Algorithm for Identifying Abnormal User Arrears Based on Composite Neural Networks

The identification of abnormal user arrears usually relies on a large amount of user data, including consumption records, payment history, credit ratings, etc. However, these data may be missing, incorrect, or inconsistent, resulting in inaccurate identification of abnormal user arrears. In addition, the real-time nature and update frequency of data can also affect the efficiency of debt recognition. To this end, a combined neural network-based algorithm for identifying abnormal user arrears is proposed. Using the feature extraction layer, deconvolution layer, and Softmax layer in deep neural network models to extract abnormal user features. Convert the extracted abnormal user features into graph data, and use a graph neural network model to detect abnormal user arrears behavior. On this basis, a feedforward neural network is used to construct a model for identifying abnormal user arrears. Deep neural networks, graph neural networks, and feedforward neural networks are used as a combination neural network to achieve abnormal user arrears recognition. The experimental results show that the average MAPE and MAE values of the proposed method are 6.2 and 2.9, respectively. The recognition time for abnormal user arrears is only 47ms, which can effectively improve the accuracy and efficiency of abnormal user arrears recognition.