融合时空-社交-顺序影响的多维兴趣点推荐  

Multi-dimensional point of interest recommendations incorporating spatio-temporal,social and sequential influences

作　　者：张稳 伊华伟[1] 兰洁 薛莹莹 ZHANG Wen;YI Hua-wei;LAN Jie;XUE Ying-ying(College of Electronics and Information Engineering,Liaoning University of Technology,Jinzhou 121001,China)

机构地区：[1]辽宁工业大学电子与信息工程学院,辽宁锦州121001

出　　处：《计算机工程与设计》2024年第9期2704-2711,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(62203201);辽宁省教育厅基础科研基金项目(LJKZZ20220085);营口市企业博士双创计划基金项目(2022-13)。

摘　　要：现有的兴趣点推荐方法忽略了不同上下文因素之间的内在联系,导致上下文因素未能得到充分利用,为此提出一种融合时空-社交-顺序影响的多维兴趣点推荐算法。根据用户的时间状态和活动轨迹刻画用户的活动区域,探索用户的时间偏好程度和活动轨迹相似度;利用高斯分布模型评估用户的地理偏好程度,使用马尔科夫链算法预测用户访问下一个兴趣点的概率。实验结果表明,该算法优于其它算法。The existing point of interest recommendation methods ignore the internal relationship between different context factors,which leads to the failure to make full use of context factors.Therefore,a multi-dimensional point of interest recommendation algorithm integrating spatio-temporal,social and sequential effects was proposed.The user’s activity area was described according to the user’s time state and activity trajectory,and the user’s time preference degree and activity trajectory similarity degree were explored.The Gaussian distribution model was used to evaluate the user’s geographical preference degree.The improved Markov chain algorithm was used to predict the user’s probability of visiting the next point of interest.Experimental results show that this algorithm is superior to other algorithms.

关 键 词：社交网络 兴趣点推荐 时空 社交 顺序影响 活动轨迹 多维 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]