Abnormal Transaction Detection and Fraud Prediction of Bank Accounts based on Big Data Association Analysis Algorithm

In order to solve the problem of abnormal transactions in bank accounts, the graph convolutional network (GCN) algorithm is introduced. The historical records in bank transaction data are used to construct a transaction behavior graph and associate elements such as accounts, transactions, merchants, and payment methods through a graph model. Then, the transaction graph is analyzed through deep learning by GCN to construct a multi-dimensional feature vector of transaction behavior. The extracted features are trained based on the anomaly detection algorithms Isolation Forest and One-Class SVM to identify the boundaries between normal and abnormal transactions. GCN combined with incremental learning performs best, with an accuracy of 91.00% and a recall of 85.80%, which proves the adaptability and flexibility of the model in processing dynamic data. The algorithm based on big data association analysis effectively improves the accuracy and real-time performance of abnormal transaction detection and fraud prediction in bank accounts.