Bitcoin Price Prediction Using Enhanced Transformer and Greed Index

Abstract

The dramatic fluctuations of cryptocurrency prices in recent years have led to various studies on price forecasting, among which transformer-based forecasting methods have better prediction results on time series data. However, a large number of studies in the past have shown that transformer has the disadvantages of high computational cost and inability to predict global information and does not consider external information. To solve these problems, we propose an adaptive decomposition method based on transformer, which combines transformer with seasonal trends and external factors, i.e. the global information of the time series obtained by the decomposition method through adaptation is combined with external information. We conducted experiments on different datasets, and the results show that our method can significantly improve the performance of the baseline model.