Highway Section Traffic Flow Prediction Method Based on Historical Data Situation Mining

Abstract

The present invention discloses a method for predicting traffic flow of a highway section based on historical data trends, which belongs to the field of data processing technology, including: establishing a historical traffic flow database of a highway section for one year, selecting traffic flow data of M×K minutes before the prediction moment from the historical traffic flow database to obtain a first matrix; selecting traffic flow data of the same time period of M×K minutes before the prediction moment to obtain a historical data set matrix; selecting traffic flow data of the same time period as the minutes after the prediction moment to obtain a prediction set matrix; calculating the Euclidean distance matrix of the first matrix and the historical data set matrix, and combining the prediction set matrix to obtain traffic flow prediction data, M is the time step, and K is the number of adjacent time points. The present invention has higher prediction accuracy when predicting traffic flow, and is applicable to different high-speed section traffic flow predictions at the same time, and has stronger applicability and scalability.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a method for predicting highway section traffic flow based on historical data situation mining. Background Art

[0002] The research on the prediction of highway traffic and the evaluation of road conditions has always been an important research topic in the field of transportation, which is of great significance to highway planning, toll decision-making, and people's travel. Traffic flow prediction has a history of several decades of research, but previous studies often focus on urban roads, and the research on highways is relatively insufficient. There are huge differences between highways and ordinary urban roads, and they cannot be discussed in the same way. Highways have the characteristics of high closure, wide coverage, more special spatiotemporal characteristics, and large fluctuations caused by the external environment. Previous traffic flow research can be roughly divided into two categories: one is a model based on statistical methods, including but not limited to the use of autoregressive moving average (ARIMA), Kalman filtering theory (KF) and adaptive spatiotemporal K-nearest neighbor model (adaptive-STKNN), which comprehensively considers the spatial characteristics of urban traffic based on parameters such as the number of adaptive neighbors, time window size, and spatiotemporal weights to predict traffic flow. The other is a model based on deep learning, which analyzes the characteristics of highway toll data, combines vehicle inspection data, and uses BP neural network to estimate the cross-sectional traffic volume of highways. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for predicting highway section traffic flow based on historical data trends, which solves the problem.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for predicting the traffic flow of a highway section based on historical data trends, comprising the following steps:

[0005] S1: Collect historical traffic flow data of highway sections and perform preprocessing to obtain a historical traffic flow database, wherein the time step of the data in the historical traffic flow database is M minutes;

[0006] S2: Obtain the prediction time to be predicted, and select the traffic flow data of M×K minutes before the prediction time from the historical traffic flow database to obtain the first matrix, where K is the number of set time points;

[0007] S3: Select the traffic flow data of the same time period as M×K minutes before the prediction time in the historical traffic flow database, and integrate and construct the historical data set matrix;

[0008] S4: According to the predicted time to be predicted, select the time corresponding to the predicted time from the historical traffic flow database. The traffic flow data of the same time period in minutes is used to obtain the prediction set matrix;

[0009] S5: Calculate the Euclidean distance between the first matrix and the historical data set matrix, and construct a Euclidean distance matrix according to the calculation results;

[0010] S6: Select the data with the smallest preset range in the Euclidean distance matrix and substitute it into the prediction set matrix to calculate and obtain the traffic flow prediction data.

[0011] The beneficial effects of the present invention are as follows: the present invention predicts the traffic flow of highway sections through historical traffic flow data, and integrates the influence of time factors and external environmental factors on traffic flow changes. Compared with traditional traffic flow prediction algorithms, the present invention has higher prediction accuracy, and is suitable for traffic flow prediction of highway sections of different types and different usage conditions, and has stronger applicability and scalability.

[0012] Furthermore: the expression of the first matrix is ​​as follows:

[0013] B=[x 1 ,x 2 ,…,x K-1 ,x K ]

[0014] Among them, B is the first matrix, x K is the historical traffic flow data at the prediction time, x K-1 is the historical traffic flow data M minutes before the prediction time, x 2 is the historical traffic flow data of M×(K-2) minutes before the prediction time, x 1 It is the historical traffic flow data of M×(K-1) minutes before the prediction time.

