Enhancing Cryptocurrency Prediction: A Fusion of Machine Learning and Neutrosophic Programming

Abstract

Fluctuations in cryptocurrency markets present a significant problem to the accuracy of forecasting
 trends and prices in this field. This paper proposes a new method of improving cryptocurrency forecasting
 by applying machine learning algorithms to a hybrid model. The framework integrates, a neural network
 model, with auto regressive integrated moving average (ARIMA) and trigonometric, Box-Cox, ARMA, Trend,
 Seasonal (TBATS) to capture the intricate relationship and dynamics in the data. Because most aspects
 affecting the cryptocurrency’s price are uncertain, we propose that fuzzy parameters be used to reflect this
 uncertainty in the market. Furthermore, we apply neutrosophic programming to optimize predictions where the
 indeterminacy of the data is considered. The hybrid model thus incorporates short-term market volatility and
 long-term market trends, making the model rigid and accurate. Here, we compare this approach’s performance
 with other forecasting models using actual cryptocurrency data. The results indicate that the hybrid model
 developed achieves better predictive accuracy and is more flexible than the conventional models. To sum
 up, this research offers significant knowledge of applying the newest machine learning methods to enhance
 cryptocurrency prediction and improve its efficiency for investors, traders, and financial institutions

DOI: 10.5281/zenodo.16710195