A New Hybrid Deep Learning Algorithm for Prediction of Wide Traffic Congestion in Smart Cities

• Main Authors: G. Kothai, E. Poovammal, Gaurav Dhiman, Kadiyala Ramana, Ashutosh Sharma, Mohammed A. AlZain, Gurjot Singh Gaba, Mehedi Masud
• Format: Article
• Language: English
• Published: Hindawi-Wiley 2021-01-01
• Series: Wireless Communications and Mobile Computing
• Online Access: http://dx.doi.org/10.1155/2021/5583874

The vehicular adhoc network (VANET) is an emerging research topic in the intelligent transportation system that furnishes essential information to the vehicles in the network. Nearly 150 thousand people are affected by the road accidents that must be minimized, and improving safety is required in VANET. The prediction of traffic congestions plays a momentous role in minimizing accidents in roads and improving traffic management for people. However, the dynamic behavior of the vehicles in the network degrades the rendition of deep learning models in predicting the traffic congestion on roads. To overcome the congestion problem, this paper proposes a new hybrid boosted long short-term memory ensemble (BLSTME) and convolutional neural network (CNN) model that ensemble the powerful features of CNN with BLSTME to negotiate the dynamic behavior of the vehicle and to predict the congestion in traffic effectively on roads. The CNN extracts the features from traffic images, and the proposed BLSTME trains and strengthens the weak classifiers for the prediction of congestion. The proposed model is developed using Tensor flow python libraries and are tested in real traffic scenario simulated using SUMO and OMNeT++. The extensive experimentations are carried out, and the model is measured with the performance metrics likely prediction accuracy, precision, and recall. Thus, the experimental result shows 98% of accuracy, 96% of precision, and 94% of recall. The results complies that the proposed model clobbers the other existing algorithms by furnishing 10% higher than deep learning models in terms of stability and performance.