Deep Q network framework on stock price prediction

Abstract

Listen

Translate article

Stock price forecasting seeks to forecast future changes in a trading exchange's share price. More profit can be made by investors with the help of correct stock price prediction. Artificial Intelligence methods such as Deep reinforcement learning has become more important for sound decision-making across a range of areas. Deep reinforcement learning has been utilized in a range of fields, including education, healthcare, transportation, finance, video games, robotics, computer vision and natural language processing. This model has proven efficient in handling a variety of challenging decision-making tasks that were previously beyond of the machine's capabilities. As a consequence, it is a model that can be used in creating smart frameworks that assists us in predicting when to buy or sell a specific stock. With limited historic information, reinforcement learning performs quite well. Within this work, we present a Deep Q-Network architecture that makes use of Q-Learning, off-policy reinforcement learning algorithm to determine which action an agent should perform in response to an action-value function. Using an action-value function, Q learning determines the value of being in a certain state and taking a particular action at that state.