A Hybrid Deep Learning-based Chaos Dynamic System for CryptoCurrency Prediction

Abstract

The financial market's price forecast for cryptocurrencies is of huge importance, particularly until the current global financial and economic crisis. Because of the nonlinear structure that involves the intrinsic fractality and chaoticity of cryptocurrencies, several studies found that a single method is not adequate to predict cryptocurrencies with quite greater resolution. While each framework used in cryptocurrency prediction has limitations and also the other capabilities, they may not provide the highest prediction performance for that duration in all scenarios. In the cryptocurrency time sequence, an efficient and modern prediction system was introduced to reduce this detrimental condition and maximize statistical performance. A new hybrid prediction model focused on long short-term memory (LSTM) neural network and Chaotic dynamic technique for cryptocurrency regression analysis is built in this research. While the Long short - term memory model's computing task is extremely high in nonlinear pattern classification as compared to brute force, deep learning rapidly proved extremely effective in predicting the underlying unpredictable nature of the cryptocurrency.