Target vehicle screening system

Abstract

The invention relates to the technical field of traffic, in particular to a target vehicle screening system, the system comprises a first storage medium, a second storage medium, a processor and a memory in which a computer program is stored, and when the computer program is executed by the processor, the following steps are realized: according to a plurality of initial travel characteristic values corresponding to each first vehicle, a plurality of first travel characteristic values corresponding to each first vehicle; and clustering the plurality of first vehicles into a plurality of initial vehicle clusters, obtaining an initial splitting coefficient and a plurality of target parameters of the self-iteration clustering model according to the plurality of initial vehicle clusters, then performing clustering processing in combination with a preset maximum merging number and an initial iteration number, obtaining a plurality of final vehicle clusters, and screening the plurality of final vehicle clusters according to a preset screening condition. Screening out a final vehicle cluster meeting conditions and target vehicles with given travel characteristics from all the final vehicle clusters; through the combination of the two clustering models, the clustering efficiency and reliability can be improved, and the required commuting vehicles can be screened out more comprehensively and accurately.

Description

Technical Field

[0001] The invention relates to the field of traffic technology, in particular to a target vehicle screening system. Background Art

[0002] In the context of urbanization, as commuting distances increase, more and more people are switching from traditional commuting to private car commuting. The rapid increase in private car commuting has had a great impact on peak-hour traffic. Commuting issues not only involve everyone's travel, but also reflect the level of traffic management in the entire city. Therefore, accurately screening out commuting private cars from traffic flows can effectively discover peak-hour commuting traffic conditions from a spatiotemporal dimension, and is also conducive to adopting real-time and efficient control strategies. At present, when judging whether a vehicle is a commuter vehicle, its travel attributes are mainly determined based on the vehicle's registration information in the transportation department, various questionnaires, or follow-up surveys of individual vehicles, which wastes a lot of manpower, has low accuracy, and is difficult to generalize to the identification of many vehicles in a city or even a larger area. Summary of the invention

[0003] In view of the above technical problems, the technical solution adopted by the present invention is:

[0004] A target vehicle screening system, the system comprising: storing a first vehicle set C = {C1, C2, ..., C i , ..., C m}, a first storage medium storing m initial travel feature sets corresponding to C, a processor, and a memory storing a computer program, wherein C i is the i-th first vehicle, i=1, 2, ..., m, m is the number of first vehicles, the first vehicles correspond to the initial travel feature sets one by one, each initial travel feature set includes a number of initial travel feature values, when the computer program is executed by the processor, the following steps are implemented:

[0005] S100, according to each C i The k-means clustering model is used to cluster C into the initial vehicle cluster U = {U1, U2, ..., U g , ..., U h}, where U g is the g-th initial vehicle cluster, g=1, 2, ..., h, h is the number of initial vehicle clusters; each initial vehicle cluster includes a number of initial vehicles.

[0006] S200, according to U, obtain the initial split coefficient ζ of the given self-iterative clustering model.

[0007] Among them, the initial splitting coefficient ζ meets the following conditions:

[0008] ζ=∑ h g＝1 ((∑ Zg v＝1 (L gv / λ g )) / Z g ) / h, where L gv For U g The travel eigenvalue vector corresponding to the vth initial vehicle in U g The distance of the center point vector, λ g For U g The travel eigenvalue vectors corresponding to all initial vehicles in are respectively g The maximum distance among the distances of the center point vector, Z g For U g The number of initial vehicles in; the travel feature vector corresponding to each initial vehicle is a vector composed of several initial travel feature values ​​corresponding to the initial vehicle itself.

[0009] S300, based on ζ, a preset maximum number of merges, a preset number of initial iterations, and several target parameters corresponding to the self-iterative clustering model obtained according to U, the self-iterative clustering model is used to iteratively cluster U to obtain several final vehicle clusters.

[0010] S400, screening out a corresponding final vehicle cluster from a plurality of final vehicle clusters according to a preset screening condition, and determining each first vehicle in the screened final vehicle cluster as a target vehicle having a given travel characteristic.

