A novel approach to evaluate machine learning models for bitcoin price prediction using multi-criteria decision making
Ajay Kumar, Anuj Kumar Singh, Ankit Garg · 2024
Bitcoin is now more often regarded as a viable investment, after the recent ups and downs in cryptocurrency valuation. Because of its volatility, accurate forecasts are required to support investment decisions. Numerous researchers have developed various models for bitcoin price prediction. However, in the presence of several competing accuracy measures, various models may exhibit differing capability for predicting bitcoin price. This study&s;s objective is to evaluate the performance of several machine learning-based forecasting models for predicting bitcoin price using multi-criteria decision making (MCDM), considering various competing accuracy measures simultaneously. The forecasting models evaluation can be described as an MCDM problem because it comprises several performance measures. An experimental study was conducted to examine the proposed approach using one MCDM method, eight forecasting models, and four performance measures over a bitcoin price dataset. Based on the final MCDM rankings, multilayer perceptron (MLP) is recommended for the bitcoin price prediction. The study&s;s experimental findings demonstrate that the suggested approach can be used as a successful decision-making tool to select the most suitable forecasting models for bitcoin price prediction by considering various accuracy metrics simultaneously.