Short-term traffic flow prediction based on optimized deep learning neural network: PSO-Bi-LSTM

Traffic flow prediction is important for urban planning and traffic congestion alleviation as well as for intelligent traffic management systems. Due to the periodic characteristics and high fluctuation in short-term periods, it is difficult to accurately estimate future patterns in traffic flow on the urban road network. Thus, to forecast short-term traffic flow, a PSO-Bi-LSTM model based on the combination of Particle Swarm Optimization (PSO) and Bidirectional-Long Short-Term Memory (Bi-LSTM) neural network is developed in this paper. The PSO approach, which searches for the best parameters of a model on a global scale is used and nonlinear variable inertial weights are considered instead of linear weight. Additionally, the Bi-LSTM network prediction model is optimized using the PSO technique, which has the advantages of rapid convergence, high robustness, and large global search ability. To test the performance of the proposed model, traffic flow data has been collected from the Inner Ring Road, South Extension, Delhi, India. The performance of proposed PSO-Bi-LSTM model has been compared with other existing neural network models, e.g., Bi-LSTM, LSTM, Extreme Learning Machine (ELM), Gated Recurrent Unit (GRU), Wavelet Neural Network (WNN), Multilayer perceptron (MLP), and Autoregressive Integrated Moving Average (ARIMA). Experimental findings demonstrated that the proposed PSO-Bi-LSTM model has significantly outperformed the other models in terms of accuracy and stability.