CN108764375A - Highway goods stock transprovincially matching process and device - Google Patents
Highway goods stock transprovincially matching process and device Download PDFInfo
- Publication number
- CN108764375A CN108764375A CN201810664921.6A CN201810664921A CN108764375A CN 108764375 A CN108764375 A CN 108764375A CN 201810664921 A CN201810664921 A CN 201810664921A CN 108764375 A CN108764375 A CN 108764375A
- Authority
- CN
- China
- Prior art keywords
- data
- outlet
- entrance
- transprovincially
- matching
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/02—Payment architectures, schemes or protocols involving a neutral party, e.g. certification authority, notary or trusted third party [TTP]
- G06Q20/023—Payment architectures, schemes or protocols involving a neutral party, e.g. certification authority, notary or trusted third party [TTP] the neutral party being a clearing house
Abstract
The present invention relates to communications and transportation statistical technique fields, and in particular to transprovincially matching process and device, method include for a kind of highway goods stock：Obtain the entry data and outlet data of the corresponding goods stock of charge station transprovincially to prestore, target algorithm model is obtained based on entry data and outlet data, to in charge station transprovincially pending entry data and pending outlet data use target algorithm model to be matched to obtain matching result, and similarity calculation is carried out to obtain similarity result to the license plate number in the license plate number and pending outlet data in the corresponding pending entry data of matching result, and the matching result is optimized based on the similarity result.By the above method with effective guarantee highway goods stock transprovincially matched accuracy, and then effectively solve expressway tol lcollection data segmentation problem and since directly goods stock matching problem transprovincially can not be carried out by license plate number caused by Car license recognition is not complete or identification mistake transprovincially.
Description
Technical field
The present invention relates to communications and transportation to count field, is transprovincially matched in particular to a kind of highway goods stock
Method and device.
Background technology
Important main line channel of the highway as highway transportation, for support national economic development, push social progress,
Safeguard national security etc. plays an important role.Especially the driving path of vehicle on a highway is reflecting regional warp
The important reference indicator for state of development of helping.
ExpresswayNetwork Toll Collection System is to save cascade network, provincial unified allocation settlement, expressway tol lcollection data at present
Essential record vehicle receives and dispatches situation in the IC card for respectively passing in and out freeway toll station inside the province, when vehicle is by adjacent province, needs
The highway passage IC card in different provinces is received and dispatched again at the provincial boundaries station of two provinces, carry out pike balance respectively, it is therefore, right
In vehicle transprovincially, traveling record is divided in different provinces.
Inventor it has been investigated that, due to the precision of car license recognition equipment is not high and expressway tol lcollection management require etc.
Reason causes the Car license recognition rate in expressway tol lcollection data not high, directly can not carry out vehicle string transprovincially by license plate number
It connects.Therefore, transprovincially matching of the goods stock on national highway is effectively solved, to obtain goods stock on a highway
Complete driving path is a technical problem to be solved urgently.
Invention content
In view of this, the purpose of the present invention is to provide a kind of highway goods stock transprovincially matching process and device,
Above-mentioned technical problem is effectively relieved.
To achieve the above object, the embodiment of the present invention adopts the following technical scheme that：
A kind of highway goods stock transprovincially matching process, the method includes：
The entry data and outlet data of the corresponding goods stock of charge station transprovincially to prestore are obtained, the outlet number is based on
The matching degree generation of the license plate number in license plate number and entry data in includes the sample number of matched data and non-matched data
According to：
By in the sample data entry data and the outlet data perform mathematical calculations to obtain overall target, and
The overall target is counted to obtain correlation metric, and the target signature collection is allocated to obtain training spy
Collection and test feature collection；
The correlation metric is handled to obtain candidate characteristic set, and important journey is carried out to the candidate characteristic set
Degree evaluation is to obtain target signature collection；
To the training characteristics collection use a variety of default machine learning algorithms be trained study with it is corresponding obtain it is a variety of across
Matching algorithm model is saved, and chooses a kind of the first algorithm model of conduct from a variety of models of matching algorithm transprovincially；
It uses the first algorithm model described in the test feature set pair to be assessed to obtain assessment result, and is commented according to this
Estimate result to be adjusted to obtain target algorithm model first algorithm model；
To in charge station transprovincially pending entry data and pending outlet data using the target algorithm model into
Row matching is to obtain matching result；
The car plate in license plate number and pending outlet data in pending entry data corresponding to the matching result
Number similarity calculation is carried out to obtain similarity result, and the matching result is optimized based on the similarity result.
Optionally, described to obtain the charge station transprovincially to prestore in above-mentioned highway goods stock transprovincially matching process
The entry data and outlet data of corresponding goods stock, based on the vehicle in the license plate number and entry data in the outlet data
The matching degree of the trade mark generates the step of matched data and non-matched data and includes：
Charge station transprovincially is obtained according to freeway net topological structure, charge station location and provincial administrative area boundary to believe
Breath, and obtain the goods stock outlet data corresponding with the information of charge station transprovincially and entry data to prestore；
Have the corresponding outlet data of complete license plate number as target outlet data using in outlet data, using preset algorithm
From searched in the entry data time with the complete license plate number in the target outlet data in consistent and outlet data and
The difference of time in entry data is located at the entry data of a setting time range as target entries data, and by the target
Entry data and with the target outlet data of the target entries Data Matching as matched data, using other data as mismatching
Data.