[0015] Further: the expression of the historical data set matrix is ​​as follows:

[0016]

[0017] Among them, C [n][] is the historical data set matrix, n is the number of days of the historical traffic flow database, is the historical traffic flow data of M×(K-1) minutes before the prediction time with day 1, is the historical traffic flow data of M×(K-2) minutes before the prediction time with day 1, is the historical traffic flow data M minutes before the prediction time on day 1, is the historical traffic flow data at the prediction time of day 1, is the historical traffic flow data of M×(K-1) minutes before the prediction time with n days, is the historical traffic flow data of M×(K-2) minutes before the prediction time of day n, is the historical traffic flow data M minutes before the prediction time on day n. is the historical traffic flow data at the prediction time with n days.

[0018] Further: the expression of the prediction set matrix is ​​as follows:

[0019]

[0020] Among them, D [n][] is the prediction set matrix, n is the number of days in the historical traffic flow database, is the historical traffic flow data M minutes after the prediction time on day 1, The historical traffic flow data is M×2 minutes after the prediction time of day 1. After the forecast time with a day number of 1 Minutes of historical traffic flow data, is the historical traffic flow data M minutes after the prediction time on day n, is the historical traffic flow data of M×2 minutes after the prediction time of day n, After the forecast time n days Minute-by-minute historical traffic data.

[0021] Further: the Euclidean distance between the first matrix and the historical data set matrix is ​​calculated, and its expression is as follows:

[0022]

[0023] Where Dis_Eculid is the Euclidean distance between the first matrix and the historical dataset matrix, x K is the Kth flow in the first matrix, is the Kth flow data on the nth day in the historical flow, and Z is the first Z groups of historical flow sequences with the smallest distance. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The figure is a flow chart of a highway section traffic flow prediction method based on historical data trends. DETAILED DESCRIPTION

[0025] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0026] Example 1

[0027] like Figure 1 As shown, it is a flow chart of a method for predicting highway section traffic flow based on historical data trends, which includes the following steps:

[0028] S1: Collect historical traffic flow data of highway sections and perform preprocessing to obtain a historical traffic flow database, wherein the time step of the data in the historical traffic flow database is M minutes;

[0029] S2: Obtain the prediction time to be predicted, and select the traffic flow data of M×K minutes before the prediction time from the historical traffic flow database to obtain the first matrix, where K is the number of set time points;

[0030] S3: Select the traffic flow data of the same time period as M×K minutes before the prediction time in the historical traffic flow database, and integrate and construct the historical data set matrix;

[0031] S4: According to the predicted time to be predicted, select the time corresponding to the predicted time from the historical traffic flow database. The traffic flow data of the same time period in minutes is used to obtain the prediction set matrix;

[0032] S5: Calculate the Euclidean distance between the first matrix and the historical data set matrix, and construct a Euclidean distance matrix according to the calculation results;

[0033] S6: Select the data with the smallest preset range in the Euclidean distance matrix and substitute it into the prediction set matrix to calculate and obtain the traffic flow prediction data.

[0034] In one embodiment of the present invention, in the process of finding the optimal common situation traffic flow, parameter setting is very important for reducing the prediction error. The preset parameters include the time window length M and the number of the nearest K, that is, the time step M of the data in the historical traffic flow database and the number of time points K set. By comparing the errors under different combinations, the combination with the smallest error is selected as: K = 12, M = 5, that is, in the historical traffic flow database, the Euclidean distances of the 12 5-minute (1-hour) flows adjacent to the prediction period are compared, and the 5 flow situations with the shortest Euclidean distance are selected as the prediction set D. Therefore, in this embodiment, the time step of the data in the historical traffic flow database is set to M = 5 minutes, the number of time points K is set to 12, the traffic flow data one hour before the prediction time is selected from the historical traffic flow database to obtain the first matrix, and the traffic flow data of the same time period half an hour after the prediction time is selected from the historical traffic flow database to obtain the prediction set matrix. At this time, it is the combination with the smallest prediction error, but it is not the only one, and it can be adjusted according to the specific highway environment.

