Enhanced genetic-chaos based grasshopper optimisation algorithm for efficient crash risk prediction using novel deep learning model

The development of the road transport advanced collision avoidance system, which aids in accident prevention, heavily relies on accident risk prediction.The prediction of crash risk before a collision has been made possible by a number of deep learning algorithms and hybrid data driven models.In the existing convolutional-neural-network-long-short-term-memory (CNN-LSTM) model, the hyper parameters are not optimised which affects the crash risk prediction system efficiency.Therefore, to improve the collision risk prediction performance, enhanced genetic-chaos based grasshopper optimisation algorithm (EGCGOA) has been developed which optimises CNN-LSTM model's hyper parameters.CNN and LSTM are employed for data extraction and prediction respectively.The performance of EGCGOA has been evaluated with test functions, which prove the efficiency algorithm.The next generation simulation project (NGSIM) dataset was utilised to test the EGCGOA-CNN-LSTM, and it performs better than conventional techniques.The accuracy has increased which is greater than the previous model, based on the results.