A Multifeature Fusion Short-Term Traffic Flow Prediction Model Based on Deep Learnings

Abstract

Short-term traffic flow prediction is an important component of intelligent transportation systems, which can support traffic trip planning and traffic management. Although existing predicting methods have been applied in the field of traffic flow prediction, they cannot capture the complex multifeatures of traffic flows resulting in unsatisfactory short-term traffic flow prediction results. In this paper, a multifeature fusion model based on deep learning methods is proposed, which consists of three modules, namely, a CNN-Bidirectional GRU module with an attention mechanism (CNN-BiGRU-attention) and two Bidirectional GRU modules with an attention mechanism (BiGRU-attention). The CNN-BiGRU-attention module is used to extract local trend features and long-term dependent features of the traffic flow, and the two BiGRU-attention modules are used to extract daily and weekly periodic features of the traffic flow. Moreover, a feature fusion layer in the model is used to fuse the features extracted by each module. And then, the number of neurons in the model, the loss function, and other parameters such as the optimization algorithm are discussed and set up through simulation experiments. Finally, the multifeature fusion model is trained and tested based on the training and test sets from the data collected from the field. And the results indicate that the proposed model can better achieve traffic flow prediction and has good robustness. Furthermore, the multifeature fusion model is compared and analyzed against the baseline models with the same dataset, and the experimental results show that the multifeature fusion model has superior predictive performance compared to the baseline models.

Urban Traffic Flow Estimation System Based on Gated Recurrent Unit Deep Learning Methodology for Internet of Vehicles

“…Additive decomposed parts have not been employed as attributes in multivariate ML models, as far as researchers are aware of. It is recommended to use a CNN-Bidirectional GRU unit with an attention technique (CNN-BiGRU-attention) and two Bidirectional GRU modules, each with their own attention function, to create a multifeature fusion model using deep learning techniques [32]. Daily and weekly periodic characteristics of the traffic flow are extracted using the CNN-BiGRU-attention component, while local trend attributes are extracted using the CNN-BiGRU-attention component and long-term dependent characteristics are extracted using the two BiGRU-attention units.…”

“…Specifically, after the input layer, the first GRU is followed the input layer with input of period of sequence equals to 32. The output shape of this GRU layer is (32,150), accordingly, the number of trainable parameters is 68850. The next layer is the dense layer (fully connected layer), which here time-dependent, TimeDistributed-dense layer.…”

“…Note that the dropoutlayer has no trainable parameters. Now, the third GRU-layer of output shape (32,50) resulting in 22800 trainable parameters. TimeDistributed-layer is used after the third GRU-layer of the same size of the previous TimeDistributed-layers, in other words, there are 5100 trainable parameters, i.e., less number of weightings by 10000, this is due to that the input size is 50 (50-inputs ×100-neurons + 100-bias = 5100).…”

“…Better representations can be learned in BiLSTM with the help of potential future context chunks. However, the dimension of this layer is (32,200) which supports 160800 trainable parameters. Consequently, the fourth GRU-layer is used after the BiLSTM-layer with 50-units, the trainable parameters are 37800 parameters, when its input is 200.…”

Short-term traffic flow prediction of parallel roads based on transfer learning

“…In recent years, the spatial distribution of traffic flow of expressways and parallel roads is uneven [1]. Research on traffic flow forecasting can effectively solve such problems [2].…”