The ACDF Algorithm in the Stream Data Analysis for the Bank Telemarketing Campaign
Jan T. Kozák, Przemysław Juszczuk · 2018
Ant Colony Decision Forrest (ACDF) and other ensemble methods have been proved to be effective in the case of the various datasets. In this article, we propose a concept, of transforming the classical ACDF algorithm in such matter, that the every newly generated decision tree becoming the element of the ensemble of classifiers is generated after the new data packages are derived to the system. Such approach allows to adapt to the still-changing data present in the system and may allow overcoming difficulties related to the problem of concept drift which is a commonly present problem in the real financial data. We investigate this problem and present a new algorithm based on the original ACDF adapted to the stream data.The proposed concept is verified experimentally on two different approaches with a different number of data packages. Statistical verification of the proposed method is presented as well.