Traffic control methods, devices, media, and equipment based on covariate-assisted frequency domain

Abstract

This invention belongs to the field of traffic control and provides a traffic control method, device, medium, and equipment based on covariate-assisted frequency domain. The method includes: acquiring original urban traffic flow time-series data and decomposing it into trend and seasonal components; extracting historical traffic flow data and time information from the adjacent area of ​​the target location as covariates for the urban traffic flow data of the target location; concatenating the trend and seasonal components with the covariates to obtain concatenated trend and seasonal information, respectively; converting the concatenated seasonal information into time series segments and then into seasonal features in the frequency domain; directly converting the concatenated trend information into trend features in the frequency domain; obtaining corresponding traffic flow prediction values ​​based on the mapping relationship between the seasonal and trend features in the frequency domain and the predicted traffic flow values, and then summing them as the final prediction result of the future traffic flow of the target location to guide traffic control.

Description

Technical Field

[0001] This invention belongs to the field of traffic control, and particularly relates to a traffic control method, device, medium and equipment based on covariate-assisted frequency domain. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Traffic flow, as a key indicator of urban transportation systems, is crucial for accurate prediction in urban traffic management and planning. Effective traffic flow forecasting provides a scientific basis for traffic congestion early warning, road network optimization, and the rational allocation of public transportation resources. It not only helps improve road utilization efficiency and reduce traffic congestion but also optimizes urban transportation energy consumption and reduces environmental pollution. However, because traffic flow data varies with different factors such as weekdays, holidays, seasons, and weather, traditional time-series models and machine learning methods are not suitable for high-precision long-term predictions of real-world data. With the continuous development of deep learning, some urban traffic flow forecasting methods based on deep learning models can better uncover patterns and characteristics in traffic data, improving prediction accuracy and precision.

[0004] Current methods for predicting urban traffic flow based on deep learning models still have the following problems:

[0005] (1) Model complexity and running efficiency: Deep learning models such as Transformer, CNN and RNN have powerful modeling capabilities, but they are not suitable for processing long-term sequences and large-scale traffic flow data due to their complex model structure and long computation time.

[0006] (2) Challenges of Time Series Decomposition: Traditional methods of directly using the original series for urban traffic flow prediction may lead to unclear internal characteristics of time series data. Although the simple moving average method is often used for data decomposition, this method is often unsatisfactory in terms of decomposition results.

[0007] (3) Intrinsic Dependencies in Time Series: Time series prediction aims to resolve the data relationships between different time steps. However, unlike the explicit semantics of words in natural language, individual time steps in a time series often lack explicit semantic meaning. The direct time-point dependencies established by existing deep learning models are difficult to fully reflect the intrinsic dependency structure of time series.

[0008] (4) Lack of covariate information: In time series analysis, covariate information refers to external information that may affect the target variable in the time series. Current forecasting methods often fail to fully utilize this valuable covariate information, thus affecting the accuracy and reliability of the forecast. Summary of the Invention

[0009] To address the aforementioned issues, this invention provides a traffic control method, apparatus, medium, and device based on covariate-assisted frequency domain. This method combines trend similarity maps with spatial relationship maps to learn associated features, and utilizes sequence features and associated features together to predict traffic flow, thereby improving the effectiveness of traffic flow prediction.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] The first aspect of the present invention provides a traffic control method based on covariate-assisted frequency domain.

[0012] In one or more embodiments, a traffic control method based on covariate-assisted frequency domain is provided, comprising:

[0013] Obtain the original time-series data of urban traffic flow and decompose it into two parts: trend component and seasonal component.

[0014] Historical traffic flow data and time information of the adjacent area of ​​the target location are extracted from the original time series data of urban traffic flow and used as covariates of the urban traffic flow data of the target location.

[0015] The trend component and seasonal component are concatenated with the covariates of the urban traffic flow data of the target area to obtain the concatenated trend information and the concatenated seasonal information, respectively.