Optionally, in above-mentioned highway goods stock transprovincially matching process, will there is complete car plate in outlet data
Number corresponding outlet data searches to go out with the target from the entry data as target outlet data using preset algorithm
The difference of time of the complete license plate number unanimously and in time and entry data in outlet data in mouth data is located at a setting
The entry data of time range as target entries data, and by the target entries data and with the target entries Data Matching
Target outlet data include as matched data, using other data as the step of non-matched data：
License plate number L in outlet data is calculated using JaroWinklerDistance algorithmsOutletWith the vehicle in entry data
Trade mark LEntranceSimilarity Slicense：
Wherein, m LOutletAnd LEntranceMatched number of characters, t are the number of transposition；
It must not be less than V kilometers of standard per hour according to the minimum speed of highway, according to the distance D of charge station transprovincially,
Calculate the running time T from outlet charge station to entrance charge station：
T=(D/V) × 60
Entry data is screened based on the running time T, is subtracted when screening obtains the entry time in entry data
Go the Outlet time of outlet data in time interval [- T, T] range and license plate number similarity SlicenseMore than a setting value
When, judge that the corresponding vehicle of corresponding outlet data and the corresponding vehicle of entry data are same vehicles, and by the entry data
And it is included in respectively as the target entries data and target outlet data with the matched outlet data of the entry data
With data, other data are included in non-matched data.
Optionally, in above-mentioned highway goods stock transprovincially matching process, the entry data includes entrance charge
It stands coding, entrance license plate number, entry time, entrance vehicle, the total number of axle of entrance vehicle, entrance vehicle goods gross weight and entrance vehicle
Freight weight limit, the outlet data include outlet charge station coding, outlet license plate number, Outlet time, outlet vehicle, outlet vehicle line shaft
Number, outlet vehicle goods gross weight and outlet vehicle weight limitation, by the sample data entry data and the outlet data carry out
Mathematical operation is to obtain overall target, and the step of being counted the overall target to obtain correlation metric includes：
Utilize the entry time T in entry dataEntrance, entrance vehicle CEntrance, the total number of axle A of entrance vehicleEntrance, entrance vehicle goods it is total
Weight WEntranceAnd entrance vehicle weight limitation LWEntranceSubtract the Outlet time T in corresponding outlet dataOutlet, outlet vehicle COutlet, outlet vehicle
Total number of axle AOutlet, outlet vehicle goods gross weight WOutletAnd outlet vehicle weight limitation LWEntranceObtain overall target：
Dtime=TEntrance-TOutlet
Dcar=CEntrance-COutlet
Daxis=AEntrance-AOutlet
Dweight=WEntrance-WOutlet
Dlimitweight=LWEntrance-LWOutlet
Wherein, DtimeFor entrance time difference, DcarFor entrance vehicle is poor, DaxisFor the total number of axle of entrance vehicle it is poor,
DweightFor the total method of double differences of entrance vehicle goods, DlimitweightIt is poor for entrance vehicle weight limitation；
Statistic of classification goes out the entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis、
The total method of double differences D of entrance vehicle goodsweight, entrance vehicle weight limitation difference DlimitweightFeature distribution, and choose judge outlet data and
The whether matched correlation metric of entry data.
Optionally, in above-mentioned highway goods stock transprovincially matching process, the correlation metric is handled
To obtain candidate characteristic set, and significance level evaluation is carried out to obtain target signature Ji Buzhoubao to the candidate characteristic set
It includes：
With nondimensionalization, qualitative features quantification, quantitative characteristic binaryzation and discrete features coding method to coming in and going out
Mouth time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweight, come in and go out
Mouth vehicle weight limitation difference DlimitweightIt is handled, forms highway goods stock matching candidate feature set transprovincially；
The significance level of the candidate characteristic set is evaluated to obtain target using correlation coefficient process or variance back-and-forth method
Feature set.
Optionally, quantitative with nondimensionalization, qualitative features in above-mentioned highway goods stock transprovincially matching process
Change, quantitative characteristic binaryzation, One-Hot coding and/or discrete features coding method are to entrance time difference Dtime, entrance vehicle
Type difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweightIt is poor with entrance vehicle weight limitation
DlimitweightHandled, formed highway goods stock transprovincially matching candidate feature set the step of include：
To the entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxisIt is respectively adopted
The One-Hot codings are handled, and to the total method of double differences D of the entrance vehicle goodsweightIt is poor with entrance vehicle weight limitation
DlimitweightSection is carried out respectively to zoom in the section of [- 1,1] to obtain highway goods stock matching candidate spy transprovincially
Collection.
Optionally, in above-mentioned highway goods stock transprovincially matching process, to the training characteristics collection using a variety of
Default machine learning algorithm is trained study and obtains a variety of matching algorithm models transprovincially with corresponding, and from it is described it is a variety of transprovincially
Include with choosing a kind of in algorithm model as the step of the first algorithm model：
Respectively value, support vector machines, naive Bayesian, decision tree, random forest and ladder are closed on using logistic regression, K
Degree hoisting machine learning algorithm is trained study to the training characteristics collection and obtains corresponding model, and calculates the standard of each model
True rate score；
Using a corresponding model of accuracy rate highest scoring in each model as the first algorithm model.
Optionally, in above-mentioned highway goods stock transprovincially matching process, described in the test feature set pair
First algorithm model assess and handled first algorithm model to obtain target algorithm mould according to assessment result
The step of type includes：
It is tested using the first algorithm model described in test feature set pair, and draws learning curve, ROC curve, calculated
AUC value；
According to learning curve, ROC curve and AUC value, the fitting state of first algorithm model is judged；
It is adjusted according to the parameter of first algorithm model of fitting state pair and characteristic variable to obtain target algorithm
Model.
