基于知识图谱的银行反欺诈模型的研究与应用

Research and Application of Bank Anti-fraud Model Based on Knowledge Graph

针对传统的银行反欺诈模型已无法满足欺诈检测要求的及时性与准确性的问题,本文提出了一种基于知识图谱的反欺诈模型。该模型基于多源信息和高维衍生特征的大数据,构建知识图谱,对信贷个体进行全方位画像,分析关联关系,并抽取网络属性。从四大方面、两大维度挖掘风险特征,四大方面指个人基本信息、账户信息、征信和行为信息,两大维度指个人节点和网络结构。最后将风险特征代入LightGBM,判断是否为欺诈类型,并得到对应概率。实验表明,相比于仅使用个人自身特征的模型,使用个人特征加网络特征的模型效果更好,AUC和F1分数分别提升5.18%和5.71%。因此,该方案能够有效地为银行对个人信贷进行欺诈评估。

In response to the issue that traditional bank anti-fraud models no longer meet the timeliness and accuracy requirements for fraud detection,this article proposes a knowledge graph-based anti-fraud model.This model uses big data based on multi-source information and high-dimensional derived features to build a knowledge graph,conduct a comprehensive profiling of credit individuals,analyze the associated relationships,and extract network attributes.Then,risk features are mined from four aspects and two dimensions.The four aspects refer to personal basic information,account information,credit history,and behavioral information,while the two dimensions refer to personal nodes and network structure.Finally,the risk features are substituted into LightGBM to determine whether it is a fraudulent type and obtain the corresponding probability.Experiments show that,compared to models using only personal characteristics,models using both personal and network characteristics perform better,with AUC and F1 scores increasing by 5.18%and 5.71%respectively.Therefore,this solution can effectively provide fraud assessment for banks’personal credit.