A data-driven emergency plan evaluation method based on improved RIMER

Implementing suitable emergency plan can greatly reduce the damage and risk caused by emergencies. However, the impact of emergency plan on the response effect to emergency is complex, which makes emergency plan evaluation difficult. Rule-base inference methodology using the evidential reasoning (RIMER) can simulate real complex systems, and adopting RIMER to evaluate emergency plans would rely on less subjective information, so RIMER can be regarded as a competitive potential evaluation tool. In conventional RIMER, the antecedent attributes are usually selected subjectively. Unfortunately, subjectively selecting the antecedent attributes that affect rescue results is difficult due to the complexity of emergencies. Therefore, the joint optimization model for antecedent attributes and belief rule base parameters is constructed in this paper. However, using meta-heuristic algorithm to solve the joint optimization model will face encoding and optimization difficulties. So the joint optimization is simplified into an optimization problem for antecedent attributes and their weights, and a data-driven scheme is designed to determine belief rule base parameters. Finally, a case study on a wildland fire rescue dataset is carried out to verify the effectiveness of the proposed method. Compared with other relative methods, the proposed method shows competitive performance in both predictive accuracy and interpretability.