Short-Term Traffic Flow Prediction Based on a K-Nearest Neighbor and Bidirectional Long Short-Term Memory Model

In the previous research on traffic flow prediction models, most of models mainly studied time series flow, and spatial correlation was not fully considered. To solve this problem, paper proposes a method to predict spatio-temporal characteristics short-term by combining k-nearest neighbor algorithm bidirectional long memory network model. By selecting real-time data observed high-speed roads in United Kingdom, K-nearest is used spatially screen station determine points with high then input BILSTM model for prediction. The experimental results show that compared SVR, LSTM, GRU, KNN-LSTM, CNN-LSTM proposed has better accuracy, its performance been improved 77%, 19%, 18%, 22%, 13%, respectively. neighbor-bidirectional short-time shows performance.