A Short-Term Vehicle Speed Prediction Method Based on BiLSTM-RVFL Model

Abstract

The present invention discloses a short-term vehicle speed prediction method based on a BiLSTM-RVFL model, including: obtaining vehicle speed data and performing preprocessing to generate a vehicle speed data set; inputting the vehicle speed data set into a pre-constructed BiLSTM-RVFL model for short-term vehicle speed prediction; wherein, the construction process of the BiLSTM-RVFL model includes: performing mathematical modeling on BiLSTM: adding a sequentially connected forgetting gate, input gate, and output gate to the LSTM network based on the RNN network; constructing a BiLSTM network model through a forward LSTM network and a backward LSTM network; performing bidirectional training on the BiLSTM network model; the present invention can accurately predict the short-term vehicle speed and provides a valuable reference for the traffic department.

Description

Technical Field

[0001] The present invention relates to a short-term vehicle speed prediction method based on a BiLSTM-RVFL model, belonging to the technical field of intelligent transportation. Background Art

[0002] With the rapid development of China's economy and the advancement of urbanization, the concept of smart city has gradually become known to the public, among which intelligent transportation is an extremely important part of the smart city. In the process of promoting the development of intelligent transportation, the prediction of road traffic conditions has always been a very important topic. Accurate short-term vehicle speed prediction not only helps travelers with route planning and time arrangement, effectively reduces congestion during peak hours, but also provides strong support for formulating traffic management strategies.

[0003] Short-term vehicle speed prediction models can generally be divided into statistical models, parametric models, and non-parametric models. The representative model in statistical learning models is the historical average model (HA). This method mainly infers the change trend according to the laws of statistical data, and the calculation is simple and fast, but it cannot well adapt to time characteristics. Common parametric models include autoregressive integrated moving average model (ARIMA), which determines parameters through time series data and then predicts based on the regression function. For non-linear data, this model has certain limitations. Non-parametric models include support vector machine (SVR) and Bayesian network model, etc. This method discovers the laws of vehicle speed changes from historical data, can well reflect the non-linearity of vehicle speed, but it is difficult to mine its characteristics for long sequence data, thus affecting the prediction accuracy.

[0004] Deep learning models have been widely applied to the field of traffic prediction. Recurrent neural network (RNN) and long short-term memory network (LSTM) can effectively utilize the self-recurrent mechanism to capture the characteristics of vehicle speed. In recent years, the combined model of LSTM and convolutional neural network (CNN) has been widely used. According to existing research, the hybrid model often has higher prediction accuracy than the single model. Due to its structural characteristics, deep learning is prone to falling into local optimum in operation, thus affecting the prediction accuracy. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a short-term vehicle speed prediction method based on a BiLSTM-RVFL model, which can accurately predict the short-term vehicle speed and provide a valuable reference for the traffic department.

[0006] To achieve the above purpose, the present invention is implemented by the following technical solutions:

[0007] In the first aspect, the present invention provides a short-term vehicle speed prediction method based on a BiLSTM-RVFL model, including:

[0008] Obtain vehicle speed data and perform preprocessing to generate a vehicle speed data set;

[0009] Input the vehicle speed data set into the pre-constructed BiLSTM-RVFL model for short-term vehicle speed prediction;

[0010] Among them, the construction process of the BiLSTM-RVFL model includes:

[0011] Perform mathematical modeling on BiLSTM: On the basis of the RNN network, the LSTM network adds a forget gate, an input gate, and an output gate connected in sequence; construct a BiLSTM network model through the forward LSTM network and the backward LSTM network; perform bidirectional training on the BiLSTM network model;

[0012] Perform mathematical modeling on BiLSTM-RVFL: Capture vehicle speed data features through two trained BiLSTM models, input the vehicle speed data features into the RVFL layer through the fully connected layer, and generate enhanced nodes through the nonlinear transformation of the activation function of the RVFL layer; optimize the BiLSTM-RVFL model.

