Predictive Learning Methods to Price European Options Using Ensemble Model and Multi-asset Data

Abstract

Option contracts are financial instruments that serve economic purposes for various institutions and individuals. Option plays a crucial role in developing the financial market due to the high innovation and liquidity associated with it. However, due to option contract’s increased adaptability and responsiveness, its pricing mechanism has become complicated. The conventional parametric models suffer from various computing restrictions and implausible economic and statistical presumptions leading to deviations from real-world dynamics. Thus, data-driven strategies built upon non-parametric models seems compelling. Machine Learning (ML) serves as a powerful tool that can increase efficiency and productivity by automated processes, decreasing human biases and errors caused by psychological or emotional factors. Most of the existing literature involves only neural networks, whereas alternative algorithms remain undiscovered. This study explores the effectiveness of various ML algorithms through different experimentation. The ML algorithms harnessed for the study are Artificial Neural Networks (ANN), XGBoost, Decision Tree Regression, Support Vector Regression, Random Forest Regression, Long short-term memory (LSTM) Network and Gated recurrent unit (GRU) Network. Furthermore, multi-asset training and ensemble modelling are carried out to enhance predictive performance. A comparison is carried out with the seminal Black-Scholes model to highlight the advantages of the ML approach. The models are evaluated for European option contracts. The underlying assets used are NIFTY50 and BANKNIFTY indices from India’s National Stock Exchange (NSE). ML algorithms performed superior to the Black-Scholes model by a significant margin. Additionally, the models are evaluated on data collected following the outbreak of the COVID epidemic to get insight into the effects of abrupt changes in market sentiment.