%0 Journal Article
%T 基于MogrifierLSTM的POI推荐算法研究<br>Research on POI Recommendation Algorithm Based on MogrifierLSTM
%A 李昊昱
%A 徐一波
%J Computer Science and Application
%P 384-395
%@ 2161-881X
%D 2024
%I Hans Publishing
%R 10.12677/CSA.2024.142039
%X <span style=\"font-size:14px;\">针对兴趣点推荐研究中忽略用户未来规划对下一POI的影响以及用户当前序列上下文信息学习不充分的问题，提出了一种基于MogrifierLSTM和未来信息的POI推荐算法——MOGREC。首先，根据用户签到行为的周期性学习用户历史行为偏好；其次，使用LSTM的变体MogrifierLSTM来对当前用户签到序列进行上下文学习；然后，利用注意力机制设计用户多步骤未来信息提取器；最后结合用户当前行为序列以及提取的多步骤未来信息来向用户推荐下一个POI。在Foursquare真实的签到数据集上的实验结果表明，MOGREC算法在长序列数据集SIN上的召回率和NDCG值比最优的CFPRec分别提高了16.17%和18.18%，验证了MOGREC算法的有效性。</span><br />
<span style=\"font-size:14px;\">In order to address the issue of neglecting the impact of user future planning on the next POI and insufficient learning of user current sequence context information in point of interest (POI) recommendation research, we propose a POI recommendation algorithm named MOGREC based on Mogri-fierLSTM and future information. Firstly, we learn user historical behavioral preferences based on the periodicity of user check-in behavior; Secondly, the MogrifierLSTM variant of LSTM is used to learn context information better on the current user check-in sequence; Then, a user multi-step fu-ture information extractor is designed using attention mechanism; Finally, based on the user’s current behavior sequence and the extracted multi-step future information, the next POI is recom-mended to the user. The experimental results on the real check-in dataset of Foursquare show that the MOGREC algorithm has s recall rate and NDCG increase of 16.17% and 18.18% compared to the optimal CFPRec on the long sequence dataset SIN, respectively, verifying the effectiveness of the MOGREC algorithm.</span>
%K POI推荐，未来信息，MogrifierLSTM，多任务，Transformer<br>POI Recommendations
%K Future Information
%K MogrifierLSTM
%K Multi-Task
%K Transformer
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=81947