[0013] Optionally, the forget gate takes the input vector x t and the hidden state h t-1 to obtain the forget state f t through the sigmoid function σ, and multiply the cell state c t-1 and the forget state f t for output;

[0014] The input gate takes the input vector x t and the hidden state h t-1 to obtain the input state i t and the cell state respectively through the sigmoid function σ and the tanh function, t multiply the input state i and the cell state, and add the result to the output of the forget gate to obtain the cell state c t ;

[0015] The output gate takes the input vector x t and the hidden state h t-1 to obtain the output state o t through the sigmoid function σ, multiply the result obtained by taking the cell state c t through the tanh function with the output state o t to obtain the hidden state h t ;

[0016] Among them,

[0017] f t = σ(Wf · [h t-1 , x t + b f )

[0018] i t = σ(W i · [h t-1 , x t + b i )

[0019]

[0020]

[0021] o t = σ(W o [h t-1 , x t + b o )

[0022] h t = o t * tanh(C t )

[0023] where t represents the time step, W f , W i , W c , W o are weight vectors respectively, b f , b i , b c , b o are damage variables respectively, * is the Hadamard product of matrices, and t is the moment.

[0024] Optionally, the bidirectional training of the BiLSTM network model includes:

[0025] Obtain vehicle speed data and perform preprocessing and sequential forgetting encoding to generate a training set;

[0026] Calculate the cell states of the vehicle speed data in the training set through the forward LSTM network and the backward LSTM network respectively;

[0027] Perform fusion calculations based on the cell states of the forward LSTM network and the backward LSTM network respectively to obtain hidden states;

[0028] Concatenate the hidden states of the forward LSTM network and the backward LSTM network to obtain the final hidden state;

[0029] Obtain a loss value based on the final hidden state based on a preset loss function, and perform iterative training based on the loss value until the preset maximum number of iterations is reached or the loss value converges.

[0030] Optionally, the enhanced node is:

[0031]

[0032] where E j is the value of the j-th enhanced node, g(·) is the activation function, X is the input of the RVFL layer, is the weight of the label value of the j-th enhanced node, b j is the threshold of the j-th enhanced node, and the weight and the threshold b j are randomly generated.

[0033] Optionally, the BiLSTM-RVFL model is:

[0034]

[0035] where y j , E j , β j are the predicted output value, enhanced node, and weight vector of the j-th enhanced node respectively, L is the number of enhanced nodes, and d is the number of eigenvalue of the input data.

[0036] Optionally, the optimization of the BiLSTM-RVFL model includes:

[0037] Construct an optimization objective function:

[0038]

[0039] where ‖·‖ 2 is the L2 norm of the Euclidean space, N is the total number of samples, y i , t i are the predicted value and label value of the i-th sample, β is the output weight vector, Hβ = Y, H is the connection matrix between the input node and the enhanced node, and Y is the predicted value matrix; calculate β by the least square method: β = H + T, T is the label value matrix, and H + is the Moore-Penrose generalized inverse matrix of the connection matrix H.

[0040] Optionally, the preprocessing includes eliminating abnormal data from the vehicle speed data according to a preset vehicle speed range, filling in the missing data by Newton interpolation method, and then normalizing the filled data.

[0041] In a second aspect, the present invention provides a short-term vehicle speed prediction method based on a BiLSTM-RVFL model, and the device includes:

[0042] A data acquisition module, configured to acquire vehicle speed data and perform preprocessing to generate a vehicle speed data set;

[0043] A vehicle speed prediction module for inputting a vehicle speed data set into a pre - constructed BiLSTM - RVFL model for short - term vehicle speed prediction;

[0044] Among them, the construction process of the BiLSTM - RVFL model includes:

[0045] Mathematically model BiLSTM: On the basis of the RNN network, the LSTM network adds a forgetting gate, an input gate, and an output gate connected in sequence; construct a BiLSTM network model through a forward LSTM network and a backward LSTM network; perform bidirectional training on the BiLSTM network model;

[0046] Mathematically model BiLSTM - RVFL: Use two trained BiLSTM models to capture vehicle speed data features, input the vehicle speed data features into the RVFL layer through a fully - connected layer, and generate enhanced nodes through the non - linear transformation of the activation function of the RVFL layer; optimize the BiLSTM - RVFL model.

