An optimal approach for fraud detection by comparing random forest algorithm and support vector machine algorithm for credit card transaction with improved accuracy

• Main Authors: Kumar, K. Yashwanth, Vani, B.
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Datasets; Fraud; Machine learning; Performance enhancement; Support vector machines
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0177045

Present investigation is targeted toward detection of credit card fake transactions. Random forest algorithm is used for innovative fraud detection in credit cards and its performance is tested by comparing with sup port vector machine algorithm. For present study two datasets were created one is training dataset [n=2,27,845 (80%)] and a test dataset [n=56,962 (20%)] (0.8g power). In the dataset 492 fraud transactions are present out of 284,807 transactions. Random forest classifier is compared with various machine learning algorithms using metrics like specificity, F-score, and accuracy. The detection accuracies of random forest algorithms and support vector machine are 97.2%, and 93.27% respectively. Findings of present study shows that the proposed random forest algorithm gives significantly enhanced performance as compared to the support vector machine algorithms in innovative fraud detection for fraudulent transactions in credit cards.