Decision-making model for emergency evacuation based on the lens model using machine learning and Monte-Carlo simulation for incomplete information environment

Preparing for emergencies reduces significant losses to infrastructure and the economy. In this study, a decision-making analysis tool, Lens model (LM), is used to characterize the decision behavior during an emergency evacuation based on multiple cues. Five Supervised Machine Learning (SML) algorithms were used to derive the LM parameters. The decision to evacuate under uncertain and incomplete information is always challenging. However, the LM consisting of the ecological-judgment models was created to under-stand evacuation behavior in uncertain environments fully. The judgment model was consolidated from historical data, whereas the ecological data, the incomplete information, was simulated using the Monte-Carlo Simulation (MCS). The SML models were evaluated using prediction accuracy (PA), and their performance validated by comparing the measures to the LM parameters. Experimental results show that k-nearest neighbor (KNN) achieved the least error in the ecology model as the LM parameter, Re corresponds to the performance of the algorithm model.