Certain Applications of Neural Networks to Finance and High Frequency Trading

Alexander Shklyarevsky 

 

Abstract:  Big Data, Machine Learning including Deep Learning / Neural Networks and other Artificial Intelligence (AI) applications are becoming more and more important in Financial Industry, as well as in other Industries – Insurance, Direct Marketing, Tech, Medical, Biochemical, Transportation, Construction, Aerospace, Commodities, etc. AI as a major direction of current STEM (Science, Technology, Engineering and Mathematics) revolution took off to a new level thanks to an improved ability of computer systems to process Big Data (billions and trillions of Big Data repository records vs. thousands and millions of database records before) and other advances in technology including a much greater speed and RPA (Robotic Process Automation), as well as improved mathematical, statistical, software and hardware foundations of Data Science and Machine Learning / Deep Learning. Our presentation is focusing on certain applications of Deep Learning / Neural Networks to Finance and High Frequency Trading. We will organize our presentation as follows: • Firstly, we will show how a particular Neural Networks construction process could be implemented to be applied to certain areas of Finance and High Frequency Trading. Specifically, the construction process – input data, its structure and frequency, input data mapping process, processing component, output data, its structure and frequency and output data mapping process – will be discussed for the following Neural Network models: o Deep Learning / Neural Network Volatility Models across asset classes o Deep Learning / Neural Network Correlation Models across asset classes • Secondly, we will describe methodology components for Neural Network Volatility Models and Neural Network Correlation Models for the following asset classes with specifics by asset class: o Equities and Equity Indices o Foreign Exchange (FX) Spot Rates o Commodities and Commodity Indices o Fixed Income Instruments and Fixed Income Indices • Thirdly, we will conduct a comparative analysis between Neural Network Volatility Models and Neural Network Correlation Models vs. other Volatility and Correlation Models (for example, Stochastic Volatility Models and Stochastic Correlation Models) with pros and cons of both approaches, as well as their cross-fertilization • Fourthly, we will show how to apply Neural Network Volatility Models and Neural Network Correlation Models to High Frequency Trading for the 4 asset classes where we described methodology components for above • And, finally, we will conclude our presentation by showing the perspectives of further development of Neural Networks - mathematical and statistical theory, mathematical and statistical modeling and its implementation - and their applications to Finance and High Frequency Trading