Project Abstract

This project aims to investigate the potential of machine learning to forecast stock market trends to address the challenges investors face due to the stock market’s complexity. This is done by investigating the implementation of a Long Short-Term Memory (LSTM) with a Recursive Neural Network (RNN) model trained on both long-term daily and minute-by-minute stock prices across volatile and stable market conditions. A diverse set of features including opening, closing, volume, high and low values were integrated to showcase the differences in predictive accuracy between basic and enhanced feature sets. To evaluate the model several performance metrics were used including Mean Squared Error, Root Mean Squared Error, Mean Absolute Error, and R2 as well as plotting the prediction for a visual representation.To further help investors with the stock market, a web application was developed to present financial data as well as perform real-time predictions using the LSTM model showcasing the model?��s performance. The findings show that while the LSTM model provides a high accuracy in predicting long-term data for stable stocks which is demonstrated by strong R�� values, its performance decreases with higher stock volatility and in short-term scenarios.This project concludes that even though the LSTM model is effective for long-term prediction, it struggles in short-term market prediction making it more of a secondary decision tool rather than primary in these situations. Overall, this highlights the machine learning potential as both primary and secondary decision-making tools in different trading contexts.

Keywords: Machine Learning, Recursive-Neural-Network, Stock Market Forecasting

 Conference Details

Session: Presentation Stream 34 at Presentation Slot 2

Location: CoFo 002 at Wednesday 8th 13:30 – 17:00

Markers: Mark Jones, Galileo Sator (GTA)

Course: MSci Computer Science, Masters 4th Year

Future Plans: I’m looking for work