Optimized Structure of the Traffic Flow Forecasting Model With a Deep Learning Approach
Citations
1,802 citations
537 citations
385 citations
339 citations
Cites methods from "Optimized Structure of the Traffic ..."
...In addition, three prevailing deep learning NNs (i.e., LSTMs, DCNNs, and SAEs) and a classical machine learning method (SVM) were compared with the SRCNs for the same dataset....
[...]
...The structures of other methods (LSTMs and SAEs) are established according to their papers....
[...]
...The results of the SRCNs, LSTMs, SAEs, DCNNs, and SVM are listed in Figure 10 and Table 3....
[...]
...In this section, we employ traffic speed data from Beijing, China, to evaluate our model—SRCNs— and compare them with other deep NNs, including LSTMs [36], SAEs [35], DCNNs, and SVM....
[...]
...In this section, we compare SRCNs with four other algorithms (LSTMs, SAEs, DCNNs, and SVM) in terms of short term prediction....
[...]
310 citations
References
16,717 citations
14,635 citations
"Optimized Structure of the Traffic ..." refers background or methods in this paper
...That is, the SAE-LM model with a 240-min sampling time, SoftSign function for the autoencoders, 3 × log2(N) hidden nodes, five hidden layers (four autoencoders), and Purelin function for the last hidden set generates the most accurate predicted results....
[...]
...We follow the practice of using the same activation function for f (·) and g(·), and Sigmoid, Tanh, LogSig, and SoftSign functions are set as Levels 1, 2, 3, and 4, respectively....
[...]
...The concept of deep learning is to use deep architectures (multiple layers of nonlinear processing units) to extract and transform the inherent features in the data from the lowest level to the highest level, and every continuous layer uses the output from the previous layer as input [4], [16]....
[...]
...Therefore, if a large volume of traffic data is adopted, we may be able to avoid many failures caused by assumptions and the accuracy of traffic prediction can be improved by learning the information and correlations hidden in the data [3], [4]....
[...]
...2) Design Factor ii: For the activation functions of the autoencoders, f (·) and g(·), the most commonly used transfer functions for deep structures are Sigmoid, Tanh, LogSig, and SoftSign [4], [34]....
[...]
7,767 citations
"Optimized Structure of the Traffic ..." refers methods in this paper
..., the collected traffic data) are real numbers, we use the squared loss function L(x, z) to measure the reconstruction error [14]....
[...]
...Deep learning is a type of machine learning method based on learning representations of data, and it has been successfully applied to assist in many fields, for instance, classification tasks, information processing, pattern recognition, and object detection [14], [15]....
[...]
6,790 citations
6,547 citations
"Optimized Structure of the Traffic ..." refers background in this paper
...Moreover, with widespread traffic sensing facilities and advancements in sensing technologies [11], [12], the amount of collected traffic flow data can be greatly increased....
[...]