CNN-DDQN based, Self Learning, Automated, Trading System

Parmar, Rushabh (2023) CNN-DDQN based, Self Learning, Automated, Trading System. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration manual) Download (2MB) | Preview

Abstract

Trading is a widely studied field in terms of psychology as well as technology. While there are ample trading systems that provide an automated trading environment based on calculative trends, there are still minimal algorithms that capture the market sentiments to bet their decisions. Recent advancements in Convolutional Neural Networks (CNN) have helped us understand the important features of a candlestick chart and capture the market trends to output a trading decision. At the same time, Deep Reinforcement Learning algorithms like Double Deep Q-Network (DDQN) have proven successful in creating a self-learning trading system. This research demonstrates a CNN-DDQN-based model, where CNN outputs a trading decision given 24-day Gramian Angular Field (GAF) images as input, acting as an agent in DDQN, which incorporates the trading environment. The GAF is what is being tested in the research with a motive to enhance the performance of the CNN-DDQN model to make better trading decisions. CNN helps us acknowledge the psychological perspective of the market, whereas we fine-tune the decisions made over time by using the Q-Network at the core. The GAF image, with 5 channels stacked vertically as features, outperforms all the other data representations with the highest weighted F1 score of 0.813, by CNN to classify the images correctly.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Makki, Ahmed UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Algebra > Algorithms > Computer algorithms H Social Sciences > HG Finance > Investment > Stock Exchange
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	28 Dec 2024 14:13
Last Modified:	28 Dec 2024 14:13
URI:	https://norma.ncirl.ie/id/eprint/7246

Actions (login required)

View Item