基于PMR架构的兴趣点推荐研究

Study of POI-s recommendation based on a PMR framework

基于位置的移动社交网络,如微博、微信等为移动社交网络的研究提供了大量用户历史行为信息,诸如用户资料、签到时间、位置轨迹、微博内容等.它们为基于情境的用户实时兴趣点推荐的研究提供了契机.本文通过借鉴PMJ模型的思想,建立了一种全新的位置推荐系统原型PMR架构,它最大的优势在于,通过把最新的情境感知技术和计算方法与PMJ架构的优势相结合,能够对用户反馈信息实时处理后依据该信息对推荐方案进行分值和排序更新,因此这种以用户需求为核心的推荐方案更加契合用户需求.在此架构的基础上,提出了一整套用于情境感知、存储、推荐和反馈的计算方法,较好地解决了情境计算研究领域中基于用户反馈机制的自适应个性化推荐这一关键问题.实验结果表明,该系统架构具有较高的召回率(96.93%)和准确率(87.27%),以及较好的用户满意度(80.12%).

Location-based mobile social networks, such as micro-blogs, web-chats, etc. have provided mass his- torical user behavior information such as user profiles, check-in times, position tracking, and tweets, for research on mobile social network The availability of these data received from users offers a good opportunity to study the users＇ real-time POI-s recommendation based on context. In this paper, a new framework for location recommendation called PMR is enhanced by drawing from the idea of a PMJ model. One of the greatest strengths of the prototype is its ability to process users＇ feedback information in real time and update recommendations by integrting the latest methods of context awareness and the strength of a PMJ framework into a PMR framework. Additionally, it can re-score the recommended scheme and update the rank of the scheme so that it fits users＇ demands. Based on this, a set of algorithms are proposed to solve the key problems in the research of context-aware computing： context awareness, context storage, context recommendation, and context feedback. Experimental results show that the model performs with a higher recall rate （96.93%） and precision rate （87.27%） and a better customer satisfaction index （80.12%）.