[0047] In a third aspect, the present invention provides a short - term vehicle speed prediction method based on a BiLSTM - RVFL model, including a processor and a storage medium;

[0048] The storage medium is used to store instructions;

[0049] The processor is used to operate according to the instructions to execute the steps of the above - mentioned method.

[0050] In a fourth aspect, the present invention provides a computer - readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above - mentioned method are implemented.

[0051] Compared with the prior art, the beneficial effects achieved by the present invention:

[0052] The short - term vehicle speed prediction method based on a BiLSTM - RVFL model provided by the present invention combines a bidirectional recurrent network and a random vector functional link network. The proposed BiLSTM - RVFL short - term vehicle speed prediction method, through experimental verification, is found to be able to overcome the defects of a single model. At the same time, compared with existing combined models, the prediction accuracy is further improved. The present invention can accurately predict short - term vehicle speeds and provides a valuable reference for the traffic department. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a structural diagram of the LSTM network provided in Embodiment 1 of the present invention;

[0054] Figure 2It is the structure diagram of the BiLSTM network provided by the first embodiment of the present invention;

[0055] Figure 3 It is the structure diagram of the BiLSTM-RVFL model provided by the first embodiment of the present invention;

[0056] Figure 4 It is the overall flowchart of the BiLSTM-RVFL model provided by the first embodiment of the present invention. Detailed implementation manners

[0057] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0058] Embodiment 1:

[0059] The embodiment of the present invention provides a short-term vehicle speed prediction method based on the BiLSTM-RVFL model, including the following steps:

[0060] 1. Obtain vehicle speed data and preprocess it to generate a vehicle speed data set;

[0061] 2. Input the vehicle speed data set into the pre-constructed BiLSTM-RVFL model for short-term vehicle speed prediction;

[0062] The following mainly introduces the construction process of the BiLSTM-RVFL model:

[0063] S1. Conduct mathematical modeling on BiLSTM:

[0064] S11. Add a forget gate, an input gate, and an output gate connected in sequence to the LSTM network on the basis of the RNN network;

[0065] Since the traditional RNN can only mine the features of some sequences, once the sequence is too long, the prediction accuracy will be significantly reduced. Therefore, the input gate, output gate, and forget gate are respectively added to the LSTM network on the basis of the RNN. Through these three gates, the LSTM network judges the input information and solves the problem of long sequence dependence in the RNN.

[0066] As Figure 1 shown, it is the structure diagram of the LSTM network;

[0067] The forget gate multiplies the input vector x t at time t and the hidden state h t-1 at time t-1 through the sigmoid function σ to obtain the forget state f t at time t, and multiplies the cell state c t-1 at time t-1 and the forget state f t at time t for output;

[0068] The input gate takes the input vector x at time step t t and the hidden state h at time step t - 1 t-1 and respectively obtains the input state i at time step t through the sigmoid function σ and the tanh function t and the cell state Multiply the input state i at time step t t and the cell state and then add the result to the output of the forget gate to obtain the cell state c t ;

[0069] The output gate takes the input vector x at time step t t and the hidden state h at time step t - 1 t-1 and obtains the output state o through the sigmoid function σ t , multiplies the result obtained by applying the tanh function to the cell state c t with the output state o t to obtain the hidden state h t ;

[0070] Among them,

[0071] f t =σ(W f ·[h t-1 ,x t +b f )

[0072] i t =σ(W i ·[h t-1 ,x t +b i )

[0073]

[0074]

[0075] o t =σ(W o [h t-1 ,x t +b o )

[0076] h t =o t *tanh(C t )

[0077] Among them, t represents the time step, and W f , W i , W c , W o are weight vectors respectively, and bf , b i , b c , b o are damaged variables respectively, * is the Hadamard product of the matrix, and t is the time.

