基于时间上下文的移动应用推荐系统研究与应用

发布时间：2018-06-06 01:38

本文选题：推荐系统 + 图模型　；参考：《山东农业大学》2017年硕士论文

【摘要】：近年来,互联网尤其是移动互联网规模和技术发展迅猛,智能移动设备如智能手机、平板电脑等大量普及,智能手机用户数量剧增。移动应用作为智能手机的重要组成部分,改变了用户的生活方式、工作方式、学习方式等,在系统设计与开发中占有重要地位。移动应用市场如Google Play Store、Apple App Store等中的移动应用数量达到了百万级别,海量的移动应用在给用户提供便利、开发人员提供参考的同时,也带来了新的挑战。从用户角度来看,在海量应用中找到自己感兴趣的应用面临巨大困难;从开发人员角度来看,在海量应用中选取合适的应用作为参考耗费大量精力。与此同时,智能手机用户低龄化是一个不可忽视的趋势,如何防止低龄用户沉迷于智能手机是一个亟待解决的问题。本文对移动应用个性化推荐进行研究,首先对推荐算法进行了改进,提出了一种基于用户分裂的资源扩散算法。在调研个性化推荐算法的过程中,发现用户在时间维度上积累的兴趣偏移对推荐算法的准确率影响很大。在基于图模型的资源扩散算法的基础上,将时间上下文信息融入算法中,提出了一种基于用户分裂的资源扩散算法,改进了传统的资源扩散算法。改进算法把兴趣随时间发生变化的用户看作多个用户,即采用用户分裂思想引入时间上下文信息,更加充分的利用了数据信息,通过实验对比发现,和传统的资源扩散算法相比,改进算法的准确率明显提高。然后将基于用户分裂的资源扩散算法应用于移动应用推荐领域,设计了一种面向智能手机用户和开发人员的移动应用推荐系统并实现了系统原型。对用户进行移动应用的个性化推荐时,不仅使用个性化推荐算法迎合用户的兴趣偏好,而且根据用户个人信息中的人口统计学特征,如年龄、性别等提供了不同的推荐策略;为方便开发人员在海量的移动应用中寻找参考信息,系统为开发人员提供了应用的用户画像、基于用户特征的应用查询、移动应用之间的关联度等。最后从信息论的角度,看待移动应用产生、生长、衰亡,进行系统维护。
[Abstract]:In recent years, the scale and technology of the Internet, especially the mobile Internet, have developed rapidly. Smart mobile devices such as smartphones and tablets have been widely used, and the number of smartphone users has increased dramatically. As an important part of smart phone, mobile application has changed the user's life style, working style, learning style and so on, which plays an important role in the system design and development. The number of mobile applications in the mobile application market, such as Google Play Store App Store, has reached a million-level. The vast amount of mobile applications provide convenience to users and provide reference to developers, but also bring new challenges. From the user's point of view, it is difficult to find the application that is interested in the mass application; from the developer's point of view, it takes a lot of energy to select the appropriate application as the reference in the mass application. At the same time, the younger age of smartphone users is a trend that can not be ignored, and how to prevent young users from indulging in smartphones is an urgent problem to be solved. In this paper, personalized recommendation for mobile applications is studied. Firstly, the recommendation algorithm is improved, and a resource diffusion algorithm based on user splitting is proposed. In the process of investigating the personalized recommendation algorithm, it is found that the interest offset accumulated by the user in the time dimension has a great influence on the accuracy of the recommendation algorithm. On the basis of the resource diffusion algorithm based on graph model, the time context information is incorporated into the algorithm, and a resource diffusion algorithm based on user splitting is proposed, which improves the traditional resource diffusion algorithm. The improved algorithm regards users whose interests change with time as multiple users, that is, using the idea of user splitting to introduce time context information, which makes full use of the data information. Compared with the traditional resource diffusion algorithm, the accuracy of the improved algorithm is obviously improved. Then, the resource diffusion algorithm based on user splitting is applied to mobile application recommendation field, and a mobile application recommendation system for smart phone users and developers is designed and implemented. In the process of personalized recommendation for mobile applications, not only the personalized recommendation algorithm is used to cater to the interests and preferences of users, but also different recommendation strategies are provided according to the demographic characteristics of users' personal information, such as age, gender, etc. In order to facilitate developers to find reference information in a large number of mobile applications, the system provides developers with user portraits, application queries based on user characteristics, and the correlation between mobile applications, etc. Finally, from the point of view of information theory, the generation, growth, decline and system maintenance of mobile applications are discussed.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】