Optionally, in above-mentioned highway goods stock transprovincially matching process, according in the pending entry data
License plate number and pending outlet data in license plate number carry out similarity calculation to obtain similarity result, and be based on the phase
Include like the step of result optimizes the matching result is spent：
It is the vehicle in the license plate number and pending outlet data in the corresponding pending entry data of matching to matching result
The trade mark using JaroWinklerDistance algorithms carries out that similarity is calculated；
When the similarity is more than a preset value, then the matching result is maintained, otherwise, the matching result is carried out
Modification.
The present invention also provides a kind of highway goods stock transprovincially coalignments, including：
Acquisition module, entry data and outlet data for obtaining the corresponding goods stock of charge station transprovincially to prestore,
Matching degree generation based on the license plate number in the license plate number and entry data in the outlet data includes matched data and not
Sample data with data；
Computing module, for by the sample data entry data and the outlet data perform mathematical calculations with
It is counted to overall target, and to the overall target to obtain correlation metric；
Processing module, for being handled the correlation metric to obtain candidate characteristic set, and to described candidate special
Collection carries out significance level evaluation to obtain target signature collection, and is allocated to obtain training characteristics to the target signature collection
Collection and test feature collection；
Training module, for using a variety of default machine learning algorithms to be trained study with right the training characteristics collection
A variety of matching algorithm models transprovincially should be obtained, and choose a kind of the first algorithm of conduct from a variety of models of matching algorithm transprovincially
Model；
Evaluation module, for using the first algorithm model described in the test feature set pair to be assessed to obtain assessment knot
Fruit, and first algorithm model is adjusted to obtain target algorithm model according to the assessment result；
Matching module, for in charge station transprovincially pending entry data and pending outlet data use the mesh
Mark algorithm model is matched to obtain matching result；
Optimization module, for the license plate number in the license plate number and pending outlet data in the pending entry data
Carry out similarity calculation with obtain the license plate number in similarity result pending entry data corresponding to the matching result and
License plate number in pending outlet data carries out similarity calculation to obtain similarity result, and is based on the similarity result pair
The matching result optimizes.
A kind of highway goods stock provided by the invention transprovincially matching process and device are prestored transprovincially by obtaining
The entry data and outlet data of the corresponding goods stock of charge station, and target algorithm is obtained based on entry data and outlet data
Model, in charge station transprovincially pending entry data and pending outlet data use target algorithm model to be matched with
Matching result is obtained, and similar with the license plate number progress in pending outlet data according to the license plate number in pending entry data
Degree is calculated to obtain similarity result, and is optimized to the matching result based on the similarity result.It is set by above-mentioned
It sets with effective guarantee highway goods stock transprovincially matched accuracy, can effectively solve expressway tol lcollection data and transprovincially divide
Cut problem and since directly goods stock can not be carried out transprovincially by license plate number caused by Car license recognition is not complete or identification mistake
Matching problem, to obtain goods stock complete driving path on a highway.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment cited below particularly, and coordinate
Appended attached drawing, is described in detail below.
Description of the drawings
Fig. 1 is the structure diagram of terminal device provided in an embodiment of the present invention.
Fig. 2 is the flow diagram of highway goods stock provided in an embodiment of the present invention transprovincially matching process.
Fig. 3 is the flow diagram of step S110 in Fig. 2.
Fig. 4 is goods stock provided in an embodiment of the present invention matched ROC curve transprovincially.
Fig. 5 is the connection block diagram of highway goods stock provided in an embodiment of the present invention transprovincially coalignment.
Icon：10- terminal devices；12- memories；14- processors；100- highways goods stock transprovincially matches dress
It sets；110- acquisition modules；120- computing modules；130- processing modules；140- training modules；150- evaluation modules；160- is matched
Module：170- optimization modules.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment only
It is a part of the embodiment of the present invention, instead of all the embodiments.The present invention being usually described and illustrated herein in the accompanying drawings
The component of embodiment can be arranged and be designed with a variety of different configurations.
As shown in Figure 1, an embodiment of the present invention provides a kind of terminal device 10, including memory 12, processor 14 and height
Fast highway freight vehicle transprovincially coalignment 100.Wherein, the terminal device 10 can be but not limited to server, intelligent hand
Machine, PC (personal computer, PC), tablet computer etc. have the electronic equipment of data-handling capacity, herein not
Make specific limit.
In this embodiment, it is directly or indirectly electrically connected between the memory 12 and processor 14, to realize
The transmission or interaction of data.For example, these elements can realize electricity by one or more communication bus or signal wire between each other
Property connection.The highway goods stock transprovincially coalignment 100 include it is at least one can be with software or firmware
(firmware) form is stored in the software function module in the memory 12.The processor 14 is for executing described deposit
The executable module stored in reservoir 12, such as the highway goods stock transprovincially software included by coalignment 100
Function module and computer program etc., to realize highway goods stock matching process transprovincially.
Wherein, the memory 12 may be, but not limited to, random access memory (Random Access Memory,
RAM), read-only memory (Read Only Memory, ROM), programmable read only memory (Programmable Read-Only
Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM),
Electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only Memory, EEPROM) etc..
Wherein, memory 12 is for storing program, and the processor 14 executes described program after receiving and executing instruction.
