A method, device and storage medium for predicting traffic flow on a highway section
By taking the abnormal detection probability as the characteristic of traffic flow and combining the Attention-BiLSTM model, the problems of low traffic flow prediction efficiency and poor real-time performance in the prior art are solved, and more accurate and real-time traffic flow prediction at highway sections are achieved.
Patent Information
- Application Number
- CN202311425412.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-30
- Publication Date
- 2024-12-31
- Estimated Expiration
- 2043-10-30
AI Technical Summary
The prior art methods for repairing abnormal detection results in the data preprocessing stage are inefficient and have poor real-time performance, and cannot effectively deal with traffic flow prediction errors caused by majeure factors such as weather and detector occlusion.
The abnormal detection probability is used as an inherent feature of traffic flow, and the probability characteristics are detected through historical data and abnormal paths. The traffic flow prediction is performed using the Attention-BiLSTM neural network model, and the pre-trained model is directly input to achieve real-time and accurate prediction.
It improves the real-time and accuracy of traffic flow prediction, reduces the dependence on abnormal detection results, avoids post-processing, and significantly improves the practicality of prediction application.
Smart Images
Figure CN117636624B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method, a device and a storage medium for predicting traffic flow on a highway section, and belongs to the field of intelligent traffic control. Background Art
[0002] The emerging Intelligent Transportation System (ITS) has established a real-time, accurate and efficient integrated transportation and management system based on advanced technology, and short-term traffic flow prediction plays an important role in it. With the continuous development of intelligent sensing technology, the detection and information extraction of road traffic operation conditions have gradually become complete, which provides strong data support for road short-term traffic flow prediction.
[0003] For travelers, real-time and accurate short-term traffic flow forecasts can provide a scientific theoretical basis for their travel decisions and reduce the time and economic costs of travel. For traffic management departments, the size of traffic flow can reflect the efficiency of road traffic and can intuitively show the changing trend of road traffic. Short-term traffic flow forecasts can effectively alleviate the imbalance of time and space resources and improve the overall traffic capacity of the road network.
[0004] However, for abnormal detection results caused by force majeure factors such as weather and detector occlusion, previous studies have used methods to repair abnormal detection results in the data preprocessing stage, which is a post-processing method with low efficiency and poor real-time performance. Summary of the invention
[0005] The purpose of the present invention is to provide a method, device and storage medium for predicting traffic flow on a highway section, aiming to no longer focus on the impact of the detection anomaly of a specific factor on the accuracy of traffic flow prediction, but to use the anomaly detection probability as an inherent feature of traffic flow, thereby achieving accurate prediction of traffic flow in future time periods based on historical data.
[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions:
[0007] In one aspect, the present invention provides a method for predicting traffic flow on a highway section, comprising:
[0008] Obtain information data from the highway gantry management system;
[0009] According to the acquired information data, the traffic flow time series data of the gantry inspection point on the expressway and the driving information of the vehicles on the expressway are obtained;
[0010] Restore the driving path of each vehicle according to the driving information to obtain a reconstructed path;
[0011] Determine the feasible and effective path for the vehicle based on the actual geographical location of the gantry;
[0012] Comparing the reconstructed path with the valid path to determine whether each reconstructed path is an abnormal detection path;
[0013] According to the judgment result of the anomaly detection path, the anomaly detection probability is calculated;
[0014] The anomaly detection probability and traffic flow time series data are input into a pre-trained traffic flow prediction model to obtain the highway section traffic flow.
[0015] In the present invention, pre-training can be performed based on historically known traffic flow data and corresponding traffic flow data sequences and anomaly detection probabilities, so that the model can infer traffic flow data for future periods based on historical traffic flow data and abnormal path detection probability characteristics. The traffic flow prediction model can use the Attention-BiLSTM neural network model. Compared with the ordinary LSTM model, BiLSTM model, BiGRU model and Attention-BiGRU model, the model prediction effect of the present invention is better.
[0016] It should be noted that the highway section mentioned in the present invention may refer to a certain highway gantry detection point location. Specifically, the highway section flow prediction may refer to the traffic flow prediction for a certain highway gantry detection point location during a certain period of time.
