The vehicular Adhoc network is a new research area in the smart transport system that offers critical information to the network's cars. Road accidents affect about 150 thousand individuals, and VANET must improve safety. Traffic congestion forecasting is critical for reducing road accidents and enhancing traffic management for people. The dynamic behavior of the cars in the network, on the other hand, reduces the accuracy of deep learning models in predicting traffic congestion on roadways. This research introduces a new hybrid boosted long short-term memory ensemble and convolutional neural network (CNN) model to solve the congestion problem. to effectively handle the vehicle's dynamic behavior and estimate traffic congestion on routes The suggested BLSTME trains and improves the weak classifiers for the prediction of congestion, while the CNN collects characteristics from traffic photos. The suggested model is built with Tensor flow Python modules and evaluated in a real-world traffic situation with SUMO and OMNeT++. Extensive testing is done, and the model's performance is assessed using the performance criteria likely prediction accuracy, precision, and recall. As a consequence, the experimental result indicates 98 percent precision, 96 percent recall, and 94 percent accuracy. The findings show that the suggested model outperforms other current algorithms by 10% in terms of stability and performance when compared to deep learning methods.