[0016] The spliced ​​seasonal information is converted into time series segments, and then into seasonal information in the corresponding frequency domain; the spliced ​​trend information is directly converted into trend information in the corresponding frequency domain; and then the seasonal and trend information in the frequency domain are subjected to corresponding frequency domain feature extraction.

[0017] Based on the mapping relationship between the seasonal and trend characteristics of the frequency domain and the traffic flow forecast, the corresponding traffic flow forecast is obtained. The two are then added together to obtain the final forecast result of the future traffic flow of the target area, so as to guide traffic control.

[0018] A second aspect of the present invention provides a traffic control device based on covariate-assisted frequency domain.

[0019] In one or more embodiments, a traffic control device based on covariate-assisted frequency domain includes:

[0020] The data decomposition module is used to acquire the original time-series traffic flow data of the city and decompose it into two parts: trend component and seasonal component.

[0021] The covariate extraction module is used to extract historical traffic flow data and time information of the adjacent area of ​​the target location from the original time series data of urban traffic flow, and use it as a covariate of the urban traffic flow data of the target location.

[0022] The information splicing module is used to splice the trend component and seasonal component with the covariates of the urban traffic flow data of the target area, respectively, to obtain the spliced ​​trend information and the spliced ​​seasonal information.

[0023] The frequency domain feature extraction module is used to convert the spliced ​​seasonal information into time series segments, and then into seasonal information in the corresponding frequency domain; to directly convert the spliced ​​trend information into trend information in the corresponding frequency domain; and then to extract corresponding frequency domain features from the seasonal and trend information in the frequency domain.

[0024] The traffic flow prediction module is used to obtain the corresponding traffic flow prediction value based on the mapping relationship between the seasonal and trend characteristics of the frequency domain and the traffic flow prediction value. The two are then added together to obtain the final prediction result of the future traffic flow of the target area, so as to guide traffic control.

[0025] A third aspect of the present invention provides a computer-readable storage medium.

[0026] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the traffic control method based on covariate-assisted frequency domain as described above.

[0027] A fourth aspect of the present invention provides an electronic device.

[0028] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the traffic control method based on covariate-assisted frequency domain as described above.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] This invention fully decomposes urban traffic flow data into trend and seasonal components, and separates the internal features of the urban traffic flow data. It effectively utilizes covariate information extracted from the raw data to assist in predicting traffic flow in target areas. The seasonal component of the data is first transformed into a time series segment, preserving the correlation information at the data sequence level. This effectively reduces computational complexity while better representing the changing patterns of the time series. Feature extraction is performed using a frequency domain multilayer perceptron, which more effectively learns global dependencies and retains clear pattern features. The frequency domain multilayer perceptron is then used to directly model the trend component of urban traffic flow. Future predictions are made based on the extracted historical features of the seasonal and trend components. The two prediction results are combined as the final prediction result of the model, thereby improving the prediction accuracy of the model for complex and ever-changing urban traffic flow data.

[0031] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0032] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0033] Figure 1 This is a schematic diagram of an electronic device according to an embodiment of the present invention;

[0034] Figure 2 This is a flowchart illustrating the traffic control method based on covariate-assisted frequency domain according to an embodiment of the present invention.

[0035] Figure 3 This is a schematic diagram of traffic control based on covariate-assisted frequency domain according to an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram illustrating the seasonal trend decomposition principle of an embodiment of the present invention;

[0037] Figure 5 This is a segmented structure diagram of an embodiment of the present invention;

[0038] Figure 6 This is a schematic diagram of a multilayer perceptron according to an embodiment of the present invention;

[0039] Figure 7 This is a schematic diagram of the traffic control device based on covariate-assisted frequency domain according to an embodiment of the present invention. Detailed Implementation

