Chaotic Type-2 Transient-Fuzzy Deep Neuro-Oscillatory Network (CT2TFDNN) for Worldwide Financial Prediction

Abstract

Over the years, financial engineering ranging from the study of financial signals to the modeling of financial prediction is one of the most exciting topics for both academia and financial community. With the flourishing AI technology in the past 20 years, various hybrid intelligent financial prediction systems with the integration of neural networks, chaos theory, fuzzy logic, and genetic algorithms have been proposed. An interval type-2 fuzzy logic system (IT2FLS) with its remarkable capability for the modeling of highly uncertain events and attributes provides a perfect tool to interpret various financial phenomena and patterns. In this paper, the author proposes a chaotic type-2 transient-fuzzy deep neuro-oscillatory network with retrograde signaling (CT2TFDNN) for worldwide financial prediction. With the extension of author's original work on Lee oscillator—a chaotic discrete-time neural oscillator with profound transient-chaotic property—CT2TFDNN provides: effective modeling of an IT2FLS with a chaotic transient-fuzzy membership function; and effective time-series network training and prediction using a chaotic deep neuro-oscillatory network with retrograde signaling. CT2TFDNN not only provides a fast chaotic fuzzy-neuro deep learning and forecast solution, but also successfully resolves the massive data overtraining and deadlock problems, which are usually imposed by traditional recurrent neural networks using classical sigmoid-based activation functions. From the implementation perspective, CT2TFDNN is integrated with 2048 trading-day time-series financial data and top-10 major financial signals as fuzzy financial signals for the real-time prediction of 129 worldwide financial products that consists of: nine major cryptocurrencies, 84 worldwide forex, 19 major commodities, and 17 worldwide financial indices.