一种基于协同过滤的APPS跨类别推荐算法

发布时间：2018-04-28 14:57

本文选题：协同过滤 + 智能应用推荐　；参考：《情报科学》2017年12期

【摘要】：【目的/意义】针对主流APPS推荐系统一般仅能推荐同类别APPS的现状,提出了一种基于协同过滤的APPS跨类别推荐算法(APPSR)。【方法/过程】该算法先对APPS进行聚类,考虑APPS簇间相似度,对未评分APPS进行评分预测,构建无缺失的用户-APPS评分矩阵。在传统协同过滤技术的基础之上,引入了时间权重函数与热门APPS惩罚机制,体现了用户兴趣的时效性,消除了热门APPS对推荐结果的影响。根据不同用户对多种APPS的评分,预测用户对其它类别APPS的喜好,为用户提供跨类别的APPS个性化推荐。【结果/结论】通过实验验证了APPSR算法的可行性与有效性,为APPS个性化推荐提供了新方法。
[Abstract]:[objective / significance] in view of the current situation that mainstream APPS recommendation systems can only recommend the same APPS, a cross-category APPS recommendation algorithm based on collaborative filtering is proposed. [method / process] this algorithm firstly clusters APPS. Considering the similarity between APPS clusters, the ungraded APPS is predicted and the user-Apps scoring matrix is constructed. Based on the traditional collaborative filtering technology, time weight function and popular APPS penalty mechanism are introduced, which reflect the timeliness of user interest and eliminate the influence of popular APPS on the recommended results. According to the scores of different users on various APPS, the preference of users to other types of APPS is predicted, and the cross-category APPS personalized recommendation is provided to users. [results / conclusions] the feasibility and effectiveness of APPSR algorithm are verified by experiments. It provides a new method for APPS personalized recommendation.
【作者单位】：江苏大学信息管理与信息系统系;江苏大学计算机科学与通信工程学院;江苏大学电子商务系;
【基金】：国家自然科学基金项目(71573107) 江苏省自然科学基础研究计划基金项目(BK20150531) 江苏省博士后科研资助计划项目(1401056C) 全国统计科学研究项目(2016LY17) 江苏大学高级人才基金项目(13JDG127)
【分类号】：TP391.3

【相似文献】