[0017] Optionally, in the present invention, the method further includes: preprocessing the acquired information data, including: acquiring the time, license plate and gantry number of the vehicle passing through the highway gantry in the information data, and deleting duplicate recorded information data. This can reduce the size of the data file to improve the efficiency of subsequent data processing.
[0018] Optionally, for any highway gantry detection point whose traffic flow is to be predicted, the traffic flow time series data is expressed as:
[0019]
[0020] Where X is a t×(2N+1) matrix, and each element x i,j (i∈(1,t),j∈(pN,p+N)) represents the traffic flow data of the highway gantry detection point at position j in time period i, p represents the number of the current highway gantry detection point for which the traffic flow is to be predicted, and N represents the number of upstream / downstream gantries of the selected current highway gantry detection point.
[0021] In the above scheme, the interval frequency of traffic flow time series data can be minute or hour level. When the frequency is daily, the traffic flow at the detection point in the corresponding period of the next day can be predicted based on the traffic flow in a certain period of multiple historical days and the abnormal detection probability of historical days.
[0022] Optionally, for the current highway gantry detection point of the traffic flow to be predicted, the number N of adjacent gantries selected at the upstream and downstream of the highway line that actually have a greater impact on the flow of the detection point is at least 2. Of course, it is also possible to select one gantry detection point upstream and downstream respectively, but if a large number of adjacent gantries are selected, the prediction result obtained will be more accurate. Considering that the farther the distance between the gantries, the smaller the mutual flow impact, and the complexity of the model calculation will increase, the number of upstream and downstream gantries is not the more the better.
[0023] Optionally, restoring the driving path of each vehicle according to the driving information to obtain a reconstructed path includes:
[0024] Filter all gantry record data of each vehicle in one day by date, and sort them by the time they passed the gantry;
[0025] The gantry number sequence corresponding to the time-sorted data of each vehicle is used as the representation of the reconstructed path.
[0026] Optionally, determining a feasible and effective path for the vehicle according to the actual geographical location of the gantry includes:
[0027] According to the geographical location of the gantry, the gantry numbers are arranged in sequence according to the driving direction, and the obtained highway gantry number sequence is used as the representation of the effective path.
[0028] Optionally, comparing the reconstructed path with a valid path to determine whether each reconstructed path is an abnormal detection path includes:
[0029] Determine whether the gantry number sequence corresponding to the reconstructed path is a subsequence of the highway gantry number sequence corresponding to the valid path; if not, the corresponding reconstructed path is an abnormal detection path;
[0030] The abnormality detection probability is calculated based on the judgment result of the abnormality detection path, and the formula is:
[0031]
[0032] Where P represents the anomaly detection probability, m represents the number of anomaly detection paths, and M represents the total number of trips of all vehicles, where a vehicle's trip in the same driving direction within a day is defined as one trip.
[0033] Optionally, the abnormal detection probability and traffic flow time series data are input into a pre-trained traffic flow prediction model to obtain the highway section traffic flow, including:
[0034] For the current highway gantry detection point whose traffic flow is to be predicted, two adjacent gantries upstream and two adjacent gantries downstream of the highway position are selected respectively, and the traffic flow sequence data of the selected upstream and downstream gantries, the traffic flow sequence data of the current highway gantry detection point, and the abnormal detection probability are used to form the input feature matrix Input of the traffic flow prediction model:
[0035]
[0036] In the formula, e t,d represents the probability of anomaly detection on day d;
[0037] The input feature matrix data is input into the traffic flow prediction model to obtain the highway section traffic flow in the next period corresponding to the current highway gantry detection point output by the traffic flow prediction model.
[0038] As mentioned above, optionally, the traffic flow sequence data is obtained by statistics according to minute-level time intervals;
[0039] The abnormal detection probability is a value calculated based on the vehicle driving information of each day;
[0040] In the input feature matrix Input, e 1,d ,e 1,d ,...,e t,d The value of is equal to , which is the average of the anomaly detection probabilities of multiple days before the prediction day. To improve the accuracy of the prediction results, the anomaly detection probability can be calculated for each highway separately. Of course, an anomaly detection probability can also be calculated for all highways every day.