[0011] Compared with the prior art, the present invention has obvious beneficial effects. By means of the above technical solution, a target vehicle screening system provided by the present invention can achieve considerable technical advancement and practicality, and has wide industrial utilization value, and has at least the following beneficial effects:

[0012] The present invention provides a target vehicle screening system. When a computer program in the system is executed by a processor, firstly, according to a plurality of initial travel characteristic values ​​corresponding to each first vehicle, a plurality of first vehicles are clustered into a plurality of initial vehicle clusters, and according to the plurality of initial vehicle clusters and the dispersion of samples in each cluster, a plurality of target parameters and an initial splitting coefficient of a self-iterative clustering model are obtained, wherein the initial splitting coefficient is obtained based on the clustering result of a k-means clustering model, which is conducive to the reasonable splitting of samples in the cluster and accelerates the convergence of the self-iterative clustering model, thereby improving the efficiency and reliability of the self-iterative clustering; then, clustering processing is performed in combination with a preset maximum merging number and an initial iteration number to obtain a plurality of final vehicle clusters, and according to a preset screening condition, a qualified final vehicle cluster is screened out from the plurality of final vehicle clusters, and each initial vehicle in the screened final vehicle cluster is determined as a target vehicle with a given travel characteristic, that is, a commuter vehicle with a stable travel characteristic during a peak period is screened out; the present invention can improve the clustering efficiency and reliability by combining two clustering models, and is conducive to screening out the required commuter vehicles more comprehensively and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 A flowchart of a processor executing a computer program in a target vehicle screening system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0016] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0017] A target vehicle screening system, the system comprising: storing a first vehicle set C = {C1, C2, ..., C i , ..., C m}, a first storage medium storing m initial travel feature sets corresponding to C, a processor, and a memory storing a computer program, wherein C i is the i-th first vehicle, i=1, 2, ..., m, m is the number of first vehicles, the first vehicle corresponds to the initial travel feature set one by one, and each initial travel feature set includes several initial travel feature values; it can be understood that: what is stored in the system is the unique identifier of the first vehicle, for example, the license plate number.

[0018] When the computer program is executed by a processor, the following steps are implemented: Figure 1 As shown:

[0019] S100, according to each C i The k-means clustering model is used to cluster C into the initial vehicle cluster U = {U1, U2, ..., U g , ..., U h}, where U g is the g-th initial vehicle cluster, g=1, 2, ..., h, h is the number of initial vehicle clusters; each initial vehicle cluster includes a number of initial vehicles.

[0020] Specifically, the first vehicle is any one of several randomly selected vehicles.

[0021] In a specific embodiment, in step S100, U is obtained by the following steps:

[0022] S101, according to C i The corresponding initial travel characteristic values ​​are obtained from C i The corresponding travel feature vector.

[0023] S102, using factor analysis to analyze C i The corresponding travel eigenvalue vector is processed by dimensionality reduction to obtain C i The corresponding target travel feature value vector; the dimension of each target travel feature value vector is the dimension obtained after dimensionality reduction processing.

[0024] For ease of understanding, the following explanation is given:

[0025] When factor analysis is used for dimensionality reduction, a multidimensional vector composed of several initial travel characteristic values ​​is input into the factor analysis model, and the number of output factors can be actively set. In the present embodiment, the number of output factors is set to three, that is, the vector composed of several initial travel characteristic values ​​is reduced to three dimensions. Since the present application directly adopts the factor analysis method of the prior art, its specific implementation process will not be described in detail.

[0026] As described above, by using the factor analysis method to perform dimensionality reduction processing on several initial travel characteristic values ​​of the first vehicle, some linear relationships between multiple indicators can be eliminated, the dimension of the vector can be reduced, and clustering can be performed using the vector after dimensionality reduction, which can reduce the complexity of clustering and thus improve the reliability of clustering.

[0027] S103, according to each C i Corresponding to the target travel characteristic value vector, the k-means clustering model is used to cluster the m first vehicles into a preset number of initial vehicle clusters to obtain U; ​​it can be understood as: the k-means clustering model is used to cluster the m target travel characteristic value vectors, and the initial vehicle cluster corresponding to each cluster is obtained according to the clustering results and the first vehicle corresponding to each target travel characteristic value; for example, when there are three dimensions after dimensionality reduction, the target travel characteristic value vector corresponding to each first vehicle is a vector composed of three dimensions. Based on the target travel characteristic value vector corresponding to each first vehicle, the k-means clustering model is used to cluster several target travel characteristic value vectors. Those skilled in the art know the specific implementation of the k-means clustering model and will not go into details here.

