基于权限和功能的APP个性化推荐算法的研究

发布时间：2018-03-27 10:38

本文选题：App推荐算法　切入点：权限隐私　出处：《武汉大学》2017年硕士论文

【摘要】：近几年随着科学技术的飞跃发展,智能手机的普及使手机App呈现指数级增长,在纷繁的App中如何帮助用户选择感兴趣的应用已成为推荐系统中的热门话题。在有关App推荐的已有系统中,大部分基于热门度、下载使用量进行推荐,因此极大程度推荐了非用户所感兴趣的App,同时越来越多的用户开始关注涉及个人安全隐私的App权限。目前结合App权限、功能和用户兴趣三方面的推荐系统较少,大部分推荐系统只考虑权限隐私而没有考虑用户兴趣做Top-N推荐,或者只将App权限隐私、功能和用户兴趣进行简单的线性组合做Top-N推荐,均隔离了App权限和用户兴趣之间的关联性。本文提出了基于App权限、功能和用户兴趣相结合的矩阵分解算法(MFPF),利用两者关联性进行App个性化推荐。本文工作的主要贡献有以下几方面:(1)结合本文实验数据进行App权限分析,分析了普通App所需权限的数量、权限种类,以及普通App与恶意App所需权限的情况。(2)分析App权限隐私与用户评分的关联性,通过App-permission二分图,建立ARSM方法量化App的危险分值,并初步验证权限隐私和用户兴趣之间存在的关联性。(3)建立兼具权限隐私和用户兴趣的App推荐模型,提出一种新颖的基于权限隐私和功能兴趣的矩阵分解算法MFPF,通过结合App权限面、App功能属性面及用户兴趣面实现App推荐。(4)分析权限在App用户评分中的比重及影响,进一步分析权限隐私与用户评分的关联性,即证明权限隐私和用户兴趣之间的关联性。本文实验数据由安智市场的App和相应用户评分组成。实验结果表明,与传统的推荐方法相比,本文提出的推荐算法推荐效果更好;与最新一篇同样采用权限和功能的App推荐系统相比,本文算法的精确度更高。
[Abstract]:In recent years, with the rapid development of science and technology, the popularity of smart phones has led to an exponential increase in mobile phone App. How to help users choose interesting applications in the numerous App has become a hot topic in the recommendation system. In the existing system of App recommendation, most of them are based on the popularity of download usage. As a result, more and more users begin to pay attention to the App rights related to personal security and privacy. At present, there are fewer recommendation systems combining App permissions, functions and user interests. Most recommendation systems only consider privilege privacy without considering user interest to make Top-N recommendation, or simply linearly combine App privilege privacy, function and user interest to make Top-N recommendation. The relationship between App permissions and user interests is isolated. This paper proposes a new method based on App permissions. The matrix decomposition algorithm which combines function and user's interest makes use of the correlation between the two to carry out App personalized recommendation. The main contributions of this paper are as follows: 1) combining with the experimental data of this paper, the App privilege analysis is carried out. This paper analyzes the number and type of permission required by ordinary App, and the situation of common App and malicious App.) it analyzes the relationship between the privacy of App permission and the score of users, and establishes the ARSM method to quantify the risk score of App by App-permission dichotomy. And preliminarily verify the relationship between privacy and user interest. 3) establish a App recommendation model with both privilege privacy and user interest. This paper presents a novel matrix decomposition algorithm based on privilege privacy and functional interest. It analyzes the weight and influence of the privilege in the App users' score by combining the App function attribute surface and the user's interest surface to realize the App recommendation. The relationship between privilege privacy and user rating is further analyzed, that is, the correlation between authority privacy and user interest. The experimental data in this paper is composed of the App of Anzhi Market and the corresponding user score. The experimental results show that, Compared with the traditional recommendation method, the recommendation algorithm proposed in this paper is more effective, and compared with the latest App recommendation system, which also uses authority and function, the accuracy of this algorithm is higher.
【学位授予单位】：武汉大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3;TP311.56

【参考文献】