Credit Data Fraud Detection using Kernel Methods with Support Vector Machine
Maryamsadat Hejazi, Yashwant Prasad Singh · 2012
This paper presents binary Support Vector Machine classifiers for detection of fraudulent credit card transactions. Support Vector Machine (SVM) is a well-known method in statistical learning theory for classification and regression problems. The objective of this paper is about evaluation of accuracy and performance of different kernel methods using Sequential Minimal Optimization (SMO), C-Support Vector Classification (C-SVC) and υ-Support Vector Classification (υ-SVC). A Comparison of computation time and classifiers’ accuracy for SMO classifier with C-SVC in LIBSVM is provided. The simulation results show that accuracy of SMO classifier in with different kernel functions is higher than C-SVC in LIBSVM on credit data dataset obtained from UCI machine learning database.