Optimizing Support Vector Machine with Polynomial Kernel for Credit Card Fraud Detection
Loading...
Date
2023-10-01
Journal Title
Journal ISSN
Volume Title
Publisher
FUOYE Journal of Pure and Applied Sciences
Abstract
The rate at which fraud relating to credit cards is increasing is alarming and requires accurate detection to curb its regular occurrence. Discovery of such fraud, especially in the financial world is still a challenge despite several research using various machine learning algorithms. Though SVM has proved its efficiency in producing good detection parameters, there is still a need to improve the accuracy for optimal results. Hence, a Polynomial Kernel optimization method of Support Vector Machine (SVM) was adopted to identify Credit Card Fraud. The data set used for this work was acquired from Kaggle and it presents transactions that occurred in two days with 492 frauds out of 284,807 transactions. The dataset was trained, tested and a parameter tuning using a polynomial kernel was performed, after which the performance of the model was evaluated. SVM was optimized with Polynomial kernel, Sigmoid kernel and Radial Basis Function (RBF) kernel and it was realized that sigmoid gave an accuracy of 76%, RBF gave an accuracy of 94% while Polynomial gave an accuracy of 100%. Hence, the result proved that SVM optimized with a Polynomial kernel outperformed other optimization techniques for detecting credit card fraud.
Description
Keywords
Citation
5. Odeniyi, L. A., Oduntan, E. O., Balogun, M. O., Ogunrinde, M. A., & Omoniyi, V. (2023). Optimizing Support Vector Machine with Polynomial Kernel for Credit Card Fraud Detection. FUOYE Journal of Pure and Applied Sciences (FJPAS), 8(3), 161-170