Abstract
Fraudulent activities are an expensive problem for many financialinstitutions, costing billions of dollars to corporations annually. Morecommonly occurring activities in this regard are credit card frauds. In thiscontext, the credit card fraud detection concept has been developed over thelines of incorporating the uncertainty in our prediction system to ensurebetter judgment in such a crucial task. We propose to use a sparse Gaussianclassification method to work with the large data-set and use the concept ofpseudo or inducing inputs. We perform the same with different sets of kernelsand the different number of inducing data points to show the best accuracy wasobtained with the selection of RBF kernel with a higher number of inducingpoints. Our approach was able to work over large financial data given thestochastic nature of our method employed and also good test accuracy with lowvariance over the prediction suggesting confidence and robustness in our model.Using the methodologies of Bayesian learning techniques with the incorporatedinducing points phenomenon, are successfully able to obtain a healthy accuracyand a high confidence score.