S&P 500 Index Price Prediction Using Twitter
Open AccessAbstract of PraxisS&P; 500 Index Price Prediction Using TwitterPredicting Standard and Poor’s (S&P;) 500 stock index price and directional movement is extremely complex (Sanboon et al., 2019), and inaccuracies result in 1% to 5% annual loss of the value of investments (Vasileiou, 2022). The accuracy of prediction for S&P; 500 enables investors to earn money not only when the index is going up, but also when the index is going through a downturn. In the United States, a stock or index equivalent can be shorted, which means it can be sold before acquiring it. As such, accuracy of prediction is rewarding in all market situations.This praxis aimed to establish a predictive model using Twitter feeds to improve the accuracy of predictions of the S&P; 500 index price and change of direction (up/down). Due to the large number of tweets related to S&P; 500, Forbes magazine was used to obtain the 100 most influential finance Twitter accounts. Information on whether investors have a positive or negative intent for buying or selling stocks was extracted from tweets on the social media platform. This was done by mining the tweet sentiment polarity from the tweet texts, leveraging multiple off-the-shelf models including BERT and FinBERT. Due to the specifics of tweet text and the S&P; 500 domain, a custom BERT model, named SP500BERT, was trained to provide a sentiment polarity score for S&P; 500 tweets. By augmenting the sentiment polarity for S&P; 500 tweets with the index price, volume, and trading date statistics, data sets were created and fed into a multitude of machine learning programs for regression and classifications. The SP500BERT derived classification led to accuracies higher than the current 66.18% benchmark. The results varied from 75% to 81%, depending on the selected algorithm. While the praxis focused on S&P500; price and directional movements, the techniques and methods used in this paper can be leveraged to predict the price for other financial instruments that are influenced by social media.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.