Chapter 4: An Intelligent Framework for Traffic Congestion Analysis System Using Deep Convergence Network

The evaluation of smart city infrastructure focuses on all areas of development, especially on the intelligent management of transportation. Smart traffic congestion management systems are demandable in the current scenario to provide hassle free instructions and alerts to make the smart move. Cities are not always overloaded with heavy traffic. Roads are less congested in certain time frames and seem highly congested in the peak timings. Development of smart traffic management systems confines the intelligent management of such time variations. It is required to control the traffic light during the low congestion time, high congestion time, and moderate congestion time. In the case of emergency vehicles passing into the junctions, the traffic lights are required to get the adaptive changes and make the controls appropriately. In similar cases, the location where the usual traffic system is changed into an emergency change needs to be updated to the traffic maintenance control room to alert the upcoming junctions. The planning and execution of dynamic status based congestion update and emergency vehicle passing update is mandatory in smart city executions. In the proposed system, real-time datasets are collected from open source as publicly available websites. The dataset consists of traffic congestion information in the form of recorded values of vehicle count, duration of congestion, location information with time stamps. The raw data is collected from the dataset window and preprocessed. In the preprocessing module, data is down sampled and framed into 1,000 samples of independent packets. The proposed model is initially developed by making the training process with the given dataset. To design the model, the given data is divided into 75% training set and 15% testing set, 10% validation set. Based on the performance measure of the proposed deep convergence network (DCN) in terms of accuracy, precision, recall, and F1 Score, of the trained model, further live data split up into test data is fetched into the system. DCN is considered as the dynamic and robust neural architecture that iteratively repeats the learning process until a very low error rate is obtained. The tested data is further evaluated by making the performance evaluations using the confusion matrix. The confusion plot is formulated with predicted results with the actual results using true positive rate, true negative rate, false positive rate, false negative rate, and so on. In case of higher correlation between the traffic congestion data of testing data with training data, then higher the accuracy will be reflected in the results. Further, the proposed system also included with emergency vehicle alert to the traffic control center. In case of emergency vehicle passed into the real time scenario, then immediate alert of vehicle information, with location data is transmitted into the traffic control room. In such cases, the fore coming junctions get immediate adaptations on traffic lights that will be highly helpful in making the route cleare to the emergency vehicle. The route optimization depends on the pattern analysis done with gradient stochastic optimizer in neural toolbox. The proposed model is developed with MATLAB 2021 Software.