尚燕飞 陈德运 杨海陆

摘 要：针对移动互联网移动端的推荐系统推荐满意度和精度较低问题，在分析应用信息系统推荐方法的基础上，提出了一种基于用户相似度和主题相似度个性化的移动APP信息推荐方法，该方法采用用户相似度和个性化加权组合的方式生成信息推荐，使推荐信息更为个性化，提高了推荐精度。同时针对多用户公用账号和兴趣发生变化的推荐问题，提出了一种基于复杂兴趣的推荐算法，该方法通过挖掘用户间的相似度、用户的行为操作及兴趣取向，使推荐信息更为准确。同时，相比于性能较好的Popular推荐算法，本算法在准确率上提高了3.91%，召回率提高了3.45%，覆盖率提高了4.84%，性能明显提高。因此，文中所提出方法用于移动APP的个性化推荐，为移动APP的个性化推荐提供了一种新方法。

关键词：移动APP;推荐算法;推荐精度;个性化信息

DOI：10.15938/j.jhust.2018.06.021

中图分类号： TP393.08

文献标志码： A

文章编号： 1007-2683（2018）06-0116-08

Abstract：For the problem of low precision to both experience satisfaction and personalized requirement of Internet mobile terminal based on the recommendation method of analyzing information system a method of mobile APP information recommendation based on user similarity and subject similarity is proposed which generated information recommendation by the weighted combination of user similarity and personalized that the recommended information is more personalized and the recommended accuracy is improved. At the same time a recommendation algorithm based on complex interest is proposed which makes the recommendation information more accurate by mining the similarity between users the behavior of users and the orientation of interest for the recommendation problem of multi-user public account and interest change. Compared with the Popular which has better performance the algorithm improves the accuracy rate by 3.91% the recall rate is 3.45% the coverage rate is improved by 4.84% and the performance is improved obviously. Therefore the method proposed in this paper is used to the personalized recommendation of APP which provides a new method for mobile APP′s personalized recommendation.

Keywords：mobile application;recommendation algorithm;the accuracy of recommending;personalized information

0 引 言

隨着互联网技术的发展，移动端与互联网的迅速融合，越来越多的移动APP出现在了用户的移动设备上。随之而来的移动APP网络信息推荐服务需要响应用户实时、连续、个性化的服务需求。目前，推荐类的系统主要有基于内容的推荐系统、协同过滤推荐系统以及混合推荐系统[1]。基于内容的推荐是根据信息的内容与用户偏好之间的相关性向用户推荐，比较适用于文本信息推荐[2]。协同过滤推荐是根据具有类似观点用户的行为对目标用户进行推荐[3]。混合推荐是为了解决单一推荐技术的不足，按照不同的混合策略（如加权、特征组合等）将不同的推荐技术进行组合从而完成推荐[4]。……