[0041] In a second aspect, the present invention provides a highway section traffic flow prediction device, comprising:
[0042] A data acquisition module is configured to acquire information data in a highway gantry management system;
[0043] The data statistics module is configured to obtain the traffic flow time series data of the gantry detection point of the expressway and the driving information of the vehicles on the expressway according to the acquired information data;
[0044] A path reconstruction module is configured to restore the driving path of each vehicle according to the driving information to obtain a reconstructed path;
[0045] An effective path determination module is configured to determine a feasible effective path for the vehicle according to the actual geographical location of the gantry;
[0046] An abnormality judgment module is configured to compare the reconstructed path with the valid path to determine whether each reconstructed path is an abnormality detection path;
[0047] An anomaly detection probability calculation module is configured to calculate the anomaly detection probability according to the judgment result of the anomaly detection path;
[0048] And, the traffic flow prediction module is configured to input the abnormal detection probability and traffic flow time series data into a pre-trained traffic flow prediction model to obtain the highway section traffic flow.
[0049] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the highway section traffic flow prediction method as described in the first aspect.
[0050] Compared with the prior art, the present invention has the following beneficial effects:
[0051] (1) The present invention proposes a method for predicting traffic flow on a highway section. The anomaly detection probability is directly introduced into the input features of the traffic flow prediction model. Compared with the two prediction methods of directly predicting traffic flow without considering the anomaly detection situation and processing the anomaly detection results afterwards, the measures in the present invention further improve the real-time performance and accuracy of the method in actual engineering. There is no need to process the anomaly detection results afterwards, and the method is highly feasible.
[0052] (2) Short-term prediction can not only capture the volatility of traffic flow more effectively, but also make the presentation of traffic flow more real-time. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Figure 1 The figure is a schematic flow chart of an embodiment of a method for predicting traffic flow on a highway section according to the present invention;
[0054] Figure 2 Shown is a schematic flow chart of another embodiment of a method for predicting traffic flow on a highway section according to the present invention;
[0055] Figure 3 The figure shows a schematic diagram of the abnormal detection path judgment process in the method of the present invention;
[0056] Figure 4 Shown is a schematic diagram of the ETC gantry layout on the expressway section. DETAILED DESCRIPTION
[0057] The technical concept of the present invention is: in order to minimize the adverse effects of abnormal detection results caused by force majeure factors such as weather and detector being blocked on traffic flow prediction, unlike previous studies that repair abnormal detection results in the data preprocessing stage, the present invention no longer focuses on a specific abnormal detection result itself, but integrates the abnormal detection probability as an inherent parameter into the traffic flow prediction model, thereby achieving accurate prediction of traffic flow and supporting real-time prediction.
[0058] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.
[0059] Example 1
[0060] This embodiment introduces a method for predicting traffic flow on a highway section. Figure 1 , which includes:
[0061] Obtain information data from the highway gantry management system;
[0062] Based on the acquired information data, the traffic flow time series data of the gantry inspection point on the highway and the driving information of vehicles on the highway are obtained;
[0063] Restore the driving path of each vehicle according to the driving information to obtain a reconstructed path;
[0064] Determine the effective path available to vehicles based on the actual geographical location of the highway gantry;
[0065] Comparing the reconstructed path with the valid path to determine whether each reconstructed path is an abnormal detection path;
[0066] According to the judgment result of the anomaly detection path, the anomaly detection probability is calculated;
[0067] The anomaly detection probability and traffic flow time series data are input into a pre-trained traffic flow prediction model to obtain the highway section traffic flow.
[0068] This embodiment takes into account that the detection information of the ETC gantry on the highway may be inaccurate due to factors such as weather and detector failure. Therefore, the abnormal detection probability is introduced and used together with the historical traffic flow data as the input feature for future traffic flow prediction. The relationship between the input features and the output prediction results is learned through a deep network model, which can improve the accuracy of traffic flow prediction.
