Modelling Stock Market Volatility Using Random Forest

Abstract

In recent years, stock market volatility has had an increasing role in investment decision making on what stock to trade in, in order to reap on the returns. Owing to the high uncertainty in the stock markets; stock market volatility has become equally helpful in many micros as well as macro-economic decision-making. Since stock activity is chaotic and volatile, it is risky to invest in stock market. Traders would consider purchasing a share whose price is expected to rise in the coming days but avoid a share whose price is likely to fall in coming days. To limit the risk of losing money, one needs a lot of knowledge about how stock prices will move in the future. As a result, it is necessary to accurately estimate stock market price movements. This research project aimed at modelling stock market volatility using random forest over a period of 12 years that is from 1st Sept 2009 to 31st Aug 2021. Random Forest is an algorithm utilized for classification and regression usually derived from a set of classification and regression trees. So many models that have been developed to model stock market volatility have focused on statistical methods such as time series analysis, regression analysis, and multivariate analysis. However; most researchers have learnt that stock prediction performs exemplary when regarded as a classification problem instead of as an analytical problem. Accordingly, the study considered modelling stock market volatility using machine learning techniques and technical indicators. Random forest had the advantage of performing well in big feature sizes and efficiently handling complex data. The random forest package in R was used for analysis, and the Miss Forest package was used for imputation. On the basis of the model's classification accuracy, model diagnostics, and interpretation, conclusions were drawn.