Short-Term Traffic Flow Prediction Based on SSA-CNN-BiLSTM Combination Model
In order to solve the problem of urban road traffic congestion and provide auxiliary means for decision-making of intelligent transportation systems,a combination model of SSA-CNN-BiLSTM was constructed to predict short-term traffic flow,taking into account the nonlinear and temporal characteristics of short-term traffic flow.Firstly,the original traffic flow data was subjected to outlier cleaning,wavelet threshold denoising,and normalization processing.Secondly,the SSA algorithm was used to iteratively optimize the three hyperparameters of the number of hidden layer units,initial learn-ing rate,and L2 regularization coefficient in the CNN-BiLSTM composite network.Finally,the opti-mal hyperparameter combination obtained from the search was input into the constructed combination network for training and prediction.The experimental results showed that compared with PSO and GWO algorithm,SSA algorithm had a faster convergence speed and stronger global optimization abili-ty in the process of network hyperparameter optimization.Compared with the three comparison models of CNN-BiLSTM,BiLSTM,and LSTM,the SSA-CNN-BiLSTM combination model showed a reduction of 5.46%,12.78%,and 20.38%in RMSE,and a decrease of 0.49%,2.24%,and 3.11%in MAPE respectively at a 5-minute time scale;and under the 15-minute time scale division the RMSE of SSA-CNN-BiLSTM model was decreased by 9.70,28.42,and 41.18,while the MAPE was decreased by 0.50%,1.98%,and 2.59%,respectively.Research has shown that the accuracy and stability of the short-term traffic flow prediction model of SSA-CNN-BiLSTM have been improved compared to existing algorithms,and road traffic conditions can be improved by providing more accurate short-term traffic travel information.
国家科技期刊平台
NSTL
万方数据
维普
