Credit Card Fraud Detection Using Online Boosting with Extremely Fast Decision Tree
Aye Aye Khine, Hint Wint Khin · 2020
Nowadays, data stream mining is a very hot and high attention research field due to the real-time industrial applications from different sources are generating amount of data continuously as the streaming style. To process these growing and large data streams, data stream mining, classification algorithms have been proposed. These algorithms have to deal with high processing time and memory costs, class imbalance, overfitting and concept drift and so on. It is sure that ensembles of classifiers are being effectively used to make improvement in the accuracy of single classifiers in either data mining or data stream mining. Thus, to get higher performance in prediction with largely no increasing memory and time costs, this paper proposes an Online Boosting(OLBoost) Approach, which is firstly use the Extremely Fast Decision Tree (EFDT) as base (weak) learner, in order to ensemble them into a single online strong learner. The experiments of the proposed method were carried out for credit card fraud detection domain with the sample benchmark datasets.