Time: 13:00-15:00 (GMT), Wednesday, 10 March 2021
Presenter: Dr Yossi Shvimer, Centre for Global Finance, SOAS University of London
Chair: Professor Victor Murinde, SOAS University of London
Online venue: Click here to join the meeting on Microsoft Teams (For any inquiry about how to join the online seminar, please contact Dr Meng Xie: xm1@soas.ac.uk)
Abstract
Forecasting stock market behavior is a challenging problem. Forecasting the risk level associated with stock price futures provides an additional complexity level since it represents a derivative (hence, less stable, with more degree of freedom) of the stock market dynamics. This paper introduces a new next day forecasting model for the Chicago Board Options Exchange (CBOE) Volatility Index (VIX). We show the advantages of adding investors’ sentiment scores to a new hybrid model encompassing the long-short term methodology (LSTM) model and the auto-regressive integrated moving average (ARIMA) model. The sentiment scores are empirically evaluated via machine learning natural language processing (NLP) tools based on commonly used economic sites. The hybrid model shows robust results on the forecasts of the next day VIX level. Based on out-of-sample for 2019-2020 end of day data (including COVID-19 out-of-sample data), the hybrid LSTM-ARIMA model shows that the addition of sentiment scores yields higher accuracy by 5% than the hybrid LSTM-ARIMA without investors’ sentiment scores, consequently generating higher trade profits.
Keywords: Forecasting; VIX index; Hybrid LSTM-ARIMA model; COVID-19 crisis; Sentiment analysis
JEL Classification: C55
Authors: Yossi Shvimer, Victor Murinde, and Avi Herbon
Presenter
Dr Yossi Shvimer
Yossi is a post-doctoral Research Associate in the Centre for Global Finance. He holds a PhD. in finance from Bar-Ilan University (Israel). Yossi is also the CEO of MCM Alternative Investments, and also serves in key positions within Migdal Capital Markets, Ltd. His research focuses on finance and applied machine learning in finance. His ongoing work seeks to find advanced techniques for financial asset pricing. His work has been published in The North American Journal of Economics and Finance and Journal of Economic Interaction and Coordination.