[0028] As mentioned above, since the k-means clustering model has the advantages of fast convergence speed and few parameters that need to be adjusted, the k-means clustering model is used for pre-clustering processing. The clustering speed is fast and several target parameters in the subsequent self-iterative clustering can be obtained, reducing the operational complexity of the self-iterative clustering.

[0029] Specifically, the initial travel feature set corresponding to each first vehicle includes the spatial stability coefficient, peak travel frequency, total travel frequency, off-peak average travel frequency, off-peak travel frequency deviation coefficient of the first vehicle itself in the historical time period, and the pre-acquired stability values ​​of the starting point and the stability values ​​of the end point corresponding to the preset morning peak period of the historical time period, and the stability values ​​of the starting point and the stability values ​​of the end point corresponding to the preset evening peak period of the historical time period; in a specific implementation, the preset morning peak period is 7:00-9:00, and the preset evening peak period is 18:00-20:00.

[0030] In a specific embodiment, when the computer program is executed by a processor, the stability value of the starting point of the first vehicle corresponding to the preset morning / evening peak period in the historical time period is obtained through the following steps:

[0031] S110, for any first vehicle in a given first vehicle set, obtain a number of first travel trajectories corresponding to a preset morning / evening peak period of the first vehicle in a historical time period; it can be understood that: the number of first travel trajectories corresponding to the preset morning peak period of the first vehicle in the historical time period are obtained in the same manner as the number of first travel trajectories corresponding to the preset evening peak period of the first vehicle in the historical time period.

[0032] Specifically, step S110 includes the following steps:

[0033] S1101, receiving a plurality of vehicle position information reported by the GPS device of the first vehicle within a historical time period; it can be understood that: the vehicle position information refers to the latitude and longitude information of the vehicle reported in real time by the GPS device.

[0034] S1102, dividing the plurality of vehicle location information into a plurality of vehicle location information sets; the reporting time of the vehicle location information in each vehicle location information set is continuous and the reporting time interval is not greater than a preset time threshold; the time interval between each two adjacent vehicle location information sets is greater than a preset time threshold.

[0035] For better understanding, the following explanation is made: for example, the preset time threshold is 15 minutes. If the GPS device reports the vehicle location information three times consecutively at 6:50, 7:10 and 7:11 respectively, the vehicle location information reported at 6:50 will be divided into the previous vehicle location information set, and the vehicle location information reported at 7:10 and 7:11 will be divided into the latter vehicle location information set.

[0036] S1103, based on the multiple vehicle location information in each vehicle location information set, obtain the initial travel trajectory corresponding to each vehicle location information set; in a specific implementation, the GPS device usually uploads the vehicle location information every few seconds, and connects the multiple vehicle location information in the vehicle location information set in series to obtain the corresponding initial travel trajectory.

[0037] S1104, when the travel time corresponding to the initial travel trajectory overlaps with the preset morning / evening peak period, the initial travel trajectory itself is obtained as a first travel trajectory corresponding to the preset morning / evening peak period of the first vehicle in the historical time period.

[0038] As mentioned above, when obtaining the first trip trajectory corresponding to the morning / evening peak period of a vehicle within multiple days, the driving time of each first trip trajectory is first obtained, without considering the overlap between the driving time and the morning / evening peak period. As long as there is an overlap, it is determined to be the first trip trajectory corresponding to the morning / evening peak period. In this way, the different working hours of different people, the different distances between home and company, etc., which lead to inconsistent travel times, are taken into account, and thus a number of vehicles with morning / evening peak travel characteristics can be determined more comprehensively and reasonably.

[0039] S130, according to several first travel trajectories, obtain several first starting point hash values ​​and several second starting point hash values ​​corresponding to the first vehicle; the first starting point hash value refers to the first geohash value corresponding to the several vehicle position information reported by the GPS device of the first vehicle in the first travel trajectory, and the second starting point hash value refers to the second geohash value corresponding to the several vehicle position information reported by the GPS device of the first vehicle in the first travel trajectory; it can be understood as: the first geohash hash block and the second geohash hash block passed in the first travel trajectory; in a specific implementation, the geohash values ​​corresponding to the first hash block and the second hash block are both seven bits.

[0040] As mentioned above, by obtaining several first starting point hash values, the starting position of each trip of the first vehicle corresponding to several preset morning / evening peak periods can be known, and then whether each starting point position is the same can be known, which is conducive to determining the stability of subsequent starting points, and at the same time can provide an important reference for the screening of vehicles with morning / evening peak travel characteristics.

