The current fraud risk in digital finance is increasing year by year, and the mainstream solutions rely on the inherent characteristics of users, which makes it difficult to explain fraud behaviors and fraud behavior patterns are less researched. To address these problems, we propose an integrated multiple relational graphs fraud detection model Tri-RGCN-XGBoost, which analyzes the impact of user association patterns on fraud detection by mining the behavioral associations of users. The model builds a heterogeneous information network based on real transaction data, abstracts three types of bipartite graphs (user–device, user–merchant, and user–address), aggregates the information of the user’s neighbor nodes under the three types of behavioral patterns, and integrates the graph convolution classification results under the three behavioral patterns with the XGBoost model to achieve fraudulent user detection with integrated multiple relational graphs. The results show that the performance of this model in fraud identification is significantly improved, especially in reducing the fraudulent user underreporting rate. Further, the behavioral associations that play a key role in fraud user identification are analyzed in conjunction with shape value to provide a reference for fraud pattern mining.

In the process of user churn modeling, due to the imbalance between lost users and retained users, the use of traditional classification models often cannot accurately and comprehensively identify users with churn tendency. To address this issue, it is not sufficient to simply increase the misclassification cost of minority class samples in cost-sensitive methods. This paper proposes using the Focal Loss hard example mining technique to add the class weight α and the focus parameter γ to the cross-entropy loss function of LightGBM. In addition, it emphasizes the identification of customers at risk of churning and raises the cost of misclassification for minority and difficult-to-classify samples. On the basis of the preceding ideas, the FocalLoss_LightGBM model is proposed, along with random forests, SVM, XGBoost, and LightGBM. Empirical analysis based on a dataset of credit card users publicly available on the Kaggle website. The AUC, TPR, and G-mean index values were superior to the existing model, which can effectively improve the accuracy and stability of potential lost users.