An Interval Knowledge Based Forecasting Paradigm for Container Throughput Prediction

Abstract

Abstract Current knowledge based forecasting models are suffering from weaknesses of subjective biases and inconsistence. In order to overcome this problem, this paper proposes a novel interval knowledge based forecasting paradigm. In the proposed forecast- ing paradigm, statistical projections of the target are first generated by statistical models. Next, a panel of experts are gathered to independently formulate their interval estimates, then this kind of interval knowledge is integrated into the statistical pro- jections. Subsequently, an expert performance validating algorithm is put forward to wipe off incompetent members from the expert system, and then a Delphi based expert system is constructed to regenerate interval judgments with less subjective biases and inconsistence. Meanwhile, the algorithm is able to determine the weight distribution, with which statistical projections and interval judgments are integrated into the united predicted values. For verification purpose, container throughput series of Qingdao Port are taken as sample data. Empirical results clearly show the superiority of the proposed interval knowledge based forecasting paradigm over its benchmark models, which indicates that the proposed forecasting paradigm is effective for container throughput prediction.