Enhanced Traffic Prediction for Smart Cities through IoT using Optimized Continual Spatio-Temporal Graph Convolutional Network

In many cities, traffic congestion has a negative impact on sustainability since it increases air pollution. Effective smart traffic management can help users avoid congested areas, thus reducing pollutant levels. It is complex and dynamic in nature that is why traffic propagation forecasting is not perfectly accurate due to the existence of numerous interactions involved in traffic flow complexity., decision makers have large datasets, whose utilisation helps in designing brilliant, sustainable transport solutions. In order to address with these challenges, the following novel framework is developed namely Enhanced Traffic Prediction for Smart Cities through IoT using Optimized Continual Spatio-Temporal Graph Convolutional Network (CSTGCN-BTGO). In this work, the author has presented the traffic propagation model based on the traffic data of Buxton, UK. The feature extracted includes time, date, length, speed, flow, and headway of the traffic video which was accomplished using Differential Synchro-squeezing Wavelet Transform (DSWT). These features were then fed into a Continual Spatio-Temporal Graph Convolutional Network (CSTGCN) for traffic propagation over the road network forecasting. This model will simulate traffic in a particularly busy municipality for a 5-minute interval by employing data whose source will be traffic sensors positioned at two terminals, they are vehicle speed data terminals. To improve the predictive accuracy of the above-mentioned approach, Banyan Tree Growth Optimization (BTGO) was incorporated into CSTGCN. Two error measure tools called accuracy and mean squared error were the two tools used to evaluate the suggested CSTGCN-BTGO method. Outcome indicates that CSTGCN-BTGO model has acquired approximately sixteen percent improvement than the baseline model. 53% increase in accuracy than existing methods such congestion prediction using LSTM-IOT, traffic control using Deep SORT-IOT and traffic accident prediction using CNN-IOT.