[0041] S150, determining the first starting point hash values ​​with the largest number and the second largest number among the plurality of first starting point hash values ​​as the first target hash value and the second target hash value respectively, and determining the second starting point hash value with the largest number among the plurality of second starting point hash values ​​as the third target hash value.

[0042] For better understanding, the following explanation is made: for example, there are a total of ten first starting point hash values, five of which represent the first area, four of which represent the second area, and one first starting point hash value represents the third area. Then the first starting point hash value corresponding to the first area is used as the first target hash value, and the first starting point hash value corresponding to the second area is used as the second target hash value.

[0043] S170, when there is a first target hash value or the second target hash value is the same as the third target hash value, the ratio of the sum of the quantities corresponding to the first target hash value and the second target hash value to the given number of days is determined as the stability value of the starting point of the first vehicle corresponding to the preset morning / evening peak period; otherwise, the ratio of the quantity corresponding to the first target hash value to the given number of days is determined as the stability value of the starting point of the first vehicle corresponding to the preset morning / evening peak period; it can be understood that: the stability value of the starting point of the first vehicle corresponding to the preset evening peak period is obtained in the same way as the stability value of the starting point corresponding to the preset morning peak period.

[0044] Specifically, the given number of days refers to the number of working days in the historical time period; in a specific implementation, the historical time period is generally selected from the past month, and the working days are Monday to Friday of each week.

[0045] In the above, according to the situation that the third target hash value is the same as the first target hash value or the second target hash value, different calculation methods of the stability values ​​of the starting points are adopted. That is, when the third target hash value is the same as the first target hash value or the second target hash value, it means that the travel trajectory is consistent. It may be that the error in reporting time of the GPS device causes the inconsistent starting point. When the third target hash value is different from the first target hash value and the second target hash value, it means that the travel trajectory is inconsistent, and thus the travel starting point is inconsistent. Therefore, through the above calculation method, the stability value of the starting point obtained is more accurate and reasonable.

[0046] In a specific embodiment, when the computer program is executed by a processor, the stability value of the first vehicle at the destination corresponding to the preset morning / evening peak period is obtained through the following steps:

[0047] S120, based on several first travel trajectories, obtain several first destination hash values ​​and several second destination hash values ​​corresponding to the first vehicle; the first destination hash value refers to the last geohash value corresponding to the several vehicle position information reported by the GPS device of the first vehicle in the first travel trajectory, and the second destination hash value refers to the second to last geohash value corresponding to the several vehicle position information reported by the GPS device of the first vehicle in the first travel trajectory; in a specific implementation, the geohash values ​​corresponding to the first to last hash block and the second to last hash block are both seven digits.

[0048] As mentioned above, by obtaining several first terminal hash values, the terminal position of each trip of the first vehicle corresponding to several preset morning / evening peak periods can be known, and then whether each terminal position is the same can be known, which is conducive to determining the stability of subsequent terminal points, and at the same time can provide an important reference for the screening of vehicles with morning / evening peak travel characteristics.

[0049] S140: Determine the largest and second largest first destination hash values ​​among the plurality of first destination hash values ​​as the fourth target hash value and the fifth target hash value, respectively; and determine the largest second destination hash value among the plurality of second destination hash values ​​as the sixth target hash value.

[0050] S160, when there is a fourth target hash value or the fifth target hash value is the same as the sixth target hash value, the ratio of the sum of the quantities corresponding to the fourth target hash value and the fifth target hash value to the given number of days is determined as the stability value of the destination corresponding to the preset morning / evening peak period of the first vehicle; otherwise, the ratio of the quantity corresponding to the fourth target hash value to the given number of days is determined as the stability value of the destination corresponding to the preset morning / evening peak period of the first vehicle; it can be understood that: the stability value of the destination corresponding to the preset evening peak period of the first vehicle is obtained in the same way as the stability value of the destination corresponding to the preset morning peak period.

[0051] As described above, in the process of obtaining the endpoint stability value, the fourth target hash value, the fifth target hash value and the sixth target hash value are determined by reverse deduction. According to the situation that the sixth target hash value is the same as the fourth target hash value or the fifth target hash value, different methods of calculating the endpoint stability value are adopted. Similarly, the obtained endpoint stability value is more accurate and reasonable.