[0069] Example 2
[0070] In this embodiment, the location point of each highway ETC gantry is used as a detection point, and the detection point corresponding to one of the highway ETC gantry is used as the highway section of the traffic flow to be predicted, and the traffic flow prediction for the next period is performed for this detection point. Figure 2 , the implementation process is described in detail below.
[0071] 1. Data acquisition and processing of high-speed gantry management system
[0072] The information data of each ETC gantry record is obtained from the highway ETC gantry management system, and useless fields are deleted. Only the three fields of time, license plate, and gantry number are retained, thereby reducing the data file size, improving the efficiency of subsequent data processing, and deleting the repeated vehicle record data in the original data.
[0073] If each detection point detects traffic flow at a set frequency, then for each detection point, a traffic flow data sequence that changes over time can be obtained, which is expressed as follows:
[0074]
[0075] Considering the correlation between the traffic flow at each detection point and the traffic flow at its upstream and downstream detection points, in order to further improve the accuracy of subsequent prediction calculations, this embodiment expands the traffic flow data of each detection point in each time period. For the detection point to be predicted, the traffic flow of the two detection points upstream and downstream is included in the traffic flow data of the current detection point. The traffic flow data of the expanded detection point p is expressed as:
[0076]
[0077] The expanded X is a t×(2N+1) matrix, where each element x i,j (i∈(1,t),j∈(pN,p+N)) represents the traffic flow data of the highway gantry detection point at position j in time period i. Since two detection points are selected as associated points upstream and downstream of the detection point, N=2 in the above formula.
[0078] Since traffic flow has certain volatility and randomness, in order to effectively capture the volatility of traffic flow and better present the changes in traffic flow, this embodiment counts traffic flow data according to the commonly used minimum time interval of 5 minutes. Based on the traffic flow data sequence and abnormal detection probability of the above formula, the traffic flow data of the next 5 minutes at the detection point p can be predicted.
[0079] 2. Pre-training of traffic flow prediction model
[0080] The traffic flow prediction model of this embodiment is implemented using a deep learning model, specifically an Attention-BiLSTM model. Compared with the ordinary LSTM model, the ordinary BiLSTM model, the ordinary BiGRU model, and the Attention-BiGRU model, the obtained prediction effect can be better.
[0081] The training samples can select traffic flow data of multiple highway gantry inspection points in multiple known historical periods, as well as the traffic flow data sequence before the corresponding historical period, and the anomaly detection probability to train the deep learning neural network. The specific training process is not described in detail.
[0082] 3. Calculation of Anomaly Detection Probability
[0083] In this embodiment, reference Figure 2 As shown, in order to simplify the calculation, an abnormal detection probability is calculated for each day's ETC gantry detection situation. For the day of traffic flow to be predicted, the average abnormal detection probability of the previous days is used. Furthermore, the abnormal detection probability of each day can also be calculated for each highway, and their average value is selected as the abnormal detection probability of the predicted day. The specific calculation principle of the abnormal detection probability is introduced as follows.
[0084] Expressways are one-way, fully enclosed traffic environments. Generally, the same vehicle will not frequently pass through a specific inspection point in a short period of time. Therefore, the vehicle travel path can be reconstructed using the time, gantry number, license plate number and other information recorded by each ETC gantry in the information data of the expressway gantry management system.
[0085] like Figure 3 As shown, in this embodiment, when restoring the driving path of each vehicle, the following method is used:
[0086] Filter all gantry record data of each vehicle in one day by date, and sort them by the time they passed the gantry;
[0087] The gantry number sequence corresponding to the time-sorted data of each vehicle is used as the representation of the reconstructed path.
[0088] The specific method of this embodiment for determining a feasible effective path for a vehicle based on the actual geographical location of the gantry is as follows: according to the geographical location of the gantry, the gantry numbers are arranged in sequence according to the driving direction, and the obtained highway gantry number sequence is used as a representation of the effective path.
[0089] After determining the valid path and the reconstructed path, the reconstructed path is compared with the valid path to determine whether the gantry numbering sequence corresponding to the reconstructed path is a subsequence of the high-speed gantry numbering sequence corresponding to the valid path. If not, the corresponding reconstructed path is an abnormal detection path.
