An intelligent quantitative trading system based on intuitionistic-GRU fuzzy neural networks

Abstract

Quantitative trading based on intelligent algorithms has been a hot topic in the financial fields. However, the high noises and outliers increase the uncertainty and non-stationary of financial data such that intelligent algorithms often cease to be effective in the application scenario. Therefore, how to reduce the influence of uncertainty and extract the main trends of financial data is meaningful for trading. Motivated by it, in this article, a novel trading system based on intuitionistic fuzzy neural networks with gated recurrent unit (GRU) is proposed. Firstly, empirical mode decomposition (EMD) is applied to preprocess the original data and obtain the main trends of financial time series on the basis of smoothing data. Secondly, in order to tackle with uncertainty of financial data, a novel interval type-2 intuitionistic fuzzy system (IT2IFS) is proposed for the reasoning process, where hesitancydegree is involved for the learning processes In view of the strong dependence on the time dimension of series data, the gated recurrent unit is integrated into the reasoning model to strengthen the temporal connection. Thirdly, by using the above IT2IFS-GRU model, the quantitative trading system with a concise trading strategy is constructed. By using commodity futures and foreign exchange data, the superior trading results including net profits and precisions, etc. can be obtained, which verifies the effectiveness and availability of the proposed model.