[0078] S12. Construct a BiLSTM network model through a forward LSTM network and a backward LSTM network;

[0079] As Figure 2 shown, it is the structure diagram of the BiLSTM network;

[0080] S13. Conduct bidirectional training on the BiLSTM network model; specifically including:

[0081] S131. Obtain vehicle speed data, perform preprocessing and sequential forgetting encoding to generate a training set;

[0082] S132. Calculate the cell states of the vehicle speed data in the training set through the forward LSTM network and the backward LSTM network respectively;

[0083] S133. Perform fusion calculation based on the cell states of the forward LSTM network and the backward LSTM network respectively to obtain the hidden state;

[0084] S134. Concatenate the hidden states of the forward LSTM network and the backward LSTM network to obtain the final hidden state;

[0085] S135. Obtain the loss value based on the final hidden state and a preset loss function, and perform iterative training according to the loss value until reaching the preset maximum number of iterations or the loss value converges.

[0086] S2. Conduct mathematical modeling on BiLSTM-RVFL: Capture the vehicle speed data features through two trained BiLSTM models, input the vehicle speed data features into the RVFL layer through a fully connected layer (Dense), and generate enhanced nodes through the nonlinear transformation of the activation function of the RVFL layer; Optimize the BiLSTM-RVFL model.

[0087] As Figure 3 shown, it is the structure diagram of the BiLSTM-RVFL model;

[0088] Generate enhanced nodes through the nonlinear transformation of the activation function of the RVFL layer:

[0089]

[0090] Among them, E j is the value of the jth enhanced node, g(·) is the activation function, X is the input of the RVFL layer, is the weight of the label value of the j-th enhanced node, b j is the threshold of the j-th enhanced node, weight and the threshold b j are randomly generated.

[0091] Therefore, the BiLSTM-RVFL model is as follows:

[0092]

[0093] where y j , E j , β j are the predicted output value, enhanced node, and weight vector of the j-th enhanced node respectively. L is the number of enhanced nodes, and d is the number of feature values of the input data.

[0094] Furthermore, optimizing the BiLSTM-RVFL model includes:

[0095] Constructing an optimization objective function:

[0096]

[0097] where ‖·‖ 2 is the L2 norm in the Euclidean space, N is the total number of samples, y i , t i are the predicted value and label value of the i-th sample respectively. β is the output weight vector, Hβ = Y, H is the connection matrix between the input nodes and the enhanced nodes, and Y is the predicted value matrix; β is calculated by the least squares method: β = H + T, T is the label value matrix, and H + is the Moore-Penrose generalized inverse matrix of the connection matrix H.

[0098] The above two preprocessings mainly involve removing abnormal data from the vehicle speed data according to a preset vehicle speed range, filling in the missing data by Newton interpolation method, and then normalizing the filled data.

[0099] As Figure 4 shown, it is the overall flowchart of the BiLSTM-RVFL model; the original vehicle speed data is preprocessed by Newton interpolation and normalization, and then the preprocessed data is input into the RVFL layer through two BiLSTM models for feature extraction, and the feature extraction is input into the RVFL layer through the fully connected layer (Dense) for prediction to obtain the final result output.

[0100] Verify the method of this embodiment:

[0101] Preprocess and extract features from the publicly available vehicle speed dataset collected on a certain road; according to the characteristics of the data, the first 90% is used as the training set and the last 10% is used as the test set. To improve the smoothness of the data, a sliding window of 6 is set. The preprocessed data is transformed into a three-dimensional tensor and input into the BiLSTM network. To prevent overfitting, a Dropout mechanism is applied after each layer of BiLSTM. Then, through the fully connected layer, the output is used as the input node of RVFL. The input node undergoes a non-linear transformation through the activation function to generate enhanced nodes, and the weights and thresholds are updated. The optimization function of the model uses the Adam function because it can design independent adaptive learning rates for different parameters and can accelerate the convergence speed of the network. The detailed hyperparameter settings are shown in Table 1:

[0102] Table 1

[0103]

[0104] To verify the effectiveness of the model proposed in the present invention, the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination R 2 _score are selected as the evaluation indicators in this paper. The defining formulas of the evaluation indicators are as follows:

[0105]

[0106]

[0107]

[0108] Among them, y i , are the actual value, predicted value, and average value respectively.