[0035] In one embodiment of the present invention, a flow chart of a method for predicting highway section traffic flow based on historical data trends includes the following steps:

[0036] S1: Collect historical traffic flow data of highway sections and pre-process them to obtain a historical traffic flow database;

[0037] In one embodiment of the present invention, the historical traffic flow data of the expressway in the past year is collected and sorted, meaningless data, duplicate data, and incomplete data are removed, and the data is processed at a time interval of 5 minutes to obtain a historical traffic flow database;

[0038] Among them, traffic travel is time-periodic, and traffic data on different dates will show similar characteristics. If a sufficient amount of historical traffic data is guaranteed, the possibility of traffic data with similar traffic trends will be greater. In sufficient historical data, the accuracy of predicting the traffic during the target period by using certain rules to select more similar traffic data will be higher, but redundant data will increase the amount of calculation and search time, and historical data that is too old is of little reference significance for current traffic prediction. Therefore, it is necessary to ensure the richness of typical data as much as possible while reducing redundant data to reflect various traffic conditions. In order to ensure that as many similar traffic trends as possible can appear on special holidays, such as the Spring Festival, National Day and other major holidays, one year of historical traffic flow data is used as the historical traffic flow database.

[0039] The traffic data collected by devices such as checkpoints may have poor quality such as data loss and data distortion due to unavoidable factors such as signal transmission. Therefore, preprocessing is required.

[0040] S2: Get the prediction time to be predicted, and select the traffic flow data one hour before the prediction time from the historical traffic flow database to obtain the first matrix, which is expressed as follows:

[0041] B=[x 1 ,x 2 ,x 3 ,x 4 ,x 5 ,x 6 ,x 7 ,x 8 ,x 9 ,x 10 ,x 11 ,x 12 ]

[0042] Among them, B is the first matrix, x 12 is the historical traffic flow data at the prediction time, x 11 is the historical traffic flow data 5 minutes before the prediction time, x 10 is the historical traffic flow data 10 minutes before the prediction time, x 9 is the historical traffic flow data 15 minutes before the prediction time, x 8 is the historical traffic flow data 20 minutes before the prediction time, x 7 is the historical traffic flow data 25 minutes before the prediction time, x 6 is the historical traffic flow data 30 minutes before the prediction time, x 5 is the historical traffic flow data 35 minutes before the prediction time, x 4 is the historical traffic flow data 40 minutes before the prediction time, x 3 is the historical traffic flow data 45 minutes before the prediction time, x 2 is the historical traffic flow data 50 minutes before the prediction time, x 1 It is the historical traffic flow data 55 minutes before the prediction time.

[0043] S3: Select the traffic flow data of the same time period as one hour before the prediction time in the historical traffic flow database, and integrate and construct the historical data set matrix, which is expressed as follows:

[0044]

