Bank abnormal transaction data processing method based on K-means++ algorithm

In the face of massive banking transaction data, this paper addresses the challenge faced by relevant authorities in identifying abnormal transactions. By analyzing the banking transaction data from an illegal tobacco trading case, the paper proposes an improved K-means++ algorithm to cluster various characteristic data in banking transactions. By analyzing the clustering results, the method can identify and filter out data unrelated to abnormal transactions, thereby improving the efficiency of authorities in detecting anomalous banking behaviors. The algorithm first forms initial clusters, calculates the distance ratio of elements within each cluster, and then determines whether a cluster should split by setting a threshold. Subsequently, the next cluster center is selected with a certain probability. Finally, the algorithm filters out data unrelated to abnormal transactions from the banking transaction data, increasing the proportion of relevant data. Experiments have demonstrated that the algorithm proposed in this paper can filter out records of non-anomalous transactions in bank transaction data. Compared to other common clustering algorithms, the algorithm proposed in this paper shows a significant advantage in terms of identification accuracy.