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0042] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0043] Reference Figure 1 A schematic diagram of an electronic device is provided. It should be noted that... Figure 1 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0044] like Figure 1 As shown, the electronic device 100 includes a central processing unit (CPU) 101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 102 or a program loaded from a storage section 108 into a random access memory (RAM) 103. The RAM 103 also stores various programs and data required for system operation. The CPU 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0045] The following components are connected to I / O interface 105: an input section 106 including a keyboard, mouse, etc.; an output section 107 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 108 including a hard disk, etc.; and a communication section 109 including a network interface card such as a local area network (LAN) card, modem, etc. The communication section 109 performs communication processing via a network such as the Internet. Drive 110 is also connected to I / O interface 105 as needed. Removable media 111, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 110 as needed so that computer programs read from them can be installed into storage section 108 as needed.

[0046] When the central processing unit 101 in the electronic device of this embodiment executes the program, it achieves the following: Figure 2 The steps in the traffic control method based on covariate-assisted frequency domain are shown.

[0047] Figure 2 This is a flowchart illustrating a traffic control method based on covariate-assisted frequency domain in an embodiment of the present invention.

[0048] The traffic control method based on covariate-assisted frequency domain in this embodiment defines the urban traffic flow prediction problem as being given a time series sample set with a backtracking window of L. and the sample set of information for the k covariates corresponding to the dataset. Predict the value of the target location at T future times.

[0049] Non-stationary data are characterized by changes in their statistical properties (such as mean, variance, etc.) and joint distribution over time. This characteristic impairs the predictability of time series data, necessitating the introduction of a normalization module to normalize the time series sample set. Sequence generation through stabilization preprocessing Will The data is input into the subsequent model for modeling, thereby reducing the non-stationarity of the time series and thus obtaining better predictability. This provides a more stable data distribution for the deep learning model. The normalization module is shown in formula (1).

[0050]

[0051] ∈ is the numerical stability factor, which is usually set to 1e-5 by default. t [X] and Var[X] represent the mean and standard deviation of the time series, as defined in formulas (2) and (3).

[0052]

[0053]

[0054] The subsequent model accepts the transformed data. As input, the normalization module reduces the distributional differences of each input time series, making the model's input more stable. The model predicts future values ​​based on the input and explicitly returns non-stationary attributes removed from the input data to the model output by applying the same statistics used in the normalization steps of formulas (2) and (3) to the model output at a symmetrical position, i.e., the reverse normalization step of the model output layer. Scaling and shifting are performed to generate the final input Y of the model, as shown in formula (4). Where E t [X] and Var[X] are defined in the same way as in formulas (2) and (3).

[0055]

[0056] like Figure 2 and Figure 3 The traffic control method based on covariate-assisted frequency domain shown in this embodiment may include:

[0057] S201, obtain the original time series data of urban traffic flow and decompose it into two parts: trend component and seasonal component.

[0058] In this embodiment, the STL (Seasonal and Trend Decomposition using Loess) method is used to decompose the original urban traffic flow time series data into two parts: a trend component and a seasonal component. STL is a powerful method for decomposing time series data, capable of decomposing a given time series into a seasonal component, a trend component, and a residual component. The advantages of STL lie in its flexibility, its ability to handle time series with any form of seasonality pattern, and its robustness to outliers. STL decomposition uses Loess smoothing (locally weighted regression) through an iterative process to fit the trend and seasonal components.

[0059] Given processed urban traffic flow data The STL decomposition algorithm is used to decompose it into a trend component T(t), a seasonal component S(t), and a residual component R(t), as shown in Equation (8).

[0060]

[0061] The main idea of ​​STL decomposition is to iteratively obtain the seasonal and trend components by alternating the use of a locally weighted regression (Loess) fitting process, while simultaneously calculating the residual term R(t). Figure 4 As shown, the specific process of STL decomposition is as follows:

[0062] a. Initialize STL parameters: Initialize the STL decomposition parameters, which includes determining the seasonality cycle length, trend, and seasonality Loess smoothing parameters.

