Evacuation time prediction method for large cruise ships based on a deep extreme learning machine

Evacuation time prediction is crucial for large-scale emergency evacuation management and ship layout optimization. This study proposes a method for predicting personnel evacuation times based on a deep extreme learning machine (DELM). First, the weighted mean of vectors algorithm is improved by incorporating opposition-based learning and a dynamic candidate mechanism to avoid local optimums and accelerate algorithm convergence. Then, a data set of personnel information related to the evacuation time is generated by simulating personnel evacuation on a large cruise ship. Finally, the evacuation time is predicted using the personnel information data set and the DELM optimized by the Improved weighted mean of vectors algorithm (INFO). The experimental results show that the mean absolute percentage error (MAPE) of IINFO-DELM with test data is 2.09 %. The results show that the IINFO-DELM can efficiently and quickly predict evacuation times. • Emergency evacuation simulations were carried out, and the data sets of evacuation time were created. • IINFO-DELM was proposed by using Opposition-based learning and dynamic candidate mechanism. • By combining IINFO-DELM and the evacuation simulation, the personnel evacuation time was predicted. • Real cruise ship models and crowd distribution data are used for experiments.