An Improved Support-Vector Network Model for Anti-Money Laundering

Abstract

The selection of parameters of SVM model will affect the identification effect of suspicious financial transactions, this paper proposes the cross validation method to find the optimal SVM classifier parameters to solve this problem. Cross validation method finds the optimal parameters based on the highest classification accuracy rate through grid search, it can effectively avoid the state of over-learning and less learning, and greatly improves the overall performance of the classifier.