Dandelion Optimization Based Feature Selection with Machine Learning for Digital Transaction Fraud Detection

Digital transactions relying on credit cards are gradually improving in recent days due to their convenience. Due to the tremendous growth of e-services (e.g., mobile payments, e-commerce, and e-finance) and the promotion of credit cards, fraudulent transaction counts are rapidly increasing. Machine learning (ML) is crucial in investigating customer data for detecting and preventing fraud. Conversely, the advent of irrelevant and redundant features in most real-time credit card details reduces the execution of ML techniques. The feature selection (FS) approach's purpose is to detect the most prominent attributes required for developing an effective ML approach, making sure that the classification and computational complexity are improved and decreased, respectively. Therefore, this study presents an evolutionary computing with fuzzy autoencoder based data analytics for credit card fraud detection (ECFAE-CCFD) technique. The purpose of the ECFAE-CCFD technique is to recognize the presence of credit card fraud (CCF) in real time. To achieve this, the ECFAE-CCFD technique performs data normalization in the earlier stage. For selecting features, the ECFAE-CCFD technique applies the dandelion optimization-based feature selection (DO-FS) technique. Moreover, the fuzzy autoencoder (FAE) approach can be exploited for the recognition and classification of CCF. FAE is a category of artificial neural network (ANN) designed for unsupervised learning that leverages fuzzy logic (FL) principles to enhance the representation and reconstruction of input data. An improved billiard optimization algorithm (IBOA) could be implemented for the optimum selection of the parameters based on the FAE algorithm to improve the classification performance. The simulation outcomes of the ECFAE-CCFD algorithm are examined on the benchmark open-access database. The values display the excellent performance of the ECFAE-CCFD method with respect to various measures.