[0045] Among them, C [n][] is the historical data set matrix, n is the number of days in the historical traffic flow database, and here it is represented by n rows of data. The historical traffic flow data for 55 minutes before the prediction time on day 1. The historical traffic flow data for 50 minutes before the prediction time on day 1. The historical traffic flow data for 45 minutes before the prediction time on day 1. The historical traffic flow data for 40 minutes before the prediction time on day 1. The historical traffic flow data for 35 minutes before the prediction time on day 1. The historical traffic flow data for 30 minutes before the prediction time on day 1. The historical traffic flow data for 25 minutes before the prediction time on day 1. The historical traffic flow data for 20 minutes before the prediction time on day 1. The historical traffic flow data for the 15 minutes before the prediction time on day 1. The historical traffic flow data for the 10 minutes before the prediction time on day 1. The historical traffic flow data for 5 minutes before the prediction time on day 1. is the historical traffic flow data at the prediction time of day 1, is the historical traffic flow data 55 minutes before the prediction time of day n. is the historical traffic flow data 50 minutes before the prediction time of day n, is the historical traffic flow data 45 minutes before the prediction time of day n, is the historical traffic flow data 40 minutes before the prediction time of day n, The historical traffic flow data for 35 minutes before the prediction time of day n, is the historical traffic flow data 30 minutes before the prediction time of day n, is the historical traffic flow data 25 minutes before the prediction time of day n, is the historical traffic flow data 20 minutes before the prediction time of day n, is the historical traffic flow data 15 minutes before the prediction time of day n, is the historical traffic flow data 10 minutes before the prediction time of day n, is the historical traffic flow data 5 minutes before the prediction time of day n, is the historical traffic flow data at the prediction time with n days.

[0046] S4: According to the prediction time to be predicted, the traffic flow data of the same time period half an hour after the prediction time is selected from the historical traffic flow database to obtain the prediction set matrix, which is expressed as follows:

[0047]

[0048] Among them, D [n][] is the prediction set matrix, n is the number of days in the historical traffic flow database, and here it is represented by n rows of data. The historical traffic flow data for 5 minutes after the prediction time of day 1. The historical traffic flow data for 10 minutes after the prediction time on day 1. The historical traffic flow data for 15 minutes after the prediction time on day 1. The historical traffic flow data for 20 minutes after the prediction time on day 1. The historical traffic flow data for 25 minutes after the prediction time on day 1. The historical traffic flow data for 30 minutes after the prediction time of day 1. The historical traffic flow data for 35 minutes after the prediction time on day 1. is the historical traffic flow data 5 minutes after the prediction time of day n, is the historical traffic flow data 10 minutes after the prediction time of day n, is the historical traffic flow data 15 minutes after the prediction time on day n. is the historical traffic flow data 20 minutes after the prediction time of day n, is the historical traffic flow data 25 minutes after the prediction time on day n. is the historical traffic flow data 30 minutes after the prediction time of day n, The historical traffic flow data is 35 minutes after the prediction time on day n.

[0049] S5: Calculate the Euclidean distance between the first matrix and the historical data set matrix, and construct a Euclidean distance matrix according to the calculation results, wherein the expression for calculating the Euclidean distance between the first matrix and the historical data set matrix is ​​as follows:

[0050]

[0051] Where Dis_Eculid is the Euclidean distance between the first matrix and the historical dataset matrix, x K is the Kth flow in the first matrix, is the Kth flow data on the nth day in the historical flow, and Z is the first Z groups of historical flow sequences with the smallest distance.

[0052] S6: Select the data with the smallest preset range in the Euclidean distance matrix and substitute it into the prediction set matrix to calculate the traffic flow prediction data; in one embodiment of the present invention, select the first 5 groups of the smallest data in the Euclidean distance matrix and substitute them into the prediction set matrix to calculate the traffic flow prediction data, which can achieve higher accuracy and calculation efficiency.

[0053] The beneficial effects of the present invention are as follows: the present invention predicts the traffic flow of highway sections through historical traffic flow data, and integrates the influence of time factors and external environmental factors on traffic flow changes. Compared with traditional traffic flow prediction algorithms, the present invention has higher prediction accuracy, and is suitable for traffic flow prediction of highway sections of different types and different usage conditions, and has stronger applicability and scalability.

[0054] Example 2

[0055] In a specific embodiment of the present invention, a section of a highway in a certain city is selected as an object, and the specific data are all assumed data;

[0056] S1: Collect historical traffic flow data of highway sections for one year, perform preprocessing, remove meaningless data, duplicate data, and delete incomplete data to obtain a historical traffic flow database, as shown in Table 1 below, which is an example table of the data format of the historical traffic flow database:

[0057] Table 1

[0058]

[0059]

[0060] Among them, total is the total traffic flow, kc is the passenger car traffic flow, hc is the truck traffic flow, and qt is the traffic flow of other vehicles except passenger cars and trucks.

