STGA-CBR: A case-based reasoning method based on spatiotemporal trajectory similarity assessment

Abstract

Decision making are critical for disaster prevention and emergency response. To fully improve the effectiveness of emergency decisions by taking spatiotemporal information into consideration, this paper proposes a new method STGA-CBR for spatiotemporal case matching by using an integrated approach comprising a newly proposed spatiotemporal trajectory similarity measurement algorithm (Position-Frequency algorithm), a genetic algorithm (GA), and a case-based reasoning (CBR) technique. It consists of three main phases: (1) similar spatiotemporal trajectory retrieval; (2) weight determination; and (3) attribute similarity calculation. The proposed approach was employed in typhoon disaster, which contains a variety of spatiotemporal information. The results of matching were validated by comparing STGA-CBR with ST-CBR, GA-CBR and traditional simple CBR. The experimental results proved that the proposed STGA-CBR effectively screens out similar spatiotemporal trajectories and demonstrates higher matching performance than there other methods, indicating the high efficiency of the proposed similar case retrieval approach. The case pair selected are then used for prediction of the post-disaster social and economic loss and the average accuracy of prediction results are calculated, among which the integrated model rank the highest, thus rendering our approach superior in comparison to other traditional methods.