Hybridization Preprocessing and Resampling Technique-Based Neural Network Approach for Credit Card Fraud Detection
Bright Keswani, Poonam Keswani, Prity Vijay, Ambarish Gajendra Mohapatra · Apple Academic Press eBooks · 2022
Credit card fraud is a form of financial fraud growing every year, causing losses to financial multinationals as well as government sectors. Traditional methods like manual detection of credit card frauds are feasible only for small datasets, but with the rise of big data, these methods are of no worth. Data mining and machine learning (ML) techniques have been widely used for fraud detection, by drawing a pattern that separates two classes (fraud and legitimate). The only necessity of ML algorithms is vast sets of examples, but messy and complex datasets are tricky for ML techniques and thus bring challenges during formulating learning rules. Almost all real-world dataset is unclean which tends to produce inaccurate models. This chapter revealed various challenges possessed by ML classification 98 algorithms because of complex imbalance dataset and therefore proposes a new hybridization preprocessing and resampling technique (HPRT), which solves two major issues: (a) cleans the dataset while (b) balancing the dataset. HPRT thus enhances the performance of ML algorithms. Several ML algorithms were observed to produce better result when combined with HPRT. HPRT-based neural network model is constructed for detection of credit card fraud and is compared with traditional neural network model. Confusion matrix and other matrices revealed moderately high and precise results for HPRT-based neural network-based model in comparison to traditional neural network model.