The processor 14 may be a kind of IC chip, the processing capacity with signal.Above-mentioned processor 14
Can be general processor, including central processing unit (Central Processing Unit, CPU), network processing unit
(Network Processor, NP) etc.；It can also be digital signal processor (DSP), application-specific integrated circuit (ASIC), scene
Programmable gate array (FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware group
Part.It may be implemented or execute disclosed each method, step and the logic diagram in the embodiment of the present invention.General processor can be with
It is microprocessor or the processor can also be any conventional processor etc..
It is appreciated that structure shown in FIG. 1 is only to illustrate, the terminal device 10 may also include more than shown in Fig. 1
Either less component or with the configuration different from shown in Fig. 1.Hardware, software may be used in each component shown in Fig. 1
Or combinations thereof realize.
In conjunction with Fig. 2, the embodiment of the present invention also provides a kind of highway shipping can be applied to above-mentioned terminal device 10
Vehicle transprovincially matching process, the method includes the steps seven steps of S110- steps S170.
Step S110：The entry data and outlet data for obtaining the corresponding goods stock of charge station transprovincially to prestore, are based on
The matching degree of the license plate number in license plate number and entry data in the outlet data includes matched data and non-matched data
Sample data.
Wherein, the mode of the entry data and outlet data that obtain the corresponding goods stock of charge station transprovincially to prestore can be with
It is according to national freeway toll station position input by user and the corresponding charge station's information in Provincial administrative division boundary
As charge station's information transprovincially, and transprovincially corresponding outlet data and the entry data in charge station's information is obtained, can also be
Charge station's information transprovincially is obtained according in the high speeds highway network topological structure such as Amap and Tencent's map, and obtains this and transprovincially receives
Take the entry data and outlet data of the corresponding goods stock of station information.
It is optionally, in the present embodiment, described to obtain the corresponding goods stock of charge station transprovincially to prestore incorporated by reference to Fig. 2
Entry data and outlet data, based on the license plate number in the license plate number and entry data in the outlet data matching degree life
At including the steps that the sample data of matched data and non-matched data includes：
Step S112：It is obtained transprovincially according to freeway net topological structure, charge station location and provincial administrative area boundary
Charge station's information, and obtain the goods stock outlet data corresponding with the information of charge station transprovincially and entry data to prestore.
Step S114：Have the corresponding outlet data of complete license plate number as target outlet data using in outlet data, adopts
With preset algorithm from the entry data search with the complete license plate number in the target outlet data consistent and outlet data
In time and entry data in the difference of time be located at the entry data of a setting time range as target entries data,
And using the target entries data and with the target outlet data of the target entries Data Matching as matched data, by other data
As non-matched data.
It should be noted that in the present embodiment, the charge station transprovincially can inbound fission, can also be to come in and go out
Stand one, when it is described go out inbound fission when, the distance between the outlet station of the charge station transprovincially and access station should one compared with
In short distance range, such as in 20,30 or 40 kilometers.
The preset algorithm can be JaroWinkler Distance algorithms, can also be Levenshtein
Distance algorithms, are chosen according to actual demand, are not specifically limited herein.
Optionally, in the present embodiment, it is searched from the entry data using preset algorithm and the target outlet number
The difference of time of the complete license plate number in unanimously and in time and entry data in outlet data is located at a setting time
The entry data of range is as target entries data, and by the target entries data and target with the target entries Data Matching
Outlet data includes as matched data, using other data as the step of non-matched data：
License plate number L in outlet data is calculated using JaroWinklerDistance algorithmsOutletWith the vehicle in entry data
Trade mark LEntranceSimilarity Slicense：
Slicense=Sj+(lp(1-Sj))
Wherein, m LOutletAnd LEntranceMatched number of characters, t are the number of transposition；
It must not be less than V kilometers of standard per hour according to the minimum speed of highway, according to the distance D of charge station transprovincially,
Calculate the running time T from outlet charge station to entrance charge station：
T=(D/V) × 60
Entry data is screened based on the running time T, is subtracted when screening obtains the entry time in entry data
Go the Outlet time of outlet data in time interval [- T, T] range and license plate number similarity SlicenseMore than a setting value
When, judge that the corresponding vehicle of corresponding outlet data and the corresponding vehicle of entry data are same vehicles, and by the entry data
And it is included in respectively as the target entries data and target outlet data with the matched outlet data of the entry data
With data, other data are included in non-matched data.
Wherein, the number t of transposition be unmatched number of characters, the setting value can be but not limited to 0.8,0.9 or
0.95, and work as the similarity and be more than the setting value, then judge the car plate in the license plate number and entry data in outlet data
It number is consistent.
It should be noted that national freeway toll station and the spatial data on Provincial administrative division boundary will use
Unified space coordinates, avoid due to space coordinates are inconsistent and caused by position offset, calculating charge station transprovincially
When distance D, national freeway toll station and Provincial administrative division boundary can be superimposed upon in same map window, be adopted
The mode for manually extracting charge station transprovincially calls the tool of GIS platform to calculate the distance between charge station transprovincially.
Step S120：By in the sample data entry data and the outlet data perform mathematical calculations it is comprehensive to obtain
Index is closed, and the overall target is counted to obtain correlation metric.
Wherein, the outlet data can include but is not limited to：Export charge station coding, outlet license plate number, Outlet time,
Export vehicle, the total number of axle of outlet vehicle, outlet vehicle goods gross weight and outlet vehicle weight limitation, the entry data may include but not
Be limited to entrance charge station coding, entrance license plate number, entry time, entrance vehicle, the total number of axle of entrance vehicle, entrance vehicle goods gross weight with
And entrance vehicle weight limitation.
