基于情绪和因子分解的协同过滤算法.docVIP

  基于情绪和因子分解的协同过滤算法 王强强1,2，孟祥武1,2，胡勋1,2** （1. 北京邮电大学智能通信软件与多媒体北京市重点实验室 北京 100876； 5 10 15 20 25 30 35 40 45 2. 北京邮电大学计算机学院 北京 100876） 摘要：面对电子商务领域中电影个性化推荐问题，本文将情绪上下文信息和协同过滤算法相 结合，提出一种基于情绪和因子分解的协同过滤算法。第一种算法首先构造“用户-项目” 二维矩阵，使用 SVD 矩阵分解方法逆向填充稀疏矩阵，然后使用情绪上下文对填充矩阵作 预过滤处理，使用 KNN 协同过滤算法进行推荐预测。第二种算法首先构建“用户-项目-情 绪”三维矩阵，在此矩阵的基础上使用模型公式计算得到用户对项目的推荐值，文中将详细 的介绍使用最小二乘化求解该公式未知变量的方法和步骤，以及如何通过增加情绪信息获 得到更为精确的预测结果。公开数据集实验表明，两种算法均能用于电影服务领域的用户 偏好预测，通过与传统的推荐算法相比表明，它们有更高的推荐精确度。 关键词：推荐系统; 协同过滤; 矩阵分解; 情绪; 因子分解机 中图分类号：TP393 · A collaborative algorithm based on emotional context and factorization WANG Qiangqiang1,2, MENG Xiangwu1,2, HU Xun1,2 (1. Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia, Beijing University of Posts and Telecommunications, Beijing 100876, China; 2. School of Computer Science, Beijing University of Posts and Telecommunications, Beijing 100876, China) Abstract: To solve the problem of movie personalized services recommendation in the field of electronic commerce, a collaborative algorithm based on emotional context is proposed by combining mood context information with collaborative filtering (CF) algorithm. The first algorithm is to build user - project two-dimensional matrix, using SVD matrix decomposition method reverse filling sparse matrix, and then use emotional context to filter the filled matrix. And then use the KNN collaborative filtering algorithm to recommend. The second algorithm is to construct user - project - emotional multidimensional matrix, the model formula used on the multidimensional matrix to predict value. The detailed information on how to use least square to solve the unknown variables will be introduced. Besides, the knowledge on how to obtain more accurate prediction results by increasing the emotional information will also be introduced. Experiments in pubic datasets show that the two algorithms have higher recommend accuracy comparing with the traditional recommendation algorithm and can be used in the field