Suspicious Transaction Recognition Based on KNN Model for Credit Card Fraud Prevention

In recent years, all kinds of illegal credit card activities disturb the financial order and destroy financial security. The problem of credit card fraud is one of the most important issues. This paper aims to use K Nearest Neighbors (KNN) model to predict credit card fraud under given factors. Specifically, the process is divided into the following steps. First, conduct Data pre-processing, second, use KNN models for dichotomous behavioral prediction, and finally, evaluate model accuracy. After performing the necessary pre-processing steps, such as repeat and gap testing, data standardization, outlier testing, and data dimensionality reduction using Principal Component Analysis (PCA), in this paper, the value of the nearest sample number k is discussed. According to the ACC value, the k value which makes the precision of the model optimal is selected. Finally, when k = 1 or 2, the accuracy of the model is about 0.9941. This experiment is a basic experiment, which can help readers to form a preliminary understanding of KNN algorithm and the field of credit card fraud detection, as well as lay a foundation for further study. Moreover, using KNN model to predict credit card fraud is of great significance to maintain financial security. It can help financial institutions more effectively detect and prevent fraud, and take timely measures to reduce losses. This helps to enhance customers’ trust in financial institutions and maintain good customer relations, thereby maintaining the stability and security of the entire financial system.