[0061] S2: Obtain the prediction time to be predicted, and select the traffic flow data one hour before the prediction time from the historical traffic flow database to obtain the first matrix;

[0062] In one embodiment of the present invention, the prediction time A=2023 / 11 / 27 20:00:00, then according to the data in the historical traffic flow database, the expression of the first matrix is ​​as follows:

[0063] B=[23,24,25,26,27,28,29,30,31,32,33,34]

[0064] S3: Select the traffic flow data of the same time period as one hour before the prediction time in the historical traffic flow database, and integrate and construct a historical data set matrix;

[0065] In one embodiment of the present invention, the traffic flow data selected by S3 is the traffic flow data from 2022 / 11 / 27 00:00:00 to 2023 / 11 / 28 00:00:00 and the time period from 19:05:00 to 20:00:00, and the historical data set matrix C is constructed. [n][], which is expressed as follows:

[0066]

[0067] Among them, n is the total number of days, and 12 is the 12 data corresponding to the first matrix every day;

[0068] S4: According to the prediction time to be predicted, traffic flow data of the same time period as half an hour after the prediction time is selected from the historical traffic flow database to obtain a prediction set matrix;

[0069] In one embodiment of the present invention, the traffic flow data selected by S4 is the traffic flow data of the period from 20:05:00 to 20:35:00 between 2022 / 11 / 27 00:00:00 and 2023 / 11 / 27 00:00:00, and the prediction set matrix is ​​obtained, and its expression is as follows:

[0070]

[0071] S5: Calculate the Euclidean distance between the first matrix and the historical data set matrix, and construct a Euclidean distance matrix according to the calculation results;

[0072] In one embodiment of the present invention, the Euclidean distance between each row of the first matrix and the historical data set matrix is ​​calculated, and its expression is as follows:

[0073]

[0074] The calculation result of the Euclidean distance of each row is constructed into a Euclidean distance matrix Dis[n][2], where the number of rows n is the number of days and the two columns are stored separately. The expression is as follows:

[0075]

[0076] S6: Select the first five groups of smallest data in the Euclidean distance matrix and substitute them into the prediction set matrix to calculate the traffic flow prediction data;

[0077] In one embodiment of the present invention, the Euclidean distance matrix is ​​sorted from small to large, and its expression is as follows:

[0078]

[0079] Among them, [x,0] is the overlapping data one hour before the prediction time A, which needs to be ignored in subsequent processing. Starting from the row [n-2,72.276], we get 5 sets of data, and the expression is as follows:

[0080] index_1 = Dis_sort[1][0]

[0081] index_2 = Dis_sort[2][0]

[0082] index_3 = Dis_sort[3][0]

[0083] index_4 = Dis_sort[4][0]

[0084] index_5 = Dis_sort[5][0]

[0085] Substitute the five sets of data into the prediction set matrix and calculate the traffic flow prediction data;

[0086] In one embodiment of the present invention, the output is:

[0087] The traffic flow forecast results for 2023 / 11 / 27 20:05:00 are:

[0088] preTotal_1=(D[index_1][0]+...+D[index_5][0]) / 5.0

[0089] The traffic flow forecast results for 2023 / 11 / 27 20:10:00 are:

[0090] preTotal_2=(D[index_1][1]+...+D[index_5][1]) / 5.0

[0091] The traffic flow forecast results for 2023 / 11 / 27 20:15:00 are:

[0092] preTotal_3=(D[index_1][2]+...+D[index_5][2]) / 5.0

[0093] The traffic flow forecast results for 2023 / 11 / 27 20:20:00 are:

[0094] preTotal_4=(D[index_1][3]+...+D[index_5][3]) / 5.0

[0095] The traffic flow forecast results for 2023 / 11 / 27 20:25:00 are:

