The School of Computer Science is pleased to present…

Deciphering Market Dynamics: Node2vec and TD-3 for Intelligent Stock Prediction

 

MSc Thesis Proposal by:

Mohammed Farhan Baluch

 

Date: December 13, 2023

Time:  11:30 am – 12:30 pm

Location: Essex Hall, Room 105

 

Abstract:

This thesis presents a novel work in the realm of stock market prediction by integrating graph embeddings, specifically node2vec, within a Reinforcement Learning (RL) framework, with a focus on the Twin Delayed Deep Deterministic Policy Gradient (TD-3) algorithm. The research emphasizes the value of interpretability in machine learning to elucidate the decision-making processes of RL agents. A central contribution of this thesis is the novel integration of graph embeddings into the RL paradigm. By constructing a stock correlation graph and applying node2vec, the study generates detailed embeddings that reflect intricate stock interrelations and characteristics. These embeddings provide a rich, context-aware input for the RL model, enhancing its ability to make informed and strategic trading decisions. The research distinctly highlights the efficacy of the TD-3 algorithm in the context of stock trading. TD-3, known for its stability and robust performance in continuous action spaces, is meticulously analyzed and demonstrated to be particularly adept at navigating the complex, dynamic environment of the stock market. This underscores the potential of TD-3 as a superior tool in financial trading strategies. Moreover, a key finding of this study is the integration of interpretable machine learning techniques, which shed light on the decision-making process of the RL agent. This aspect not only increases the trustworthiness of the AI models but also provides valuable insights into how the models process and respond to various market conditions.

 

Keywords: Reinforcement Learning, Stock Prediction, TD-3, Node2Vec, Interpretable Machine Learning

 

Thesis Committee:

Internal Reader: Dr. Robin Gras 

External Reader: Dr. Esam Abdel-Raheem            

Advisor: Dr. Luis Rueda