手机应用推荐算法及分布式实现研究

发布时间：2018-05-13 03:13

本文选题：个性化推荐 + 可见度　；参考：《浙江大学》2016年硕士论文

【摘要】：随着智能手机的普及,各种各样的手机应用层出不穷,手机应用下载平台面临着严重的信息过载问题。个性化推荐是解决信息过载问题最直接的方法,但是目前大部分应用下载平台的推荐服务还停留在热门榜单推荐上,没有一个个性化的推荐系统。这主要是因为在手机应用推荐场景中,一般传统的商品推荐算法无法取得良好的推荐效果,其推荐的结果有时甚至还不如热门榜单的推荐。除此以外,手机应用的数量过大,使的手机应用推荐算法的算法时间复杂度过高,也是传统推荐算法难以直接应用在手机应用推荐场景中的原因。本文首先介绍了个性化推荐的国内外研究现状,并分析了商品推荐算法不适用于手机应用推荐场景的原因。之后针对以上问题引入了应用可见度,并将其加入到现有的推荐算法中,使得算法能够适应手机应用的推荐场景。并且为了使得推荐算法能够处理超大规模的数据,本文通过对推荐算法的改写,使之能够运行在分布式平台上,加快算法的运行速度。文章的最后通过对比实验,验证了算法引入可见度之后的有效性,并构造了不同的数据集,检验了分布式推荐算法的运行效率。
[Abstract]:With the popularity of smart phones, a variety of mobile applications emerge, mobile application download platform is facing serious information overload problem. Personalized recommendation is the most direct way to solve the problem of information overload, but at present, most of the recommendation services of application and download platform still stay on the hot list recommendation, and there is not a personalized recommendation system. This is mainly because in the mobile phone application recommendation scenario, the general traditional product recommendation algorithm can not achieve good results, its recommendation results are sometimes even worse than the popular list of recommendations. In addition, the number of mobile phone applications is too large, which makes the time complexity of the mobile application recommendation algorithm too high, which is also the reason why the traditional recommendation algorithm is difficult to be directly applied in the mobile phone application recommendation scene. This paper first introduces the research status of personalized recommendation at home and abroad, and analyzes the reason why the commodity recommendation algorithm is not suitable for mobile phone application recommendation. Then, the application visibility is introduced to solve the above problems, and it is added to the existing recommendation algorithm, which makes the algorithm adapt to the mobile phone application recommendation scenario. And in order to enable the recommendation algorithm to deal with large scale data this paper rewrites the recommendation algorithm to make it run on the distributed platform and accelerate the speed of the algorithm. At the end of the paper, the effectiveness of the algorithm is verified by comparing the results with visibility, and different data sets are constructed to test the efficiency of the distributed recommendation algorithm.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.3

【相似文献】