Implementation and Evaluation of Node Construction Criterion for Decision Tree Algorithm in Inhomogeneous Distribution Data Stream

Abstract

Recently, because of increasing amount of data in the society, data stream mining targeting large scale data has attracted attention. The data mining is a technology of discovery new knowledge and patterns from the massive amounts of data, and what the data correspond to data stream is data stream mining. In this paper, we regard credit card transaction data as data stream and detect fraud use to apply decision tree algorithm to data stream mining. In applying decision tree algorithm, we propose new evaluative criterion for decision tree. We apply VFDT (Very Fast Decision Tree learner) algorithm to online decision tree construction. At first, we construct VFDT using those data and create detection of fraud use model. At second, we verify the classification accuracy of the model.