基于知识图谱的用户特征-关系推荐模型在电力安全教育中的应用

The Application of Knowledge Graph-based User Feature-relation Recommendation Model in Power Safety Education

电力安全教育中员工群体分布的长尾特征,导致采样到的用户交互数据具有明显的稀疏性,使用传统推荐算法无法获得理想的效果,为此文章提出了基于知识图谱的用户特征-关系推荐模型。建立一个基于多任务迁移学习的神经网络,通过引入F-R单元实现了用户特征与实体关系的深度聚合,能识别出对推荐有重要影响的“用户特征-实体关系”组合,并通过神经网络挖掘其内在作用规律,从而利用实体关系信息强化了用户特征对推荐算法的影响,显著提升了模型的预测性能。实验证明,本模型能够有效解决电力安全教育场景中‘长尾’群体用户交互数据的稀疏问题,明显缓解冷启动效应。

In the context of power safety education,the long-tail distribution of the employee population leads to severe sparsity in the sampled user interaction data,having a negative impact on the performance of traditional recommendation algorithms.This paper proposes a User Feature-Relation Recommendation model based on knowledge graph.It establishes a multi-task transfer-learning neural network.By introducing the F-R unit,the model can identify crucial user feature-entity relation combinations and mine the inherent rules.By utilizing entity relationship information,it significantly improves the performance of the model.Experiments have shown that this model can effectively address the sparsity issue of user interaction data for the'long-tail'group in power safety education scenarios,significantly alleviating the cold start effect.