高级检索
    钱忠胜, 杨家秀, 李端明, 叶祖铼. 结合用户长短期兴趣与事件影响力的事件推荐策略[J]. 计算机研究与发展, 2022, 59(12): 2803-2815. DOI: 10.7544/issn1000-1239.20210693
    引用本文: 钱忠胜, 杨家秀, 李端明, 叶祖铼. 结合用户长短期兴趣与事件影响力的事件推荐策略[J]. 计算机研究与发展, 2022, 59(12): 2803-2815. DOI: 10.7544/issn1000-1239.20210693
    Qian Zhongsheng, Yang Jiaxiu, Li Duanming, Ye Zulai. Event Recommendation Strategy Combining User Long-Short Term Interest and vent Influence[J]. Journal of Computer Research and Development, 2022, 59(12): 2803-2815. DOI: 10.7544/issn1000-1239.20210693
    Citation: Qian Zhongsheng, Yang Jiaxiu, Li Duanming, Ye Zulai. Event Recommendation Strategy Combining User Long-Short Term Interest and vent Influence[J]. Journal of Computer Research and Development, 2022, 59(12): 2803-2815. DOI: 10.7544/issn1000-1239.20210693

    结合用户长短期兴趣与事件影响力的事件推荐策略

    Event Recommendation Strategy Combining User Long-Short Term Interest and vent Influence

    • 摘要: 事件社交网络的快速发展引起的信息过载问题是当前面临的主要挑战,深度学习等技术可从大量的数据中挖掘潜在的关联信息,从而有效应对该问题.同时,有研究表明用户兴趣在长期和短期的时序上具有不同的特征模式,深度挖掘用户的时序特征和兴趣可有效地为用户提供个性化的事件推荐信息.基于此,提出一种将用户长短期兴趣与事件影响力相结合的推荐策略.通过带注意力机制的图神经网络和长短期记忆网络获取用户的长短期兴趣,同时,对候选事件构建针对目标用户的影响力.根据用户长短期兴趣和事件影响力预测目标用户的参与概率,最终通过排序后的参与概率向用户推荐TOP-K兴趣事件.实验结果表明,所提推荐模型在多个指标上均有所改善,其推荐性能优于已有对比模型,具备很好的推荐效果.

       

      Abstract: The problem of information overload caused by the rapid development of event-based social network is the main challenge currently. Those technologies such as deep learning can mine potential relationship from a large amount of data to effectively cope with this problem. At the same time, some studies show that user interests have different characteristic patterns in long-term time series and short-term time series. In-depth mining of users’ time series characteristics and interests can effectively provide users with personalized event recommendation information. Based on this, a recommendation strategy is proposed, which combines long-short term user interests with event influences. It obtains the long-short term user interests by graph neural network and long-short term memory network together with attention mechanism. At the same time, it constructs the influence on the target user for candidate events, and predicts the participation probability of the target user based on the long-short term interests and event influences. Finally, TOP-K events are recommended to the users according to the sorted participation probability. The experimental results show that the proposed recommendation model has been improved on multiple indicators. Its recommendation performance is better than the several existing compared models, and it has a good recommendation effect.

       

    /

    返回文章
    返回