[0109] The model of this embodiment is experimentally compared with another 4 models on 3 indicators on the publicly available dataset, and the obtained result comparison is shown in Table 2:

[0110]

[0111] As can be seen from Table 2, compared with other prediction models, BiLSTM-RVFL has good performance in terms of root mean square error, mean absolute error, and coefficient of determination indicators, and the prediction accuracy has been further improved. The present invention can accurately predict the short-term vehicle speed situation and provides certain reference value for travelers.

[0112] Embodiment 2:

[0113] The embodiment of the present invention provides a short-term vehicle speed prediction device based on the BiLSTM-RVFL model. The device includes:

[0114] A data acquisition module for acquiring vehicle speed data and performing preprocessing to generate a vehicle speed data set;

[0115] A vehicle speed prediction module for inputting the vehicle speed data set into a pre-constructed BiLSTM-RVFL model for short-term vehicle speed prediction;

[0116] Among them, the construction process of the BiLSTM-RVFL model includes:

[0117] Performing mathematical modeling on BiLSTM: adding a sequentially connected forget gate, input gate, and output gate to the LSTM network based on the RNN network; constructing a BiLSTM network model through a forward LSTM network and a backward LSTM network; performing bidirectional training on the BiLSTM network model;

[0118] Performing mathematical modeling on BiLSTM-RVFL: capturing vehicle speed data features through two trained BiLSTM models, inputting the vehicle speed data features into the RVFL layer through a fully connected layer, and generating enhanced nodes through the activation function non-linear transformation of the RVFL layer; optimizing the BiLSTM-RVFL model.

[0119] Example 3:

[0120] Based on Example 1, an embodiment of the present invention provides a short-term vehicle speed prediction method based on a BiLSTM-RVFL model, including a processor and a storage medium;

[0121] The storage medium is used to store instructions;

[0122] The processor is used to operate according to the instructions to execute the steps of the above method.

[0123] Example 4:

[0124] Based on Example 1, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0125] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0126] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0127] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0129] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A short-term vehicle speed prediction method based on the BiLSTM-RVFL model, characterized in that, Including: Obtain vehicle speed data and perform preprocessing to generate a vehicle speed data set; Input the vehicle speed data set into a pre-constructed BiLSTM-RVFL model for short-term vehicle speed prediction; Among them, the construction process of the BiLSTM-RVFL model includes: Perform mathematical modeling on BiLSTM: On the basis of the RNN network, the LSTM network adds a forgetting gate, an input gate, and an output gate connected in sequence; construct a BiLSTM network model through a forward LSTM network and a backward LSTM network; perform bidirectional training on the BiLSTM network model; Perform mathematical modeling on BiLSTM-RVFL: Use two trained BiLSTM models to capture vehicle speed data features, input the vehicle speed data features into the RVFL layer through a fully connected layer, and generate enhanced nodes through the activation function non-linear transformation of the RVFL layer; optimize the BiLSTM-RVFL model; Among them, the enhanced node is: Among them, E j is the value of the j-th enhanced node, g(·) is the activation function, X is the input of the RVFL layer, is the weight of the label value of the j-th enhanced node, b j is the threshold of the j-th enhanced node, the weight and the threshold b j are randomly generated; The BiLSTM-RVFL model is: where y j , E j , β j are the predicted output value, boosting node, and weight vector of the j-th boosting node respectively, L is the number of boosting nodes, and d is the number of feature values of the input data; The optimization of the BiLSTM-RVFL model includes: Construct an optimization objective function: where, ‖·‖ 2 is the L2 norm of the Euclidean space, N is the total number of samples, y i , t i are the predicted value and the label value of the i-th sample, β is the output weight vector, Hβ = Y, H is the connection matrix of the input nodes and the enhanced nodes, Y is the predicted value matrix; β is calculated by the least squares method: β = H + T, T is the label value matrix, H + is the Moore-Penrose generalized inverse matrix of the connection matrix H.