[0090] According to the judgment results of the anomaly detection path, the anomaly detection probability is calculated, and the formula is:
[0091]
[0092] Where P represents the anomaly detection probability, M represents the total number of trips of all vehicles, and m represents the number of anomaly detection paths in the total number of trips. A vehicle's trip in the same driving direction within a day is defined as one trip.
[0093] If the anomaly detection probability is calculated for a highway, it is only necessary to analyze the valid paths and reconstructed paths for the highway and calculate the corresponding anomaly detection probability. The number of trips considered corresponds to the number of trips of all vehicles passing through the corresponding highway.
[0094] 4. Prediction of traffic flow at detection points
[0095] In this embodiment, for the ETC gantry inspection point p for which traffic flow is to be predicted, the above analysis can obtain the traffic flow time series data before the prediction period and the average value of the abnormal detection probability for multiple days before the prediction date. The obtained average value of the abnormal detection probability and the traffic flow time series data are input into the pre-trained traffic flow prediction model to obtain the highway section traffic flow at the current inspection point in the next period.
[0096] For the current highway gantry detection point whose traffic flow is to be predicted, the corresponding traffic flow sequence data and anomaly detection probability constitute the input feature matrix Input of the traffic flow prediction model, which is expressed as:
[0097]
[0098] In the formula, e t,d represents the probability of abnormal detection on the dth day. In this embodiment, e 1,d ,e 1,d ,...,e t,d is equal to the value of , which is the average of the anomaly detection probabilities of multiple days before the prediction day.
[0099] The above input feature matrix Input is input into the traffic flow prediction model to obtain the next period of highway section traffic flow x corresponding to the current highway gantry detection point output by the traffic flow prediction model t+1,p .
[0100] The above theoretical speculations and practical proofs show that the present embodiment introduces the anomaly detection probability, which significantly improves the real-time and accuracy of its application in actual projects compared to the two prediction methods of directly predicting traffic flow without considering the anomaly detection situation and processing the anomaly detection results afterwards. It does not need to process the anomaly detection results afterwards and has strong feasibility.
[0101] Example 3
[0102] This embodiment provides a highway section traffic flow prediction device, which includes:
[0103] A data acquisition module is configured to acquire information data in a highway gantry management system;
[0104] The data statistics module is configured to obtain, based on the acquired information data, the time series data of the traffic flow of the expressway gantry and its upstream and downstream gantry, as well as the driving information of the vehicles passing through the expressway gantry;
[0105] A path reconstruction module is configured to restore the driving path of each vehicle according to the driving information to obtain a reconstructed path;
[0106] An effective path determination module is configured to determine a feasible effective path for the vehicle according to the actual geographical location of the gantry;
[0107] An abnormality judgment module is configured to compare the reconstructed path with the valid path to determine whether each reconstructed path is an abnormality detection path;
[0108] An anomaly detection probability calculation module is configured to calculate the anomaly detection probability according to the judgment result of the anomaly detection path;
[0109] And, the traffic flow prediction module is configured to input the abnormal detection probability and traffic flow time series data into a pre-trained traffic flow prediction model to obtain the highway section traffic flow.
[0110] Combination Figures 1 to 3 The highway section traffic flow prediction device can execute the highway section traffic flow prediction method introduced in the above-mentioned embodiment 1 or 2, and the specific functional implementation of each functional module is not repeated here.
[0111] Example 4
[0112] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the highway section traffic flow prediction method described in Embodiment 1 or 2 are implemented.
