\n"," '''\n"," :params: train, 训练数据集\n"," :params: K, 超参数，设置取TopK相似用户数目\n"," :params: N, 超参数，设置取TopN推荐物品数目\n"," :return: GetRecommendation, 推荐接口函数\n"," '''\n"," # 计算item->user的倒排索引\n"," item_users = {}\n"," for user in train:\n"," for item in train[user]:\n"," if item not in item_users:\n"," item_users[item] = []\n"," item_users[item].append(user)\n"," \n"," # 计算用户相似度矩阵\n"," sim = {}\n"," num = {}\n"," for item in item_users:\n"," users = item_users[item]\n"," for i in range(len(users)):\n"," u = users[i]\n"," if u not in num:\n"," num[u] = 0\n"," num[u] += 1\n"," if u not in sim:\n"," sim[u] = {}\n"," for j in range(len(users)):\n"," if j == i: continue\n"," v = users[j]\n"," if v not in sim[u]:\n"," sim[u][v] = 0\n"," sim[u][v] += 1\n"," for u in sim:\n"," for v in sim[u]:\n"," sim[u][v] /= math.sqrt(num[u] * num[v])\n"," \n"," # 按照相似度排序\n"," sorted_user_sim = {k: list(sorted(v.items(), \\\n"," key=lambda x: x[1], reverse=True)) \\\n"," for k, v in sim.items()}\n"," \n"," # 获取接口函数\n"," def GetRecommendation(user):\n"," items = {}\n"," seen_items = set(train[user])\n"," for u, _ in sorted_user_sim[user][:K]:\n"," for item in train[u]:\n"," # 要去掉用户见过的\n"," if item not in seen_items:\n"," if item not in items:\n"," items[item] = 0\n"," items[item] += sim[user][u]\n"," recs = list(sorted(items.items(), key=lambda x: x[1], reverse=True))[:N]\n"," return recs\n"," \n"," return GetRecommendation"]},{"cell_type":"code","execution_count":8,"metadata":{},"outputs":[],"source":["# 4. 基于改进的用户余弦相似度的推荐\n","def UserIIF(train, K, N):\n"," '''\n"," :params: train, 训练数据集\n"," :params: K, 超参数，设置取TopK相似用户数目\n"," :params: N, 超参数，设置取TopN推荐物品数目\n"," :return: GetRecommendation, 推荐接口函数\n"," '''\n"," # 计算item->user的倒排索引\n"," item_users = {}\n"," for user in train:\n"," for item in train[user]:\n"," if item not in item_users:\n"," item_users[item] = []\n"," item_users[item].append(user)\n"," \n"," # 计算用户相似度矩阵\n"," sim = {}\n"," num = {}\n"," for item in item_users:\n"," users = item_users[item]\n"," for i in range(len(users)):\n"," u = users[i]\n"," if u not in num:\n"," num[u] = 0\n"," num[u] += 1\n"," if u not in sim:\n"," sim[u] = {}\n"," for j in range(len(users)):\n"," if j == i: continue\n"," v = users[j]\n"," if v not in sim[u]:\n"," sim[u][v] = 0\n"," # 相比UserCF，主要是改进了这里\n"," sim[u][v] += 1 / math.log(1 + len(users))\n"," for u in sim:\n"," for v in sim[u]:\n"," sim[u][v] /&#
