NxLFTNet: NARX LSTM Forward Taylor Network for Traffic Forecasting

Due to the growing level of transportation systems, numerous social troubles arise and create serious hazards like air pollution, the greenhouse effect, and traffic congestion.Hence, Traffic forecasting is essential for providing numerous applications in smart city management like congestion avoidance, navigation guidance, and urban traffic control.Still, the traffic prediction is difficult, since each road network is greatly dependent on other networks in a temporally and spatially manner.For capturing the temporal and spatial dependency, the Nonlinear autoregressive models with exogenous inputs (NARX) -Long Short Term Memory (LSTM) Forward Taylor Network (NxLFTN)based traffic forecasting is proposed in this paper.The traffic network is considered as a graph, which contains the spatio-temporal data.The spatio-temporal data is high-dimensional and complex.To embed the data into a lowdimensionality region, the spatio-temporal embedding (STE) generator is employed.Moreover, the STE generator is used for predicting the daily embedding and the weekly embedding of the time related to the present traffic signal.At last, the traffic forecasting is carried out using the NxLFTNet.In addition, the Mean absolute percentage error (MAPE), Root Mean square error (RMSE), and Mean Absolute Error (MAE) are used to validate the effectiveness of the NxLFTNet.It offered the finest MAPE, RMSE, and MAE of 0.003, 0.046, and 0.04 when considering the data from the Contra Costa country in Greater Bay Area (GBA) of LargeST dataset for delay is 50 minutes.