[0063] b. Preliminary estimation of seasonal components: By applying Loess smoothing to the data points within each seasonal period of the time series, a preliminary estimate of the seasonal components S′(t) is obtained. Where, bandwidth... seasonal It is the bandwidth parameter that controls the seasonal Loess fit:

[0064]

[0065] c. Estimation of trend components: Subtract the initially estimated seasonal components from the original time series to obtain the deseasoned time series. Then, Loess smoothing is applied to the deseasoned time series to obtain an estimate of the trend component, T(t). Where, bandwidth... seasonal It is the bandwidth parameter that controls the seasonal Loess fit:

[0066]

[0067] d. Seasonal component updates: using detrended time series. The seasonal components are re-estimated, and Loess smoothing is applied again to the data points in each seasonal cycle to obtain the updated seasonal components S(t).

[0068] e. Calculation of residual components: The residual component R(t) is calculated, which is obtained by subtracting the trend component and the updated seasonal component from the original time series.

[0069]

[0070] f. Repeat steps b to e until the changes in the estimates of the seasonal and trend components in consecutive iterations are less than a preset threshold, indicating that convergence has been achieved.

[0071] Ultimately, the seasonal portions will be obtained. Trend Components and residual components As a result of STL decomposition, they together form the complete representation of the original time series.

[0072] It should be noted that in other embodiments, other existing methods may also be used to decompose the original urban traffic flow time series data. Those skilled in the art can choose the appropriate method based on the actual situation, which will not be described in detail here.

[0073] S202 extracts historical traffic flow data and time information of the adjacent area of ​​the target location from the original time series data of urban traffic flow, and uses it as a covariate of the urban traffic flow data of the target location.

[0074] In time series analysis, covariates refer to variables other than the main analytical target, which exhibit some correlation with it. Covariates can be broadly categorized into static covariates (variables that do not change over time) and time-varying covariates (variables that change over time). The original urban traffic flow data includes historical traffic flow data for the target area, the time information of the historical traffic flow data collection (including year, month, day, and time), and historical traffic flow data for areas other than the target area. Considering data availability, the time information of the collected data (including year, month, day, and time) and the historical traffic flow data for areas other than the target area are used as covariate information to assist in the prediction of the target area. Assuming there are k covariate information, a sample set of covariate information is constructed based on the dataset. Used for subsequent data modeling.

[0075] In the specific implementation process, the covariates of the urban traffic flow data of the target area are represented in tensor form.

[0076] S203, the trend component and seasonal component are concatenated with the covariates of the urban traffic flow data of the target area to obtain the concatenated trend information and the concatenated seasonal information respectively.

[0077] The sample set of seasonal component S(t) and covariate information from the decomposition results. By piecing them together, we get the assembled result. It is used for seasonal component feature extraction.

[0078] Input season =Concat(S(t),ε)(9)

[0079] S204 transforms the spliced ​​seasonal information into time series segments, and then into seasonal information in the corresponding frequency domain; it directly transforms the spliced ​​trend information into trend information in the corresponding frequency domain; and then extracts corresponding frequency domain features from the seasonal and trend information in the frequency domain.

[0080] Next, we will work on the Input. season Execute along the time dimension L as follows Figure 5 The segmentation process shown has a segment length of S, dividing the input modal components into segmented input structures. Where N is the number of segments. A total of k+1 historical traffic flow data and covariate historical data from target locations are used in the segmentation operation.

[0081] By segmenting, the original time series is converted into segments of several lengths. On the one hand, since the number of segments is much smaller than the number of points in the original time series, the computational burden of subsequent frequency domain feature extraction (such as multilayer perceptron feature extraction) is significantly reduced, and memory usage and computational complexity are significantly decreased. Furthermore, since the segmented structure is used as the input to the model, each segment retains the semantic information at the data sequence level, thereby ensuring that adjacent time points containing similar features can be effectively aggregated together, better reflecting the true dependencies inherent in the time series.

[0082] For example, Fourier transform can be used to convert the trend information into time series segments and spliced ​​data into corresponding frequency domain information.