2. The short-term vehicle speed prediction method based on the BiLSTM-RVFL model according to claim 1, wherein The forgetting gate takes the input vector x t and the hidden state h t-1 to obtain the forgetting state f through the sigmoid function σ t , and multiplies the cell state c t-1 and the forgetting state f t for output; The input gate takes the input vector x t and the hidden state h t-1 to obtain the input state i and the cell state through the sigmoid function σ and the tanh function respectively t and the cell state Multiply the input state i t and the cell state and add the result to the output of the forget gate to obtain the cell state c t ; The output gate takes the input vector x t and the hidden state h t-1 to obtain the output state o through the sigmoid function σ t , and multiplies the cell state c t obtained through the tanh function with the output state o t to get the hidden state h t ; Among them, f t = σ(W f · [h t-1 , x t + b f ) i t = σ(W i · [h t-1 , x t + b i ) o t = σ(W o [h t-1 , x t + b o ) h t = o t *tanh(C t ) where t represents the time step, and W f , W i , W c , W o are weight vectors respectively, and b f , b i , b c , b o are corrupted variables respectively, * represents the Hadamard product of matrices, and t represents the time instant.

3. A short-term vehicle speed prediction method based on the BiLSTM-RVFL model according to claim 1, characterized in that The bidirectional training of the BiLSTM network model includes: Obtain vehicle speed data and perform preprocessing and sequential forgetting encoding to generate a training set; Calculate the cell state of the vehicle speed data in the training set through the forward LSTM network and the backward LSTM network respectively; Perform fusion calculation based on the cell states of the forward LSTM network and the backward LSTM network respectively to obtain the hidden state; Perform concatenation based on the hidden states of the forward LSTM network and the backward LSTM network to obtain the final hidden state; Based on the final hidden state, obtain a loss value based on a preset loss function, and perform iterative training according to the loss value until the preset maximum number of iterations is reached or the loss value converges.

4. A short-term vehicle speed prediction method based on the BiLSTM-RVFL model according to claim 1 or 3, characterized in that The preprocessing includes removing abnormal data from the vehicle speed data according to a preset vehicle speed range, filling in the missing data by Newton interpolation method, and then normalizing the filled data.

5. A short-term vehicle speed prediction device based on the BiLSTM-RVFL model, characterized in that, The device is configured to execute the steps of the method according to any one of claims 1-4, and the device includes: A data acquisition module for obtaining vehicle speed data and performing preprocessing to generate a vehicle speed data set; A vehicle speed prediction module for inputting the vehicle speed data set into a pre-constructed BiLSTM-RVFL model for short-term vehicle speed prediction; Among them, the construction process of the BiLSTM-RVFL model includes: Perform mathematical modeling on BiLSTM: On the basis of the RNN network, the LSTM network adds a forgetting gate, an input gate, and an output gate connected in sequence; construct a BiLSTM network model through a forward LSTM network and a backward LSTM network; perform bidirectional training on the BiLSTM network model; Perform mathematical modeling on BiLSTM-RVFL: Use two trained BiLSTM models to capture vehicle speed data features, input the vehicle speed data features into the RVFL layer through a fully connected layer, and generate enhanced nodes through the activation function non-linear transformation of the RVFL layer; optimize the BiLSTM-RVFL model.

6. A short-term vehicle speed prediction device based on the BiLSTM-RVFL model, characterized in that, Including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-4.

Images