Credit card fraud detection using predictive features and machine learning algorithms

CCF can destabilise economies, reduce confidence between customers and banks, and severely affect other people and businesses. The primary objective of banks and businesses is to identify fraudulent transactions with a high level of accuracy and to also reduce false alerts and the costs of manual investigation activities. When identifying CCF in large datasets, feature selection is very important to improve accuracy performance and rapid detection of fraud. One of the most widely used methods of feature selection is the random forest classifier (RFC), which is well suited for large datasets. The RFC works well; it tends to identify more predictive features, which can significantly improve the classification performance for a CCF detection model. In this paper, we suggest a CCF detection method based on feature selection using random forest classifier and machine learning algorithms such as support vector machines (SVM), isolation forest (IF) to detect fraudulent transactions. The proposed model is applied to a large real-world dataset to study the accuracy of its fraud detection performance. A comparison is made between the proposed model and other machine learning methods.