[0113] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0114] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes 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 a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0115] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0117] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A method for predicting traffic flow on a highway section, characterized in that: include: Obtain information data from the highway gantry management system; Based on the acquired information data, the traffic flow time series data of the gantry inspection point on the highway and the driving information of vehicles on the highway are obtained; Restore the driving path of each vehicle according to the driving information to obtain a reconstructed path; Determine the effective path available to vehicles based on the actual geographical location of the highway gantry; Comparing the reconstructed path with the valid path to determine whether each reconstructed path is an abnormal detection path; According to the judgment result of the anomaly detection path, the anomaly detection probability is calculated; Input the anomaly detection probability and traffic flow time series data into a pre-trained traffic flow prediction model to obtain the highway section traffic flow; in, For any highway gantry detection point whose traffic flow is to be predicted, the traffic flow time series data is expressed as: ， in, yes The matrix, each element Indicated in Time Traffic flow data of the gantry inspection point on the highway at the location, Indicates the number of the current highway gantry inspection point for which traffic flow is to be predicted. Indicates the number of upstream / downstream gantries of the selected current highway gantry inspection point; The restoring the driving path of each vehicle according to the driving information to obtain a reconstructed path includes: Filter all gantry record data of each vehicle in one day by date, and sort them by the time they passed the gantry; The gantry number sequence corresponding to the time-sorted data of each vehicle is used as the representation of the reconstructed path; Determining a feasible and effective path for the vehicle according to the actual geographical location of the gantry includes: According to the geographical location of the gantry, the gantry numbers are arranged in sequence according to the driving direction, and the obtained highway gantry number sequence is used as the representation of the effective path; The comparing the reconstructed path with the valid path to determine whether each reconstructed path is an abnormal detection path includes: Determine whether the gantry number sequence corresponding to the reconstructed path is a subsequence of the high-speed gantry number sequence corresponding to the valid path; if not, the corresponding reconstructed path is an abnormal detection path; The abnormality detection probability is calculated based on the judgment result of the abnormality detection path, and the formula is: ， In the formula, represents the probability of anomaly detection, represents the number of anomaly detection paths, It represents the total number of trips of all vehicles, where a vehicle's trip in the same driving direction within a day is defined as a trip; The anomaly detection probability and traffic flow time series data are input into a pre-trained traffic flow prediction model to obtain the highway section traffic flow, including: For the current highway gantry detection point whose traffic flow is to be predicted, two upstream adjacent gantries and two downstream adjacent gantries of the highway position are selected respectively, and the traffic flow sequence data of the selected upstream and downstream gantries, the traffic flow sequence data of the current highway gantry detection point, and the abnormal detection probability are used to form the input feature matrix of the traffic flow prediction model. : ， In the formula, Indicates Anomaly detection probability of the day; The input feature matrix data is input into the traffic flow prediction model to obtain the highway section traffic flow in the next period corresponding to the current highway gantry detection point output by the traffic flow prediction model.
2. The highway section traffic flow prediction method according to claim 1 is characterized in that: Also includes: Preprocessing the acquired information data includes: Obtain the time, license plate number and gantry number of the vehicle passing through the highway gantry in the information data; And, delete the information data of duplicate records.
3. The highway section traffic flow prediction method according to claim 1, characterized in that: The minimum value is 2.
4. The highway section traffic flow prediction method according to claim 1, characterized in that: The traffic flow sequence data is obtained by statistics at minute-level time intervals; The abnormal detection probability is a value calculated based on the vehicle driving information of each day; The input feature matrix middle, is equal to the value of , which is the average of the anomaly detection probabilities of multiple days before the prediction day.
5. A highway section traffic flow prediction device using the highway section traffic flow prediction method according to any one of claims 1 to 4, characterized in that: include: A data acquisition module is configured to acquire information data in a highway gantry management system; The data statistics module is configured to obtain the traffic flow time series data of the gantry detection point of the highway and the driving information of the vehicle on the highway according to the acquired information data; A path reconstruction module is configured to restore the driving path of each vehicle according to the driving information to obtain a reconstructed path; An effective path determination module is configured to determine a feasible effective path for the vehicle according to the actual geographical location of the gantry; An abnormality judgment module is configured to compare the reconstructed path with the valid path to determine whether each reconstructed path is an abnormality detection path; The anomaly detection probability calculation module is configured to calculate the anomaly detection probability according to the judgment result of the anomaly detection path; And, the traffic flow prediction module is configured to input the abnormal detection probability and traffic flow time series data into a pre-trained traffic flow prediction model to obtain the highway section traffic flow.
6. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the highway section traffic flow prediction method as claimed in any one of claims 1 to 4 are performed.