面向指挥员关注的聚类推荐模型

Clustering Recommendation Model for Commander’s Concerns

信息技术快速发展的条件下海量军事信息的大量涌现,人工查阅情报的传统方法已经不再适用,而人工制定或自动挖掘的关联规则难以捕获指挥员的关注方向,从而限制了大规模军事信息处理的性能。为了对不同关注点的指挥员提供个性化的信息推荐,文章提出了一种面向指挥员关注的基于强化学习的聚类推荐模型,通过强化学习从特定指挥员近期关注内容中学习到其行为模式,并使用聚类模型将尚未查看的信息推送到对应的指挥员,实现了从无差别信息共享到考虑角色关注的个性化关联目标。在真实世界的新闻数据集上验证表明,该模型针对具体指挥员进行新闻推荐的准确率较高。

With the development of information technology and the digitization of massive military information,the traditional method of manually reviewing intelligence is no longer applicable,while manually formulated or automatically mined association rules are difficult to capture the concerns of commanders,thus limiting the performance of large-scale military information processing.In order to personalize information recommendation for commanders with different concerns,a reinforcement learning-based clustering recommendation model for commanders’concerns is proposed,which learns the behavior patterns from the recent concerns of specific commanders through reinforcement learning and uses the clustering model to recommend the unviewed information to the corresponding commander,achieving the goal of personalized association from undifferentiated information sharing to considering role concerns.Validation on a real-world news dataset shows that the model is highly accurate in targeting news recommendations to specific commanders.