[0083] In this embodiment, a multilayer perceptron is also used to extract corresponding frequency domain features from seasonal and trend information in the frequency domain.

[0084] As shown in formula (10), This is a trainable linear layer used to project the piecewise structure onto the model dimension, where PE represents the position encoding matrix. The positional encoding is represented by sine and cosine positional encoding, as shown in formulas (11) and (12). pos represents the position in the segment structure input, i represents the dimension in the positional encoding, and D represents the embedding dimension of the model.

[0085] X emb =Input season W mlp +PE (10)

[0086]

[0087]

[0088] For the input time series data X emb Perform a Fast Fourier Transform (FFT) to convert to the frequency domain.

[0089] X′ emb =FFT(X) emb (13)

[0090] like Figure 6 As shown, a frequency-domain MLP is constructed to transform X′. emb Further processing is performed, and the model follows the variable independence setting during the seasonal component modeling process. The frequency domain MLP consists of multiple fully connected (linear) layers and activation functions, designed to capture complex patterns in frequency domain features.

[0091] For the first hidden layer of a frequency-domain MLP, the input X′ emb After linear transformation:

[0092] Z (1) =X′ emb W (1) +b (1) (14)

[0093] in H1 is the weight matrix of the first hidden layer, and H1 is the dimension of the first hidden layer. It is the bias term, Z (1) It is the layer output before the activation function is applied.

[0094] Then, a non-linear activation function (such as ReLU) is applied to obtain the output of the first hidden layer:

[0095] X (1) =φ(Z (1) (15)

[0096] The second hidden layer receives the output X from the previous layer. (1) Perform linear transformations and nonlinear activations:

[0097] Z (2) =X (1) W (2) +b (2) (16)

[0098] in, H1 is the weight matrix of the second hidden layer, and H2 is the dimension of the second hidden layer. It is the bias term, Z (2) It is the layer output before the activation function is applied.

[0099] Apply a non-linear activation function:

[0100] X (2) =φ(Z (2) (17)

[0101] Then for X (2) Perform residual linking and reproject it back to the time domain using inverse Fourier transform.

[0102]

[0103] Finally, the seasonal component segmentation module for time-series data outputs feature extraction results for the seasonal components.

[0104] Throughout the frequency domain MLP modeling process, W (l) b (l)These are trainable parameters that are learned through training data. The activation function φ is typically chosen as ReLU or another non-linear function to introduce non-linearity, enabling the model to capture more complex data relationships.

[0105] The processed time-series segments are transformed from their time-domain representation to their frequency-domain representation using Fourier transform. A frequency-domain Multilayer Perceptron (MLP) is then used to model the transformed frequency-domain data. Compared to time-domain MLP, frequency-domain MLP, by modeling in the frequency domain, can obtain a global perspective of the signal, more effectively learn global dependencies, and preserve clear pattern features, thus achieving better prediction results.

[0106] For trend components The trend component T(t) and the sample set of covariate information in the decomposition results are used to... Piecing them together, the result Used for trend component feature extraction.

[0107] Input trend =Concat(T(t),ε) (18)

[0108] Then, the model is embedded and the position is encoded in the same way, as shown in formula (19). For a trainable linear layer, PE represents the position encoding matrix. Position coding is represented by sine and cosine position coding, and the formula is the same as the sine and cosine position coding formula above.

[0109] X emb =Input trend W mlp +PE (19)

[0110] Then, the input time series data X... emb Perform a Fast Fourier Transform (FFT) to convert to the frequency domain.

[0111] X′ emb =FFT(X) emb (20)

[0112] Next, the constructed frequency domain MLP is used to process the transformed X′. emb Further processing will be carried out.

[0113] For the first hidden layer of a frequency-domain MLP, the input X′ emb After linear transformation:

[0114] Z (1) =X′ emb W (1) +b(1) (twenty one)

[0115] in H1 is the dimension of the first hidden layer. It is the bias term, Z (1) It is the layer output before the activation function is applied.

