Stock Market Prediction and Investment using Deep Reinforcement Learning- a Continuous Training Pipeline
Amritha Sharma R1, Debjyoti Guha2, Hitesh Agarwal3, Kothiya Meetkumar Harshadbhai4
1Amritha Sharma R*, CSE, MVJ College of Engineering, Bengaluru, India.
2Debjyoti Guha, CSE, MVJ College of Engineering, Bengaluru, India.
3Hitesh Agarwal, CSE, MVJ College of Engineering, Bengaluru, India.
4Kothiya Meetkumar Harshadbhai, CSE, MVJ College of Engineering, Bengaluru, India.
Manuscript received on December 02, 2020. | Revised Manuscript received on December 05, 2020. | Manuscript published on December 30, 2020. | PP: 93-98 | Volume-10 Issue-2, December 2020. | Retrieval Number: 100.1/ijeat.B20341210220 | DOI: 10.35940/ijeat.B2034.1210220
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Fluctuating nature of the stock market makes it too hard to predict the future market trends and where to invest. Hence, there is a need for a cross application backed by an ultramodern architecture. With the latest advancement in Deep Reinforcement Learning, successive practical problems can be modeled and solved with human level accuracy. In this paper, an agent-based Deep Deterministic Policy Gradient system is proposed to imitate professional trading strategies which is a state-of-the-art framework that can predict and make investment of customers money with high return. In addition to this, dealing with interday trading strategy, the proposed architecture is designed as a continuous training pipeline so that the model saved is up-to-date with the recent market trends by giving higher accuracy in prediction. The framework outperforms the base reinforcement learning algorithms and maximizes portfolio return. The experimental result shows how natural language processing and statistical prediction can help us to choose the trending stock based on news headlines and historical data so that model invests money only in the market which gives higher return. To evaluate the performance of the proposed method, comparison of our portfolio results was done with various other reinforcement learning algorithms by keeping the same configuration.
Keywords: Deep Reinforcement Learning, Artificial Neural Network, Artificial Intelligence, Machine Learning, Deep Learning, Stock Market
Scope of the Article: Deep Learning