Non-Equidistant Grey Model Based on Background Value and Initial Condition Optimization and Its Application

Abstract

As the background value and the initial condition are important factors affecting the precision of grey system model, in this paper, we put forward an improved non-equidistant GM(1,1) model based on PSO according to the practical application need, in which the background value is optimized firstly and a new initial condition is presented based on the principle of new information priority. Under the algorithm of minimizing the square sum of the relative error between the original series and the forecasting sequences, the solution to the optimized time parameter is given, the particle swarm optimization (PSO) algorithm is used as a tool to optimize the parameter in the background value and the initial condition. The experimental result shows the effectiveness and applicability of the proposed non-equidistant GM(1,1) model.