By in the sample data entry data and the outlet data perform mathematical calculations to obtain overall target, and
The step of being counted the overall target to obtain correlation metric include：
Utilize the entry time T in entry dataEntrance, entrance vehicle CEntrance, the total number of axle A of entrance vehicleEntrance, entrance vehicle goods it is total
Weight WEntranceAnd entrance vehicle weight limitation LWEntranceSubtract the Outlet time T in corresponding outlet dataOutlet, outlet vehicle COutlet, outlet vehicle
Total number of axle AOutlet, outlet vehicle goods gross weight WOutletAnd outlet vehicle weight limitation LWEntranceObtain overall target：
Dtime=TEntrance-TOutlet
Dcar=CEntrance-COutlet
Daxis=AEntrance-AOutlet
Dweight=WEntrance-WOutlet
Dlimitweight=LWEntrance-LWOutlet
Wherein, DtimeFor entrance time difference, DcarFor entrance vehicle is poor, DaxisFor the total number of axle of entrance vehicle it is poor,
DweightFor the total method of double differences of entrance vehicle goods, DlimitweightIt is poor for entrance vehicle weight limitation；
Statistic of classification goes out the entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis、
The total method of double differences D of entrance vehicle goodsweight, entrance vehicle weight limitation difference DlimitweightFeature distribution, and choose judge outlet data and
The whether matched correlation metric of entry data.
Step S130：The correlation metric is handled to obtain candidate characteristic set, and to the candidate characteristic set
Significance level evaluation is carried out to obtain target signature collection, and to the target signature collection be allocated to obtain training characteristics collection and
Test feature collection.
Wherein, the mode for training characteristics collection and test feature collection being obtained to the target signature and being allocated is can be with
It is to the target signature and to be allocated to obtain training characteristics collection and test feature collection according to a setting ratio, for example, can be with
It is to the target signature and to be allocated to obtain training characteristics collection and test feature collection according to 8: 2 ratio or 7: 3 ratio.
The correlation metric is handled to obtain candidate characteristic set, and important journey is carried out to the candidate characteristic set
Degree evaluation to include the step of obtaining target signature collection：
With nondimensionalization, qualitative features quantification, quantitative characteristic binaryzation and discrete features coding method to coming in and going out
Mouth time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweight, come in and go out
Mouth vehicle weight limitation difference DlimitweightIt is handled, forms highway goods stock matching candidate feature set transprovincially；
The significance level of the candidate characteristic set is evaluated to obtain target using correlation coefficient process or variance back-and-forth method
Feature set.
Specifically, in the present embodiment, to the entrance time difference Dtime, entrance vehicle difference Dcar, entrance vehicle
Total number of axle difference DaxisThe One-Hot codings are respectively adopted to be handled, and to the total method of double differences D of the entrance vehicle goodsweightWith
Entrance vehicle weight limitation difference DlimitweightSection is carried out respectively to zoom in the section of [- 1,1] to obtain highway freight
Matching candidate feature set transprovincially.
Step S140：A variety of default machine learning algorithms are used to be trained study with to deserved the training characteristics collection
To a variety of matching algorithm models transprovincially, and a kind of the first algorithm mould of conduct is chosen from a variety of models of matching algorithm transprovincially
Type.
Wherein, a variety of default machine learning algorithms can include but is not limited to logistic regression, K closes on value, support vector machines,
Naive Bayesian, decision tree, random forest and gradient are promoted.
To the training characteristics collection use a variety of default machine learning algorithms be trained study with it is corresponding obtain it is a variety of across
Matching algorithm model is saved, and chooses from a variety of models of matching algorithm transprovincially and a kind of to be wrapped as the step of the first algorithm model
It includes：
Respectively value, support vector machines, naive Bayesian, decision tree, random forest and ladder are closed on using logistic regression, K
Degree hoisting machine learning algorithm is trained study to feature set, and calculates the accuracy rate score of each model.
Using a corresponding model of accuracy rate highest scoring in each model as the first algorithm model.
Step S150：The first algorithm model described in the test feature set pair is used to be assessed to obtain assessment result,
And first algorithm model is adjusted to obtain target algorithm model according to the assessment result.
Specifically, the first algorithm model described in the test feature set pair is used to be assessed to obtain assessment result, and
First algorithm model is calibrated to include the step of obtaining target algorithm model according to the assessment result：
It is tested using the first algorithm model described in the test feature set pair, and draws learning curve, ROC curve,
Calculate AUC value.
According to learning curve, ROC curve and AUC value, the fitting state of first algorithm model is judged.
It is adjusted according to the parameter of first algorithm model of fitting state pair and characteristic variable to obtain target algorithm
Model.
Step S160：To in charge station transprovincially pending entry data and pending outlet data using the target calculate
Method model is matched to obtain matching result.
Specifically, by the way that the pending entry data and pending outlet data are input to the target algorithm mould
Type, so that the target algorithm model matches the outlet data and entry data.
Step S170：License plate number in pending entry data corresponding to the matching result and pending outlet data
In license plate number carry out similarity calculation to obtain similarity result, and based on the similarity result to the matching result into
Row optimization.
Wherein, in the license plate number and pending outlet data in pending entry data corresponding to the matching result
It can be to use JaroWinkler Distance algorithms in a manner of obtaining similarity result that license plate number, which carries out similarity calculation,
Similarity calculation is carried out, can also be that similarity calculation is carried out using Levenshtein Distance algorithms, according to practical need
It asks and is chosen, is not specifically limited herein.