[0116] Then, a non-linear activation function (such as ReLU) is applied to obtain the output of the first hidden layer:

[0117] X (1) =φ(Z (1) ) (twenty two)

[0118] The second hidden layer receives the output X from the previous layer. (1) Perform linear transformations and nonlinear activations:

[0119] Z (2) =X (1) W (2) +b (2) (twenty three)

[0120] in, H2 is the dimension of the second hidden layer. It is the bias term, Z (2) It is the layer output before the activation function is applied.

[0121] Apply a non-linear activation function:

[0122] X (2) =φ(Z (2) ) (twenty four)

[0123] Then for X (2) Perform residual linking and reproject it back to the time domain using inverse Fourier transform.

[0124]

[0125] The final time series data trend component feature extraction module outputs the feature extraction results for the trend components.

[0126] S205. Based on the mapping relationship between the seasonal and trend characteristics of the frequency domain and the traffic flow forecast, the corresponding traffic flow forecast is obtained. The two are then added together to obtain the final forecast result of the future traffic flow of the target area, so as to guide traffic control.

[0127] Extracted seasonal and trend component features and First, a flattening operation is performed, then the data is projected onto specified dimensions, and the sum of the two projections is used as the model's final prediction result. This process is shown in formulas (25)-(29).

[0128]

[0129]

[0130]

[0131]

[0132] Output the model results Scaling and shifting are performed to fill in the non-stationary information removed in the initial normalization module, generating the final input Y of the model. The specific process is shown in formula (29). Y is the final prediction result of the model for urban traffic flow data, where E t [X] and Var[X] are defined in the same way as in formulas (2) and (3).

[0133]

[0134] The prediction results are evaluated using two evaluation indicators, MSE and MAE, as shown in equations (30) and (31).

[0135]

[0136]

[0137] Where y represents the actual value of the time series. This represents the predicted value, and T is the length of the predicted sequence.

[0138] The invention method was compared with three other methods: Temporal Convolutional Neural Network (TCN), Gated Recurrent Unit (GRU), and Transformer Neural Network. The invention method achieved the lowest scores in both metrics, ranking first among the compared methods, thus demonstrating the effectiveness of our method.

[0139] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including functions for executing... Figure 2The program code for the method shown. In such an embodiment, the computer program can be downloaded and installed from a network via communication section 109, and / or installed from removable medium 111. When the computer program is executed by central processing unit 101, it performs the various functions defined in the apparatus of this application.

[0140] in, Figure 2 The computer program instructions corresponding to the method shown may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in the process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0141] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0142] Figure 7 This is a schematic diagram of a traffic control device based on covariate-assisted frequency domain in an embodiment of the present invention. This embodiment is similar to... Figure 2 Corresponding to the traffic control method based on covariate-assisted frequency domain, such as Figure 7 As shown, the traffic control device based on covariate-assisted frequency domain in this embodiment may include:

[0143] The data decomposition module 701 is used to acquire the original time-series traffic flow data of the city and decompose it into two parts: trend component and seasonal component.

[0144] The covariate extraction module 702 is used to extract historical traffic flow data and time information of the adjacent area of ​​the target location from the original urban traffic flow time series data, and use it as a covariate of the urban traffic flow data of the target location.

[0145] The information splicing module 703 is used to splice the trend component and the seasonal component with the covariates of the urban traffic flow data of the target area, respectively, to obtain the spliced ​​trend information and the spliced ​​seasonal information.

[0146] The frequency domain feature extraction module 704 is used to convert the spliced ​​seasonal information into time series segments, and then into seasonal information in the corresponding frequency domain; to directly convert the spliced ​​trend information into trend information in the corresponding frequency domain; and then to extract corresponding frequency domain features from the seasonal and trend information in the frequency domain.

