A novel three-step procedure to forecast the inspection volume

摘要

The inspection process of freight traffic at Border Inspection Posts (BIPs) generates significant time delays and,congestion within the transport system. The use of forecasting methods to anticipate these situations could be a good solution. Traditional methodologies for time series prediction usually consist on: applying single techniques, combining these techniques with some others such as clustering techniques or hybridizing single prediction techniques. A novel methodology based on a three-step procedure is proposed in this paper in order to better predict the number of inspections at BIPs, integrating a clustering technique and a hybrid prediction model. Specifically, the seasonal auto-regressive integrated moving averages (SARIMA) is used first to predict the data. Then, self-organizing maps (SOM) decomposes the time series into smaller regions with similar statistical properties. Finally, Artificial Neural Networks (ANNs) are applied in each homogeneous regions to forecast the inspections volume, testing different hybrid approaches based on the inputs of the model. The experimental results show that the performance of inspection prediction can be enhanced by using the novel three-stage procedure, providing relevant information for resource planning and turning into a powerful decision-making tool, not only at the inspection process of seaports or airports, but also in the field of time series prediction. (C) 2015 Elsevier Ltd. All rights reserved.