Optionally, in the present embodiment, according in the pending entry data license plate number and pending outlet data
In license plate number carry out similarity calculation to obtain similarity result, and based on the similarity result to the matching result into
Row optimization the step of include：
It is the vehicle in the license plate number and pending outlet data in the corresponding pending entry data of matching to matching result
The trade mark using JaroWinklerDistance algorithms carries out that similarity is calculated.
When the similarity is more than a preset value, then the matching result is maintained, otherwise, the matching result is carried out
Modification.
Wherein, which can be 0.7,0.75,0.8,0.85 or 0.9, be not specifically limited herein, according to practical need
It asks and is configured.
By above-mentioned setting highway can be effectively solved effectively to realize the matching transprovincially of highway goods stock
Charge data transprovincially segmentation problem and due to Car license recognition is not complete or identification mistake caused by can not directly by license plate number into
Row goods stock transprovincially matching problem, to restore complete driving path of the goods stock on national highway, for high speed
The analysis decisions such as highway transportation statistics analysis, economic operation analysis provide base support.
In the present embodiment, include using Guizhou Xin Zhai charge stations as outlet station transprovincially and Guangxi with the charge station transprovincially
Qian Guiliuzhai charge stations for access station transprovincially as illustrating.Goods stock in June, 2017 Guizhou to prestore is obtained respectively
Xin Zhai charge stations outlet data and charge station of the Guangxi stockaded villages Qian Guiliu entry data, and reject that license plate number is empty, data target is empty
Record, specific entry data include entrance charge station number, entrance charge station name, entrance license plate number, entry time, enter
Mouth vehicle, the total number of axle of entrance vehicle, entrance vehicle goods gross weight and entrance vehicle weight limitation, outlet data include that outlet charge station compiles
Number, outlet charge station name, outlet license plate number, Outlet time, outlet vehicle, the total number of axle of outlet vehicle, outlet vehicle goods gross weight with
And outlet vehicle weight limitation.
Guizhou Xin Zhai charge stations in June, 2017 goods stock outbound data is refering to table 1：
Table 1
Charge station of the Guangxi stockaded villages Qian Guiliu in June, 2017 goods stock inbound data is refering to table 2：
Table 2
Wherein, license plate number is generally made of Chinese character+letter+number, wherein first Chinese character is the abbreviation in province, second letter
General proxy city, behind be made of letter or number, license plate number length be 7.According to the coding rule of license plate number, from outlet number
The complete license plate number record for meeting license plate number coding rule is filtered out in.
Calculate the transit time between Guizhou Xin Zhai charge stations and charge station of the Guangxi stockaded villages Qian Guiliu.It is charged according to Guizhou Xin Zhai
The road network structure stood with the latitude and longitude coordinates of charge station of the Guangxi stockaded villages Qian Guiliu and freeway net, can calculate Guizhou Xin Zhai
Charge station is 8.4 kilometers at a distance from charge station of the Guangxi stockaded villages Qian Guiliu.Do not considering freeway net topological structure abnormal conditions
Under, 60 kilometers per hour of standard, Guizhou Xin Zhai charge stations and Guangxi Guizhou Province osmanthus six must not be less than according to the minimum speed of highway
The transit time of charge station of stockaded village up to 8.4 minutes.
On the basis of the Outlet time in outlet data, entry time difference is filtered out from entry data in [- 9,9] minute
Interior entry record collection.
The similarity of outlet license plate number and entrance license plate number is calculated using JaroWinklerDistance algorithms.Calculate knot
Fruit is refering to table 3：
Table 3
If license plate number similarity is 0.9 or more, judgement outlet vehicle and entrance vehicle are same vehicles, are included in coupling number
According to conversely, judging that outlet vehicle and entrance vehicle for non-same vehicle, are included in non-matched data.Finally by matched data and not
The highway goods stock transprovincially matched sample data that matched data is constituted, please refers to table 4：
Table 4
By performing mathematical calculations to sample data middle outlet data and entry data, by entry and exit discretization index
Be converted to can compare, analyzable overall target.By the regularity of distribution of the analysis integrated index of the statistical graphs such as histogram, grasp
To judging, the whether matched correlation metric of entry data.It specifically calculates, the indicator difference of entry data, such as comes in and goes out
Mouth time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweight, come in and go out
Mouth vehicle weight limitation difference Dlimitweight.Result of calculation is refering to table 5：
Table 5
According to whether matching index, counts entry time difference Dtime, entrance vehicle difference Dcar, entrance vehicle line shaft
Number difference Daxis, the total method of double differences D of entrance vehicle goodsweight, entrance vehicle weight limitation difference DlimitweightData distribution.