[0052] Specifically, the peak travel frequency corresponding to the first vehicle is the ratio of the number of working days with peak travel in the historical time period to the given number of days.

[0053] Specifically, the total travel frequency corresponding to the first vehicle is the ratio of the number of working days for travel in the historical time period to the given number of days.

[0054] Specifically, the off-peak average travel frequency P corresponding to the first vehicle meets the following conditions:

[0055] P=(∑ n j＝1 (a j / A j )) / n, where n is the total number of days in the historical period, a j is the number of non-peak trips on the jth day in the historical time period, A j is the total number of trips on the jth day in the historical time period.

[0056] Specifically, the off-peak travel frequency deviation coefficient σ corresponding to the first vehicle meets the following conditions:

[0057] σ=sqrt((∑ n j＝1 (a j / A j -P) 2 ) / n), where sqrt() is the square root function.

[0058] Further, the processor obtains the space stability coefficient corresponding to the first vehicle through the following steps:

[0059] S10, obtain the spatial distance O1D1 between the first target hash value corresponding to the preset morning rush hour period and the fourth target hash value corresponding to the preset morning rush hour period, the spatial distance O1D2 between the first target hash value corresponding to the preset morning rush hour period and the fourth target hash value corresponding to the preset evening rush hour period, the spatial distance O2D1 between the first target hash value corresponding to the preset evening rush hour period and the fourth target hash value corresponding to the preset morning rush hour period, and the spatial distance O2D2 between the first target hash value corresponding to the preset evening rush hour period and the fourth target hash value corresponding to the preset evening rush hour period.

[0060] S20, obtaining a distance matrix D of the starting and ending points according to O1D1, O1D2, O2D1 and O2D2.

[0061] Specifically, the distance matrix D of the start and end points meets the following conditions:

[0062]

[0063] S30, calculating the spatial stability coefficient η corresponding to the first vehicle according to O1D1, O1D2, O2D1, O2D2 and the distance matrix of the starting and ending points.

[0064] Specifically, the space stability coefficient η corresponding to the first vehicle meets the following conditions:

[0065] η=|D| / (max{O1D1,O2D2})2 -(O1D2+O2D1) / max{O1D1,O2D2}.

[0066] As mentioned above, the calculation of the spatial stability coefficient has changed the meaning of the starting point and the end point on the basis of the existing spatial stability coefficient formula. Through the calculation of the spatial stability coefficient, it can reflect the changes in the starting point and the end point of the vehicle during peak hours and the stability of the travel trajectory during peak hours. The spatial stability coefficient, as a subsequent parameter, is conducive to screening out vehicles with peak travel characteristics.

[0067] S200, according to U, obtain the initial split coefficient ζ of the given self-iterative clustering model.

[0068] Among them, the initial splitting coefficient ζ meets the following conditions:

[0069] ζ=∑ h g＝1 ((∑ Zg v＝1 (L gv / λ g )) / Z g ) / h, where L gv For U g The travel eigenvalue vector corresponding to the vth initial vehicle in U g The distance of the center point vector, λ g For U g The travel eigenvalue vectors corresponding to all initial vehicles in are respectively g The maximum distance among the distances of the center point vector, Z g For U g The number of initial vehicles in; the travel feature vector corresponding to each initial vehicle is a vector composed of several initial travel feature values ​​corresponding to the initial vehicle itself.

[0070] As mentioned above, the initial split coefficient is the difference between the self-iterative clustering model and the k-means clustering model. On the basis of the clustering results of the k-means clustering model, the initial split coefficient corresponding to the self-iterative clustering model is obtained according to the dispersion of several initial vehicles in each initial vehicle cluster in the clustering results, that is, the distance between the travel feature value vector corresponding to the initial vehicle and the center point vector of the initial vehicle cluster. Using it as an initial parameter of the self-iterative clustering model is conducive to the reasonable splitting of samples within the cluster, and accelerates the convergence of the self-iterative clustering model, thereby improving the efficiency and reliability of secondary clustering.

[0071] S300, based on ζ, a preset maximum number of merges, a preset number of initial iterations, and several target parameters corresponding to the self-iterative clustering model obtained according to U, the self-iterative clustering model is used to iteratively cluster U to obtain several final vehicle clusters; those skilled in the art are aware of the specific implementation of the self-iterative clustering model, which will not be described in detail here.

