An Integrated Fuzzy Analytic Network Process and Fuzzy Regression Method for Bitcoin Price Prediction

Predicting the prices of cryptocurrencies is more complicated than that of classical financial assets because they do not seem to have reached the maturity stage of their life. In addition, many known and unknown factors may affect Bitcoin prices; these factors and their importance seem to be changing faster than other financial assets. Therefore, the data used to predict the prices of cryptocurrencies can be considered big data challenging to manage due to their volume, variety, and variability. This study presents an integrated approach to managing the data when predicting the Bitcoin price. We first prepare a list of factors affecting the Bitcoin price. We then use the Fuzzy Analytic Network Process (FANP) to screen these factors and select the most important ones based on the experts' opinions. The selected factors are considered independent variables affecting the Bitcoin price. Next, we extract a fuzzy regression model using the historical data in which the Bitcoin price is considered the dependent variable. Finally, this model is validated with different confidence levels, and the appropriate level is selected to predict the Bitcoin price. The results show that Bitcoin prices fall within the forecasting intervals obtained from the fuzzy regression model for a 99% confidence level. Unlike crisp regression models, the fuzzy regression model used in this study does not predict the Bitcoin price as a crisp value; instead, it predicts the price as an interval value. The contributions of this study are fourfold: (1) identifying the factors affecting the Bitcoin price and investigating their mutual impacts on each other; (2) determining the most influential factors using the FANP method; (3) using fear and greed as essential sentimental independent variables in regression to predict the Bitcoin price; (4) and predicting the Bitcoin price as an interval instead of a crisp value.