Traffic Congestion Prediction in Smart Cities using Multilevel-Gated Recurrent Unit

Traffic Congestion (TC) is a main issue in smart cities because of increase in urbanization, as a result in financial losses. However, existing prediction algorithms failed to predict TC effectively due to dynamic changes of road network. To overcome this problem, a Multilevel-Gated Recurrent Unit (MGRU) is proposed for accurate prediction of traffic flow and congestion avoidance in smart cities. The main advantage of proposed M-GRU model is that it reduces errors while prediction by multilevel layers in softmax that passes information to adapt dynamic changes of road network. The data acquired based on weather conditions, number of vehicles improved the prediction process which utilized a few factors for prediction than previous prediction methods. Then the traffic density is estimated to predict TC on the road. Finally, the TC is predicted accurately by the proposed M-GRU model. The experimental results of the proposed method attained better results accuracy of 0.887, Mean Absolute Error (MAE) is 82.34 which is less than compared to existing method such as Convolutional Neural Network and Bidirectional-Long Short-Term Memory (Conv-Bi-LSTM).