Entrance time difference DtimeData distribution refering to table 6：
Table 6
Entrance vehicle difference DcarData distribution refering to table 7：
Table 7
| Entrance vehicle is poor | Whether match | Sample size |
| -4 | 0 | 84389 |
| -3 | 0 | 193407 |
| -2 | 0 | 101191 |
| -1 | 0 | 129205 |
| 0 | 0 | 1121537 |
| 1 | 0 | 143759 |
| 2 | 0 | 101843 |
| 3 | 0 | 140178 |
| 4 | 0 | 145883 |
| -4 | 1 | 6 |
| -3 | 1 | 1 |
| -2 | 1 | 5 |
| -1 | 1 | 109 |
| 0 | 1 | 42986 |
| 1 | 1 | 2061 |
| 2 | 1 | 30 |
| 3 | 1 | 11 |
The total number of axle difference D of entrance vehicleaxisData distribution refering to table 8：
Table 8
Entrance vehicle borrows total method of double differences DweightData distribution refering to table 9：
Table 9
Entrance vehicle weight limitation difference DlimitweightData distribution refering to table 10：
Table 10
According to table 6- tables 10：Entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle of entrance vehicle it is poor
Daxis, the total method of double differences D of entrance vehicle goodsweight, entrance vehicle weight limitation difference DlimitweightThere is significantly distribution to advise sample data
Rule, for example, sample data entrance time difference DtimeIt concentrates within the scope of [- 5,7] minute, entrance vehicle difference DcarIt concentrates on
In [- 1,1] range, the total number of axle difference D of entrance vehicleaxisIt concentrates in [- 1,1] range, the total method of double differences D of entrance vehicle goodsweight、
Entrance vehicle weight limitation difference DlimitweightAlso it all concentrates in a certain range.It is thereby possible to select entrance time difference Dtime, go out
Entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweight, entrance vehicle weight limitation it is poor
DlimitweightAs judging highway goods stock matched correlation metric transprovincially.
For entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxisThree indexs are adopted
It is encoded with One-Hot, converts Discrete Eigenvalue to the binary feature for including multiple mode bits；For the total method of double differences of entrance vehicle goods
Dweight, entrance vehicle weight limitation difference DlimitweightTwo indices carry out section scaling using MinMaxScalar, uniformly zoom to
In the section of [- 1,1], to form highway goods stock matching candidate feature set transprovincially, specifically include
WeightDiffefence, LimitWeightDifference, TD_-14, TD_-13, TD_12, TD_-11,
TD_10, TD_9, TD_8, TD_7, TD_6, TD_5, TD_4, TD_3, TD_2, TD_1, TD_0, TD_1, TD_2, TD_3, TD_4,
TD_5, TD_6, TD_7, TD_8, TD_9, TD_10, TD_11, TD_12, TD_13, TD_14, AD_-4, AD_-3, AD_-2, AD_-
1, AD_0, AD_1, AD_2, AD_3, AD_4, MD_4, MD_-3, MD_-2, MD_-1, MD_0, MD_1, MD_2, MD_3, MD_4 },
Wherein, TD_ beginnings are characterized as the entrance time difference using the index generated after One-Hot codings, and MD_ beginnings are characterized as out
For entrance vehicle difference using the index generated after One-Hot codings, AD_ beginnings are characterized as that the total number of axle difference of entrance vehicle uses
The index generated after One-Hot codings.
The significance level of candidate characteristic set is evaluated using Random Forest methods.Evaluation result please refers to table
11：
Table 11
According to the significance level ranking of feature, before ranking 10 feature is chosen as target signature collection, and according to 0.8 He
0.2 ratio is split target signature collection, and 80% data are for training so that as training characteristics collection, 20% data are used
In test to be modeled as test feature collection, and using training characteristics collection.
I.e. to include TD_0, WeightDifference, LimitWeightDifference, MD_0, AD_0, MD_1,
The target signature collection of AD_-2, AD-1, AD_1, MD-3 close on value using logistic regression, K, support vector machines, naive Bayesian, determine
The machine learning algorithms such as plan tree, random forest, gradient promotion are trained, and calculate predictablity rate.Result of calculation is refering to table
12：
Table 12
| Serial number | Algorithm title | Predictablity rate |
| 1 | RandomForest | 0.985375 |
| 2 | LogisticRegression | 0.975025 |
| 3 | KNN | 0.075936 |
| 4 | GradientBoosting | 0.986650 |
| 5 | AdaBoosting | 0.986975 |
| 6 | DecisionTree | 0.985550 |
| 7 | GaussianNativeByes | 0.900350 |
| 8 | SVC | 0.974725 |
As can be seen from Table 12, the predictablity rate of AdaBoosting algorithms compares other algorithm highers, therefore selects
First algorithm model of the AdaBoosting algorithms as highway goods stock transprovincially matching problem.
It is tested using the first algorithm model of test feature set pair, test result is refering to table 13：
Table 13
Specifically, by drawing learning curve, ROC curve, calculates the methods of AUC value and the first algorithm model is commented
Estimate, master of ceremonies's algorithm model model optimize accordingly and improve to obtain target algorithm model according to assessment result.
It is 0.9798609318093167 with the AUC value that AdaBoosting algorithms are calculated, ROC curve please refers to
Fig. 4 can be seen that AdaBoosting algorithms from ROC curve and AUC value and be highly suitable for solving highway goods stock
Matching problem transprovincially.
The target algorithm model in charge station of trained Guizhou Xin Zhai charge stations to the Guangxi stockaded villages Qian Guiliu in June, 2017 is answered
For differentiating whether the highway goods stock in July, 2017, August transprovincially matches, the generalization ability of testing model.Statistics knot
Fruit is shown in Table 14：
Table 14
Can be seen that model from 14 statistical result of table integrally has good generalization ability, but to the prediction of matched data
Accuracy rate is relatively low.
