Evaluating forecasting performance for interval data

Abstract

From the overlapping parts and the non-overlapping parts of the actual intervals and the forecast intervals, it should be defined a criterion which is more efficient to evaluate forecasting performance for interval data. In this paper, we present evaluation techniques for interval time series forecasting. The forecast results are compared by the mean squared error of the interval, mean relative interval error and mean ratio of exclusive-or. Simulation and empirical studies show that our proposed evaluation techniques for interval forecasting can provide a more objective decision space in interval forecasting to policymakers.