Using COVID-19 pandemic sentiment and machine learning to predict stock market price direction

Abstract

The buying and selling of financial instruments, such as stocks and bonds, has for long been an essential activity to maximise investors’ wealth. Stock markets including the likes of the New York Stock Exchange, NASDAQ, and many others, have facilitated this trading activity to take place. A pertinent question in this field is to determine whether a particular security would increase or decrease in value in the foreseeable future. Algorithmic Trading, the buying and selling of instruments made using algorithms, has gained traction. This was possible through the exponential improvements in computational speed, along with the introduction of diverse Machine Learning algorithms. This work tries to exploit Machine Learning to predict the general price direction of securities for the three days following each day in the dataset range. This is referred to as time-series forecasting. The solution proposed uses data for stocks domiciled in the United States found in the S&P500 index, which is an indicator representative of the largest 500 US-listed companies. Two stocks per S&P500 sector are reviewed, to have a fair representation of the US market. A baseline Artificial Neural Network (ANN) model, and a Long Short-Term Memory (LSTM) model are used, for which performance is evaluated and compared accordingly. The latter model has recently become popular in time-series problems due to its ability to remember the previous outcome of neurons. The COVID-19 pandemic has affected businesses in a global manner; thus, this work also tries to identify whether there is value to be derived from the sentiment towards the pandemic, in terms of prediction accuracy. Google Trends data involving pandemic-related terminology are used to derive additional features to be used within the models implemented, providing an effective comparison between the model implementations in this work. Results proved to be encouraging, whereby Google Trends data improved the results in terms of Accuracy Score and Receiver Operating Characteristic Area Under the Curve (ROC AUC) score for the majority of equities selected. Furthermore, the LSTM model attested to being slightly superior for stock market prediction when compared with ANN. Potential for future work to be done in this area is discussed with relevance to the work presented.