[0072] In a specific embodiment, in step S300, the maximum number of merging is determined by the following steps:

[0073] S301, obtain U g The distance Y between the travel feature vectors corresponding to any two initial vehicles in .

[0074] S302, when Y≤ρ, merge any two initial vehicles and divide them into one initial vehicle group, where ρ is a preset distance threshold; it can be understood that: each initial vehicle group has a number of initial vehicles, and the distance between the travel feature vectors corresponding to any two initial vehicles in the group is not greater than the preset distance threshold.

[0075] S303, obtaining a target vehicle group corresponding to each initial vehicle cluster; the target vehicle group corresponding to each initial vehicle cluster refers to an initial vehicle group containing the most initial vehicles among a plurality of initial vehicle groups corresponding to the initial vehicle cluster itself.

[0076] S304, obtaining the initial number of vehicles in each target vehicle group, and determining the minimum initial number of vehicles as the maximum merging number.

[0077] As mentioned above, based on the clustering results of the first clustering, the maximum merger number corresponding to each initial vehicle cluster is calculated according to the dispersion of several initial vehicles in each initial vehicle cluster in the clustering results, that is, the distance between several travel characteristic value vectors corresponding to the initial vehicles. Using it as the initial parameter of the self-iterative clustering model can reduce the number of iterations of the secondary clustering, accelerate convergence, and improve the reliability of clustering.

[0078] Specifically, the several target parameters include the number of clusters, the center point vector corresponding to each cluster, the initial number of samples corresponding to each cluster, the sample standard deviation threshold corresponding to each cluster and the initial shortest distance between clusters; it can be understood that: the number of clusters corresponding to the self-iterative clustering model is the number of initial vehicle clusters obtained after clustering by the k-means clustering model, the center point vector corresponding to each cluster is the center point vector corresponding to each initial vehicle cluster, the initial number of samples corresponding to each cluster is the minimum number of samples among all initial vehicle clusters, the sample standard deviation threshold corresponding to each cluster is not greater than the standard deviation of each initial vehicle cluster, and the initial shortest distance between clusters is not greater than the shortest distance between any two initial vehicle clusters among all initial vehicle clusters.

[0079] As mentioned above, based on the k-means clustering model, the self-iterative clustering model adds three parameters: the maximum number of merges, the initial number of iterations, and the initial split coefficient. It also adds the operations of merging and splitting the clustering results and iterates them to make the sample features in each cluster more similar, thereby obtaining more accurate classification results.

[0080] S400, screening out a corresponding final vehicle cluster from a plurality of final vehicle clusters according to a preset screening condition, and determining each first vehicle in the screened final vehicle cluster as a target vehicle having a given travel characteristic.

[0081] In a specific embodiment, step S400 also includes the following steps:

[0082] S401, according to the number of vehicles in each final vehicle cluster, obtain the total number of vehicles corresponding to several final vehicle clusters; for example: when there are three final vehicle clusters, and the number of vehicles in the clusters is 100, 150, and 200 respectively, the total number of vehicles is 100+150+200=450.

[0083] S402, according to the total number of vehicles and the initial travel characteristics corresponding to each first vehicle, obtain the first spatial stability coefficient average value η1, the first peak travel frequency average value F1, the first total travel frequency average value Q1, the first non-peak average travel frequency average value K1, the first non-peak travel frequency deviation coefficient average value θ1, the first starting point stability average value S1 and the first end point stability average value E1 corresponding to the preset morning peak period, and the first starting point stability average value S1 corresponding to the preset evening peak period 0 1 and the average stability of the first endpoint E 0 1.

[0084] Specifically, η1 is the average value calculated based on the spatial stability coefficients of all first vehicles, F1, Q1, K1, θ1, S1, E1, S 0 1 and E 0 The method of obtaining 1 is the same as that of η1, which will not be repeated here.

[0085] S403, according to the number of vehicles in each final vehicle cluster and a number of initial travel characteristics corresponding to each first vehicle in each final vehicle cluster, obtain the second spatial stability coefficient average value η2, the second peak travel frequency average value F2, the second total travel frequency average value Q2, the second off-peak average travel frequency average value K2, the second off-peak travel frequency deviation coefficient average value θ2, the second starting point stability average value S2 and the second end point stability average value E2 corresponding to the preset morning peak period, and the second starting point stability average value S2 corresponding to the preset evening peak period. 0 2 and the average stability of the second endpoint E 0 2.

