Optimization and Application Exploration of Quantitative Trading Strategy of Reinforcement Learning in Cryptocurrency Trading

Abstract

The popularity of cryptocurrency markets has significantly increased, which has encouraged many financial traders to pursue huge gains in cryptocurrency trading. Technical analysis and machine learning have also been integrated by some researchers and investors to predict future market patterns. Nevertheless, creating profitable trading strategies is still seen to be a very difficult undertaking, even with the application of these techniques.  This manuscript proposes an exploration of quantitative trading strategy of reinforcement learning in cryptocurrency trading with multi-scale fusion self attention generative adversarial network (EQTS-CT-MFSGAN). Initially, the data is collected from Open-High Low-Close-Volume (OHLCV) market data. Afterward, the data’s are fed to pre-processing. In pre-processing segment, Affine-Mapping Based Variational Ensemble Kalman Filter (AM-VEnKF) is used to clean the data. The outcome from the pre-processing data is transferred to the MFSGAN.  The MFSGAN method is used to classify the cryptocurrency trading such as high risk, low risk and no risk. The Hippopotamus Optimization Algorithm (HOA) is used to optimize the weight parameter of MFSGAN. The proposed technique is implemented in Python and the efficiency of the proposed EQTS-CT-MFSGAN technique is estimated with the help of several performances like accuracy, precision, recall, sensitivity, specificity, F measure and cumulative profit. Proposed SS-ABSR-MFSGAN method attains 21.34%, 23.54% and 23.76% higher accuracy, 21.34%, 21.39% and 20.28% higher precision, 18.82%, 20.53% and 23.79% higher f1-score are analyzed with existing techniques like Deep reinforcement learning for the optimal placement of cryptocurrency limit orders (OPCLO-DDQN), UNSURE-A machine learning approach to cryptocurrency trading (UCT-TCN) and Multi-Agent Deep Reinforcement Learning With Progressive Negative Reward for Cryptocurrency Trading (PNRCT-MAPPO) respectively.