Analysis of bitcoin prices using a heavy-tailed version of Dagum distribution and machine learning methods.

Statistical modeling and forecasting are very important for decision-making in any field of life. This paper has two major objectives, namely, statistical modeling and forecasting of real phenomena. For covering the first aim (i.e., statistical modeling), we introduce a new probabilistic model. The new model is introduced by mixing the Dagum distribution with the weighted T X family approach. The proposed model is called the weighted T X Dagum distribution and possesses heavy-tailed characteristics. The new model is illustrated by analyzing real-life data related to Bitcoin prices. To cover the second aim (i.e., forecasting), we take into account six macroeconomic and financial indicators to investigate their impact on Bitcoin prices such as the Adaptive least absolute shrinkage and selection operator (Alasso), elastic net, and minimax concave penalty. After analyzing the data, it is found that Alasso and MCP have retained all the included predictors, except import, while Enet holds all the predictors. The root means square error and mean absolute error associated with MCP are lower than Alasso and Enet, which reveals that MCP fits the data very well as compared to rival methods. [ABSTRACT FROM AUTHOR]