Publication Date

Spring 2019

Degree Type

Master's Project

Degree Name

Master of Science (MS)


Computer Science

First Advisor

Katerina Potika

Second Advisor

Sami Khuri

Third Advisor

Mike Wu


Reinforcement Learning, Artificial Neural Networks, Deep Learning, Recurrent Neural Networks, Long Short-Term Memory, Time Series Analysis, Deep Q-learning, Direct Reinforcement Learning


Can we train a stock trading bot that can take decisions in high-entropy envi- ronments like stock markets to generate profits based on some optimal policy? Can we further extend this learning for any general trading problem? Quantitative Al- gorithms are responsible for more than 75% of the stock trading around the world. Creating a stock market prediction model is comparatively easy. But creating a prof- itable prediction model is still considered as a challenging task in the field of machine learning and deep learning due to the unpredictability of the financial markets. Us- ing biologically inspired computing techniques of reinforcement learning (RL) and artificial neural networks(ANN), this project attempts to train an agent who takes decisions based on the optimal decision policies learned. Different existing RL tech- niques and their slightly modified variants will be used to train the agent, and the trained model is then tested against different stock prices and also stock portfolio settings to see if the agent has learned the rules of the game and can it act optimally irrespective of the testing data provided. This work aims to provide general users with simple recommendations about the possible investment decisions of selected stocks in the portfolio. Results of the implemented approaches do seem to work somewhat well on specific periods of stock market time series, but they are observed to be fragile. Selected strategies do not guarantee similar results on all historical time-periods, nor they are guaranteed to provide exceptional performance on unpredictable future stock market time-series data.