The advent of the digital economy and Industry 4.0 enables financial organizations to adapt their processes and mitigate the risks and losses associated with the fraud. Machine learning algorithms facilitate effective predictive models for fraud detection for Industry 4.0. This study aims to identify an efficient and stable model for fraud detection platforms to be adapted for Industry 4.0. By leveraging a real credit card transaction dataset, this study proposes and compares five different learning models: logistic regression, decision tree, k-nearest neighbors, random forest, and autoencoder. Results show that random forest and logistic regression outperform the other algorithms. Besides, the undersampling method and feature reduction using principal component analysis could enhance the results of the proposed models. The outcomes of the studies positively ascertain the effectiveness of using features selection and sampling methods for tackling business problems in the new age of digital economy and industrial 4.0 to detect fraudulent activities.
Bibliographical noteFunding Information:
This work is partly supported by VC Research (VCR 0000158) for Prof Chang.