Does Interval Knowledge Sharpen Forecasting Models? Evidence from China’s Typical Ports

Abstract

Substantial studies integrating experts’ point knowledge with statistical forecasting modes have been implemented to investigate a long-lasting and disputing issue which is whether or not expert knowledge could improve forecasting performance. However, a large body of current forecasting studies neglect the application of experts’ interval knowledge where experts are expected to be more competent, considering that humans do much better in fuzzy calculation like interval estimation than in accurate computation like point estimation. To fill in this gap, this paper first proposes a novel forecasting paradigm incorporating interval knowledge generated by a Delphi-based expert system into the SARIMA and SVR models. For validation purposes, the proposed paradigm is applied to several representative seaports from the top three dynamic economic regions in China. The empirical results clearly show that interval knowledge, following the proposed paradigm, significantly improves the forecasting performance. This finding implies that the proposed forecasting paradigm has the good potential to be an effective method for sharpening the statistical models for container throughput forecasting.