Effectiveness of various machine learning methods in stock price prediction

Abstract

Stock investing has always been a risky endeavor and returns are never guaranteed. The way the stock price fluctuates is very hard to predict, as it can be affected by a lot of external factors and does not depend solely on the historical data of the stock. However, with the rapid improvement of machine learning models, these models can identify the patterns and predict stock prices with a high degree of accuracy. This report is going to test and evaluate three of these machine learning models, Particle Swarm Optimization Long Short-Term Memory, Particle Swarm Optimization Gated Recurrent Unit and Transformer, and find out which model performs the best. Using evaluation metrics such as Mean Square Error, Root Mean Square Error, Mean Absolute Error, Mean Absolute Percentage Error and R squared. This report then tests the models on larger datasets finding out how well the models perform on them. Followed by adding a technical indicator in the form of Bollinger Bands and investigated the difference in performance.