[0086] Specifically, η2 is the average value calculated based on the spatial stability coefficients of all first vehicles in any final vehicle cluster, F2, Q2, K2, θ2, S2, E2, S 0 2 and E 0 The method of obtaining 2 is the same as that of η2, which will not be repeated here; it can be understood that each final vehicle cluster corresponds to 9 indicators.

[0087] S404, select from all final vehicle clusters those that satisfy η2>η1, F2>F1, Q2>Q1, K2<K1, θ2<θ1, S2>S1, E2>E1, S 0 2>S 0 1 and E 0 2>E 0 1’s final vehicle cluster.

[0088] As mentioned above, F2 and Q2 are used to reflect the travel conditions during peak hours, K2 and θ2 are used to reflect the travel conditions during non-peak hours, and the remaining five indicators are used to reflect the stability of the start and end points and the spatial stability. Therefore, through the above screening method, it is possible to more comprehensively and accurately screen out vehicles in each final vehicle cluster that meet the peak travel conditions and have relatively stable start and end points during peak travel, that is, to screen out the required commuter vehicles, which is conducive to the subsequent management and planning of traffic in the region.

[0089] The present embodiment provides a target vehicle screening system. When the computer program in the system is executed by the processor, first, according to the initial travel feature values ​​corresponding to each first vehicle, a plurality of first vehicles are clustered into a plurality of initial vehicle clusters, and according to the initial vehicle clusters and the dispersion of the samples in each cluster, a plurality of target parameters and the initial split coefficient of the self-iterative clustering model are obtained, wherein the initial split coefficient is obtained based on the clustering result of the k-means clustering model, which is conducive to the reasonable splitting of the samples in the cluster and accelerates the convergence of the self-iterative clustering model, thereby improving the efficiency and reliability of the self-iterative clustering; then, clustering processing is performed in combination with the preset maximum number of merges and the initial number of iterations to obtain a plurality of final vehicle clusters, and according to the preset screening conditions, the final vehicle clusters that meet the conditions are screened out from the plurality of final vehicle clusters, and each initial vehicle in the screened final vehicle clusters is determined as a target vehicle with a given travel feature, that is, commuter vehicles with stable travel characteristics during peak hours are screened out; the present invention can improve the clustering efficiency and reliability by combining the two clustering models, and is conducive to screening out the required commuter vehicles more comprehensively and accurately.

[0090] Although some specific embodiments of the present invention have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It will also be appreciated by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A target vehicle screening system, characterized in that: The system comprises: storing a first vehicle set C={C1, C2, . . . , C i , ..., C m }, a first storage medium storing m initial travel feature sets corresponding to C, a processor, and a memory storing a computer program, wherein C i is the i-th first vehicle, i=1, 2, ..., m, m is the number of first vehicles, the first vehicles correspond to the initial travel feature sets one by one, each initial travel feature set includes a number of initial travel feature values, when the computer program is executed by the processor, the following steps are implemented: S100, according to each C i The k-means clustering model is used to cluster C into the initial vehicle cluster U = {U1, U2, ..., U g , ..., U h }, where U g is the g-th initial vehicle cluster, g = 1, 2, ..., h, h is the number of initial vehicle clusters; each initial vehicle cluster includes a number of initial vehicles; S200, according to U, obtaining the initial split coefficient ζ of the given self-iterative clustering model; Among them, the initial splitting coefficient ζ meets the following conditions: ζ=∑ h g＝1 ((∑ Zg v＝1 (L gv / λ g )) / Z g ) / h, where L gv For U g The travel eigenvalue vector corresponding to the vth initial vehicle in U g The distance of the center point vector, λ g For U g The travel eigenvalue vectors corresponding to all initial vehicles in are respectively g The maximum distance among the distances of the center point vector, Z g For U g The number of initial vehicles in; the travel feature vector corresponding to each initial vehicle is a vector composed of several initial travel feature values ​​corresponding to the initial vehicle itself; S300, based on ζ, a preset maximum number of merges, a preset number of initial iterations, and a number of target parameters corresponding to the self-iterative clustering model obtained according to U, the self-iterative clustering model is used to iteratively cluster U to obtain a number of final vehicle clusters; S400, according to a preset screening condition, a corresponding final vehicle cluster is screened out from a plurality of final vehicle clusters, and each first vehicle in the screened final vehicle cluster is determined as a target vehicle having a given travel characteristic.

