Fraud Identification in Financial Transactions: Machine Learning-Based Anomaly Detection Method
Sweta Sharma, Netala Kavitha, Navdeep Dhaliwal, Balaji Rajalakshmi, Ippa Sumalatha, R. Kavitha · 2024
Two major financial crimes that have reached epidemic proportions worldwide are credit card holder fraud and financial statement fraud. The efficacy of corporate governance, the calibre of financial reports, and the legitimacy of audit functions have all suffered from the collapse of well-known corporations. Credit card theft and financial statement fraud have grown to be serious problems for organizations worldwide. The predicted accuracy scores of Support Vector Classifiers (SVM) and Logistic Regression were found to be greater than those of the other classifiers employed in the unbalanced training dataset. An experimental model that first applies the class balancing technique to transform the unbalanced dataset into a balanced data set was developed in alignment with the research purpose and to design a model to improve results and reduce False Negative Rate. Using a nonlinear support vector machine model, a classifier was trained on the dataset. The AUC score of the suggested model was higher. By creating a new, selected hybrid class balancing technique and using the Random Forester algorithm to create an optimum classifier and analyse the outcomes, the experimental model of class balancing was further refined. Based on the results of this experiment, it was found that the suggested approach outperforms the prior model, generates a good AUC score, and maximizes false negatives. The approach can be used to address other comparable categorization issues for better results. It is not just restricted to online financial transactions; it also aids in lowering false negatives in credit card fraud transactions.