Options volatility modelling and trading system incorporating EFSM

• Main Author: Chan, Andy Jia Wei.
• Other Authors: Quek Hiok Chai
• Format: Final Year Project
• Language: English
• Published: 2012
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/50117
• Institution: Nanyang Technological University

It is well acknowledged that financial volatility implies financial risk. Therefore, an accurate prediction of financial volatility is of critical significance. A substantial part of professional option trading focuses strictly on volatility and ignores the direction of the underlying market. Part of the attractiveness of volatility trading may be explained by the well documented fact that forecasting directional changes in the underlying asset is very difficult, whereas volatility clustering, whereby periods of high volatility and low volatility tend to cluster together, tend to have mean reverting tendencies.The goal of non-parametric pricing models is to price and risk-manage financial derivatives in a model-free approach, thus addressing the limitations of traditional models which employ various assumptions to obtain closed form solutions Non-parametric pricing methods rely on available data to detect non-linear patterns and relationships between inputs to determine asset dynamics and pricing processes. As such, we will attempt to apply eFSM to model and forecast the volatility of the Dow Jones-UBS Commodity Index for wheat and corn. The evolving fuzzy semantic memory model(efSM), is a self-organizing neural-fuzzy semantic model that evolves dynamically with the input of each individual set of training data. The eFSM adds new fuzzy rules dynamically upon their interpretation on data arrival. Old rules that are not longer descriptive of the current data(beyond a certain threshold) are also pruned. This ensures that the model accurately models the non-stationary state of data coming in, and is in line with the information processing capabilities of the human brain. In addition, the fuzzy rules allow for easy interpretation and tractability which are desirable features in a forecasting system. This overcomes some of the existing problems plaguing neural network systems, and lend credence to the idea of applying soft computing techniques to financial data. Results obtained from the experiments outlined in this paper prove to be promising. This research should prove useful for options writers and buyers looking to derive an accurate representation of the fair value of an option/derivative in today’s markets.