[0096] preTotal_5=(D[index_1][4]+...+D[index_5][4]) / 5.0

[0097] The traffic flow forecast results for 2023 / 11 / 27 20:30:00 are:

[0098] preTotal_6=(D[index_1][5]+...+D[index_5][5]) / 5.0

[0099] The beneficial effects of the present invention are as follows: the present invention predicts the traffic flow of highway sections through historical traffic flow data, and integrates the influence of time factors and external environmental factors on traffic flow changes. Compared with traditional traffic flow prediction algorithms, the present invention has higher prediction accuracy, and is suitable for traffic flow prediction of highway sections of different types and different usage conditions, and has stronger applicability and scalability.

Claims

1. A method for predicting highway section traffic flow based on historical data trends, characterized in that: The following steps are involved: S1: Collect historical traffic flow data of highway sections and perform preprocessing to obtain a historical traffic flow database, wherein the time step of the data in the historical traffic flow database is M minutes; S2: Obtain the prediction time to be predicted, and select the traffic flow data of M×K minutes before the prediction time from the historical traffic flow database to obtain the first matrix, where K is the number of set time points; S3: Select the traffic flow data of the same time period as M×K minutes before the prediction time in the historical traffic flow database, and integrate and construct the historical data set matrix; S4: According to the predicted time, select the time corresponding to the predicted time from the historical traffic flow database. The traffic flow data of the same time period in minutes is used to obtain the prediction set matrix; S5: Calculate the Euclidean distance between the first matrix and the historical data set matrix, and construct a Euclidean distance matrix according to the calculation results; S6: Select the data with the smallest preset range in the Euclidean distance matrix and substitute it into the prediction set matrix to calculate and obtain traffic flow prediction data; The expression of the historical data set matrix is ​​as follows: Among them, C [n][K] is the historical data set matrix, n is the number of days of the historical traffic flow database, is the historical traffic flow data of M×(K-1) minutes before the prediction time with day 1, is the historical traffic flow data of M×(K-2) minutes before the prediction time of day 1, is the historical traffic flow data M minutes before the prediction time on day 1, is the historical traffic flow data at the prediction time of day 1, is the historical traffic flow data of M×(K-1) minutes before the prediction time with n days, is the historical traffic flow data of M×(K-2) minutes before the prediction time of day n, is the historical traffic flow data M minutes before the prediction time on day n. is the historical traffic flow data at the prediction time with n days.

2. The highway section traffic flow prediction method based on historical data situation according to claim 1 is characterized in that: The expression of the first matrix is ​​as follows: B=[x1,x2,…,x K-1 ,x K ] Among them, B is the first matrix, x K is the historical traffic flow data at the prediction time, x K-1 is the historical traffic flow data M minutes before the prediction time, x2 is the historical traffic flow data M×(K-2) minutes before the prediction time, and x1 is the historical traffic flow data M×(K-1) minutes before the prediction time.

3. The highway section traffic flow prediction method based on historical data situation according to claim 1 is characterized in that: The expression of the prediction set matrix is ​​as follows: Among them, D [n][K] is the prediction set matrix, n is the number of days in the historical traffic flow database, is the historical traffic flow data M minutes after the prediction time on day 1, The historical traffic flow data is M×2 minutes after the prediction time of day 1. After the forecast time with a day number of 1 Minutes of historical traffic flow data, is the historical traffic flow data M minutes after the prediction time on day n, is the historical traffic flow data of M×2 minutes after the prediction time of day n, After the forecast time n days Minute-by-minute historical traffic data.

4. The highway section traffic flow prediction method based on historical data situation according to claim 1 is characterized in that: The Euclidean distance between the first matrix and the historical data set matrix is ​​expressed as follows: Where Dis_Eculid is the Euclidean distance between the first matrix and the historical dataset matrix, x K is the Kth flow in the first matrix, is the Kth flow data on the nth day in the historical flow, and Z is the first Z groups of historical flow sequences with the smallest distance.

Images