Predictions of macroeconomic indicators based on matrix multiple regression: examples

Abstract

For formulation and solution of the prediction problem based on the method of least squares the class of matrix functions with the set of matrix arguments (matrix multiple regression) is used. The successive solution of the prediction problem was performed using the mathematical apparatus of the singular value decomposition and the Moore–Penrose pseudoinverse technique under the development of tuple operators. The algorithm for estimating the unknown parameters was implemented in Wolfram Mathematica. The proposed method was demonstrated for predictions of basic macroeconomic indicators of Ukrainian economics. The approach was verified using statistical data about economic indicators of Ukraine for the period of 2007–2016. The results of calculations were presented. As shown in examples, the matrix multiple regression can be an effective prediction instrument in economics with an acceptable for planning processes accuracy.