Outperforming Cryptocurrency Price Prediction Using Deep Reinforcement Learning Approach

Abstract

Cryptocurrency price prediction is a complex and dynamic task. Cryptocurrencies are highly volatile digital assets impacted by a wide range of variables, including investor mood, legislative changes, market demand, and technical breakthroughs. This research presents a novel approach to cryptocurrency price prediction using Deep Reinforcement Learning (DRL). The proposed model incorporates proximal policy optimization (PPO) and a Hybrid Convolutional Neural Network Long Short-Term Memory(CNN-LSTM)model. The proposed model was evaluated for three benchmark datasets such as Bitcoin (BTC), Litecoin (LTC), and Ethereum (ETH). This novel approach significantly improves the prediction rate by investigating price movements and optimizing investment returns. The experimental results show the anticipated daily and weekly returns and the actual and forecasted prices using a sentiment analysis approach. The proposed model outperforms existing approaches by obtaining profits scores such as 1.04 for the LTC dataset, 0.98 ETH dataset, and 0.97 BTC dataset. These scores justify the effectiveness of the proposed model by generating favorable returns in cryptocurrency investments.