[0147] The traffic flow prediction module 705 is used to obtain the corresponding traffic flow prediction value based on the mapping relationship between the seasonal and trend characteristics of the frequency domain and the traffic flow prediction value, and then add the two together as the final prediction result of the future traffic flow of the target area to guide traffic control.

[0148] Figure 7 The specific implementation process of modules 705-705 in the traffic control device based on covariate-assisted frequency domain shown is as follows: Figure 2 The specific implementation process of steps S201 to S205 is the same, and will not be described in detail here.

[0149] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A traffic control method based on covariate-assisted frequency domain, characterized in that, include: Obtain the original time-series data of urban traffic flow and decompose it into two parts: trend component and seasonal component. Historical traffic flow data and time information of the adjacent area of ​​the target location are extracted from the original time series data of urban traffic flow and used as covariates of the urban traffic flow data of the target location. The trend component and seasonal component are concatenated with the covariates of the urban traffic flow data of the target area to obtain the concatenated trend information and the concatenated seasonal information, respectively. The spliced ​​seasonal information is converted into time series segments, and then into seasonal information in the corresponding frequency domain. The spliced ​​trend information is directly converted into the corresponding frequency domain trend information; then, the seasonal information and trend information in the frequency domain are subjected to corresponding frequency domain feature extraction. Based on the mapping relationship between the seasonal and trend characteristics of the frequency domain and the traffic flow forecast, the corresponding traffic flow forecast is obtained. The two are then added together to obtain the final forecast result of the future traffic flow of the target area, so as to guide traffic control.

2. The traffic control method based on covariate-assisted frequency domain as described in claim 1, characterized in that, The STL method was used to decompose the original urban traffic flow time series data into two parts: trend component and seasonal component.

3. The traffic control method based on covariate-assisted frequency domain as described in claim 1, characterized in that, The covariates of the urban traffic flow data for the target area are represented in tensor form.

4. The traffic control method based on covariate-assisted frequency domain as described in claim 1, characterized in that, The Fourier transform is used to convert the time series segmentation and splicing trend information into corresponding frequency domain information.

5. The traffic control method based on covariate-assisted frequency domain as described in claim 1, characterized in that, Multilayer perceptron is used to extract corresponding frequency domain features from seasonal and trend information in the frequency domain.

6. A traffic control device based on covariate-assisted frequency domain, characterized in that, include: The data decomposition module is used to acquire the original time-series traffic flow data of the city and decompose it into two parts: trend component and seasonal component. The covariate extraction module is used to extract historical traffic flow data and time information of the adjacent area of ​​the target location from the original time series data of urban traffic flow, and use it as a covariate of the urban traffic flow data of the target location. The information splicing module is used to splice the trend component and seasonal component with the covariates of the urban traffic flow data of the target area, respectively, to obtain the spliced ​​trend information and the spliced ​​seasonal information. The frequency domain feature extraction module is used to convert the spliced ​​seasonal information into time series segments, and then into seasonal information in the corresponding frequency domain; to directly convert the spliced ​​trend information into trend information in the corresponding frequency domain; and then to extract corresponding frequency domain features from the seasonal and trend information in the frequency domain. The traffic flow prediction module is used to obtain the corresponding traffic flow prediction value based on the mapping relationship between the seasonal and trend characteristics of the frequency domain and the traffic flow prediction value. The two are then added together to obtain the final prediction result of the future traffic flow of the target area, so as to guide traffic control.

7. The traffic control device based on covariate-assisted frequency domain as described in claim 6, characterized in that, In the data decomposition module, the STL method is used to decompose the original urban traffic flow time series data into two parts: trend component and seasonal component.

8. The traffic control device based on covariate-assisted frequency domain as described in claim 6, characterized in that, In the frequency domain feature extraction module, a multilayer perceptron is used to extract corresponding frequency domain features from seasonal and trend information in the frequency domain.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the traffic control method based on covariate-assisted frequency domain as described in any one of claims 1-5.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the traffic control method based on covariate-assisted frequency domain as described in any one of claims 1-5.

Images