An incremental type-2 fuzzy classifier for stock trend prediction

Abstract

Due to its complex and unstable nature, stock market prediction is one of the most challenging problems. Lots of research has tried to explore the possibility of predicting stock trends using numerous machine learning methods. However, little effort has been made to predict the future trend online. In this paper, we propose an incremental type-2 fuzzy rule-base classifier capable of considering a high amount of uncertainty in stock data in addition to self-adaptation in an online manner. In this approach, interval type-2 fuzzy sets are built by combining type-1 sets that are directly extracted from the data. The proposed classifier can dynamically modify its parameters in order to react to the alteration of concepts without re-training the model. The efficiency of the proposed method is tested on some market indexes such as SP500, TAIEX, etc. and compared to various incremental methods. Moreover, to show the effectiveness of the proposed technique even in offline mode, we have compared it against some outstanding offline approaches, including deep neural networks. The obtained results demonstrate that our proposed method performs well under different settings.