基于人员和项目多关系表达的非对称二部图模型在电力项目人员角色推荐中的应用

Application of Asymmetric Bipartite Graph Model Based on Multi Relationship Embeding Between Personnel and Works in Role Recommendation of Power Project

在电力作业项目的主要参与人员选派中由于人员属性和项目属性的多样性,而且要求一个作业项目的各类主要角色只有一人,同时参与项目的主要人员中有“负责人”“签发人”“许可人”这3种角色。对于这样的应用场景,如果有效地选择最合适的人员和项目匹配是很困难的。利用经典的推荐模型,如“双塔模型”、“Pinsage模型”等均存在明显的不足,本文提出了融合多关系预测和Pinsage无监督hingeloss的双目标图神经网络模型,较好解决了实际应用场景的人员角色推荐问题。

Due to the diversity of personnel attributes and project attributes in the selection and assignment of main participants in electric power operation projects,it is required that there is only one main role in each operation project.Meanwhile,there are three roles of"principal","issuer"and"Licensor"among the main participants in the project.For such an application scenario,it is difficult to effectively select the most appropriate personnel and project matching.Using classical recommendation models,such as"double tower model"and"pinsage model",there are obvious shortcomings.Our research group proposes a dual objective graph neural network model integrating multi relationship prediction and pinsage unsupervised hinge loss,which better solves the problem of personnel color recommendation in practical application scenarios.