Type 2 fuzzy neural network for strangle-based option trading

• Main Author: Seah, Kevin Zhixiang.
• Other Authors: Quek Hiok Chai
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/55140
• Institution: Nanyang Technological University

Many existing neural fuzzy systems are capable of self-learning and adapt their initial structure as well as their parameters. These systems are also known as evolving neural fuzzy systems. One such example is the evolving Type-2 fuzzy inference system, eT2FIS. It is common that these evolving systems have a fixed set of thresholds for their learning mechanism. This assumes that the environment that they are tasked to model changes within a certain known boundaries. However, this is not true for many real-life application such as financial time-series. In different periods, price of a security can fluctuates in different magnitude, which requires great adaptability of the modelling system in order to produce sound prediction. Without a set of dynamic learning thresholds, these models is unable to optimally match the rate of learning to the various rate and degree of change in the environment. In this paper, a dynamic evolving neural-fuzzy system eT2FIS++, extended from the original eT2FIS model is proposed. The proposed model uses a simple rate of change measure, inspired by a commonly used momentum technical indicator, to match the model learning mechanism to the observed environment rate of change. eT2FIS++ also possesses an additionally recurrent neuron that enables it to store information on error signal produced on previous prediction. This locally recurrent feature at the output layer of the model allows it to convergence more quickly by factoring in a faction of previously made prediction error in the current back-propagation process. Encouraging results were produced in experiments that aim to test the performance of the proposed structure of eT2FIS++ in the modelling of chaotic time-series and financial time-series. Finally, an intelligent strangle trading system that consists of volatility projection module (VPM), trend identification module (TIM) and a trade decision module (TDM) is proposed for financial volatility trading. The proposed trading system employs eT2FIS++ as its financial time-series prediction model involve in the VPM and TIM before generating trade signals in the TDM with the help of some well-established financial technical indicators. The returns of the proposed strangle trading system observed in the trading simulation, when benchmarked against an existing eFSM-based straddle trading system, are encouraging.