2. The target vehicle screening system according to claim 1, characterized in that: The initial travel feature set corresponding to each first vehicle includes the spatial stability coefficient, peak travel frequency, total travel frequency, off-peak average travel frequency, off-peak travel frequency deviation coefficient of the first vehicle itself in the historical time period, and the pre-acquired stability values ​​of the starting point and the stability values ​​of the end point corresponding to the preset morning peak period of the historical time period, and the pre-acquired stability values ​​of the starting point and the stability values ​​of the end point corresponding to the preset evening peak period of the historical time period.

3. The target vehicle screening system according to claim 1, characterized in that: In step S100, U is obtained by the following steps: S101, according to C i The corresponding initial travel characteristic values ​​are obtained from C i The corresponding travel feature vector; S102, using factor analysis to analyze C i The corresponding travel eigenvalue vector is processed by dimensionality reduction to obtain C i The corresponding target travel feature value vector; the dimension of each target travel feature value vector is the dimension obtained after dimensionality reduction processing; S103, according to each C i The corresponding target travel feature value vector is clustered into a preset number of initial vehicle clusters using the k-means clustering model to obtain U.

4. The target vehicle screening system according to claim 3, characterized in that: In step S300, the maximum number of merging is determined by the following steps: S301, obtain U g The distance Y between the travel feature vectors corresponding to any two initial vehicles in ; S302, when Y≤ρ, merging any two initial vehicles and dividing them into an initial vehicle group, where ρ is a preset distance threshold; S303, obtaining a target vehicle group corresponding to each initial vehicle cluster; the target vehicle group corresponding to each initial vehicle cluster refers to an initial vehicle group containing the most initial vehicles among a plurality of initial vehicle groups corresponding to the initial vehicle cluster itself; S304, obtaining the initial number of vehicles in each target vehicle group, and determining the minimum initial number of vehicles as the maximum merging number.

5. The target vehicle screening system according to claim 1, characterized in that: The target parameters include the number of clusters, the center point vector corresponding to each cluster, the initial number of samples corresponding to each cluster, the sample standard deviation threshold corresponding to each cluster and the initial shortest distance between clusters.

6. The target vehicle screening system according to claim 2, characterized in that: The step S400 also includes the following steps: S401, obtaining the total number of vehicles corresponding to a number of final vehicle clusters according to the number of vehicles in each final vehicle cluster; S402, according to the total number of vehicles and the initial travel characteristics corresponding to each first vehicle, obtain the first spatial stability coefficient average value η1, the first peak travel frequency average value F1, the first total travel frequency average value Q1, the first non-peak average travel frequency average value K1, the first non-peak travel frequency deviation coefficient average value θ1, the first starting point stability average value S1 and the first end point stability average value E1 corresponding to the preset morning peak period, and the first starting point stability average value S1 corresponding to the preset evening peak period 0 1 and the average stability of the first endpoint E 0 1; S403, according to the number of vehicles in each final vehicle cluster and a number of initial travel characteristics corresponding to each first vehicle in each final vehicle cluster, obtain the second spatial stability coefficient average value η2, the second peak travel frequency average value F2, the second total travel frequency average value Q2, the second off-peak average travel frequency average value K2, the second off-peak travel frequency deviation coefficient average value θ2, the second starting point stability average value S2 and the second end point stability average value E2 corresponding to the preset morning peak period, and the second starting point stability average value S2 corresponding to the preset evening peak period. 0 2 and the average stability of the second endpoint E 0 2; S404, select from all final vehicle clusters those that satisfy η2>η1, F2>F1, Q2>Q1, K2<K1, θ2<θ1, S2>S1, E2>E1, S 0 2>S 0 1 and E 0 2>E 0 1’s final vehicle cluster.

7. The target vehicle screening system according to claim 6, characterized in that: η1 is the average value calculated based on the spatial stability coefficients of all first vehicles, F1, Q1, K1, θ1, S1, E1, S 0 1 and E 0 The method of obtaining 1 is the same as η1.

8. The target vehicle screening system according to claim 6, characterized in that: η2 is the average value calculated from the spatial stability coefficients of all first vehicles in any final vehicle cluster, F2, Q2, K2, θ2, S2, E2, S 0 2 and E 0 The method of obtaining 2 is the same as η2.

Images