Abstract

Traffic flow prediction is an essential element in intelligent transportation models, aiming to forecast vehicular movement patterns on road networks. This system aids in diminishing traffic issues, particularly traffic congestion, to enhance traffic quality. This system uses extensive data information and transmission technology to provide more efficient traffic control and real-time road infrastructure analysis. This system relies on traffic prediction as a crucial component. Traffic prediction main intention is to forecast future traffic conditions on transportation networks based on historical analysis. Using artificial intelligence technology, especially machine learning, to generate and create traffic flow prediction is ground-breaking. It provides conventional techniques for generating and developing prediction of traffic flow. Deep learning (DL) is a machine learning (ML) subclass strongly associated with each other in artificial intelligence. In this aspect, this manuscript devises an effective traffic flow prediction using the Kookaburra Optimization Algorithm with a hybrid DL (ETFP-KOAHDL) model. The aim of the ETFP-KOAHDL method is to provide a reliable and efficient solution for real-time traffic flow prediction using advanced DL models. Initially, the ETFP-KOAHDL technique follows a min-max scaler to pre-process the input data. Besides, the HDL algorithm implements the prediction process, which incorporates convolution long short-term memory with an autoencoder (CLSTM-AE) system. Finally, the ETFP-KOAHDL model designs the KOA for the optimal hyperparameter selection. An extensive set of experiments was conducted to examine the improvised prediction results of the ETFP-KOAHDL method. The simulation exploration of the ETFP-KOAHDL system exposed an excellent MAPE value of 04.82% over other approaches under diverse runs. Also, the process can predict traffic flow at a consecutive time, which helps in effective travel planning for autonomous vehicles.