Therefore, solve the problems, such as that model is indifferent to exactly matching data generaliza-tion using license plate number similarity algorithm, into
One step improves highway goods stock matching precision transprovincially.Result of calculation is refering to table 15：
Table 15
If entry and exit license plate number similarity is more than 0.8, model is maintained to judge as a result, otherwise, then judging to tie by model
Fruit modification is modified.Therefore, indifferent to exactly matching data generaliza-tion using license plate number similarity algorithm solution model
Problem further increases highway goods stock matching precision transprovincially.Specially：To model be judged as it is matched as a result, by
According to license plate number similarity calculating method, the similarity of outlet license plate number and entrance car plate is calculated, if entry and exit license plate number is similar
Degree is more than a setting value, then maintains model to judge as a result, otherwise, then model judgement result is modified.
Incorporated by reference to Fig. 5, on the basis of the above, the present invention also provides a kind of highway goods stock transprovincially coalignments
100, including acquisition module 110, computing module 120, processing module 130, training module 140, evaluation module 150, matching module
160 and optimization module 170.
The acquisition module 110 is used to obtain entry data and the outlet of the corresponding goods stock of charge station transprovincially to prestore
Data, based on the license plate number in the license plate number and entry data in the outlet data matching degree generation include matched data and
The sample data of non-matched data.In the present embodiment, the acquisition module 110 can be used for executing step S110 shown in Fig. 2,
Specific descriptions about the acquisition module 110 are referred to the description to step S110 above.
The computing module 120 is used for the entry data and outlet data progress mathematics fortune in the sample data
It calculates to obtain overall target, and the overall target is counted to obtain correlation metric.In the present embodiment, the meter
It calculates module 120 and can be used for executing step S120 shown in Fig. 2, before the specific descriptions about the computing module 120 are referred to
Description of the text to step S120.
The processing module 130 is for handling the correlation metric to obtain candidate characteristic set, and to described
Candidate characteristic set carries out significance level evaluation to obtain target signature collection, and is allocated and is instructed to the target signature collection
Practice feature set and test feature collection.In the present embodiment, the processing module 130 can be used for executing step S130 shown in Fig. 2
It is referred to above to step S130 in the specific descriptions of the processing module 130.
The training module 140 using a variety of default machine learning algorithms to the training characteristics collection for being trained
It practises and a variety of matching algorithm models transprovincially is obtained with correspondence, and choose from a variety of models of matching algorithm transprovincially and a kind of being used as the
One algorithm model.In the present embodiment, the training module 140 can be used for executing step S140 shown in Fig. 2, about the instruction
The specific descriptions for practicing module 140 are referred to the description to step S140 above.
The evaluation module 150 is for using the first algorithm model described in the test feature set pair to be assessed to obtain
Assessment result, and first algorithm model is adjusted to obtain target algorithm model according to the assessment result.In this reality
It applies in example, the evaluation module 150 can be used for executing step S150 shown in Fig. 2, and specific about the evaluation module 150 is retouched
It states and is referred to the description to step S150 above.
The matching module 160 be used for in charge station transprovincially pending entry data and pending outlet data use
The target algorithm model is matched to obtain matching result.In the present embodiment, the matching module 160 can be used for executing
Step S160 shown in Fig. 2, the specific descriptions about the matching module 160 are referred to the description to step S160 above.
The optimization module 170 is used for in the license plate number and pending outlet data in the pending entry data
License plate number carries out similarity calculation to obtain the vehicle in similarity result pending entry data corresponding to the matching result
License plate number in the trade mark and pending outlet data carries out similarity calculation to obtain similarity result, and is based on the similarity
As a result the matching result is optimized.In the present embodiment, the optimization module 170 can be used for executing step shown in Fig. 2
Rapid S170, the specific descriptions about the optimization module 170 are referred to the description to step S170 above.
To sum up, a kind of highway goods stock provided by the invention transprovincially matching process and device is prestored by obtaining
The corresponding goods stock of charge station transprovincially entry data and outlet data, target is obtained based on entry data and outlet data
Algorithm model, in charge station transprovincially pending entry data and pending outlet data using target algorithm model carry out
It is equipped with to obtain matching result, and is carried out according to the license plate number in the license plate number and pending outlet data in pending entry data
Similarity calculation optimizes the matching result based on the similarity result with obtaining similarity result.By upper
Setting is stated with effective guarantee highway goods stock transprovincially matched accuracy, can effectively solve expressway tol lcollection data across
Save segmentation problem and since directly goods stock can not be carried out by license plate number caused by Car license recognition is not complete or identification mistake
Matching problem transprovincially.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in a computer read/write memory medium.Based on this understanding, technical scheme of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be expressed in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be
People's computer, electronic equipment or network equipment etc.) execute all or part of step of each embodiment the method for the present invention
Suddenly.
Claims (10)
1. a kind of highway goods stock transprovincially matching process, which is characterized in that the method includes：
The entry data and outlet data for obtaining the corresponding goods stock of charge station transprovincially to prestore, based in the outlet data
License plate number and entry data in license plate number matching degree generate include matched data and non-matched data sample data；
By in the sample data entry data and the outlet data perform mathematical calculations to obtain overall target, and to institute
Overall target is stated to be counted to obtain correlation metric；
The correlation metric is handled to obtain candidate characteristic set, and significance level is carried out to the candidate characteristic set and is commented
Valence is to obtain target signature collection, and is allocated to obtain training characteristics collection and test feature collection to the target signature collection；
To the training characteristics collection use a variety of default machine learning algorithms be trained study with it is corresponding obtain it is a variety of transprovincially
With algorithm model, and a kind of the first algorithm model of conduct is chosen from a variety of models of matching algorithm transprovincially；
The first algorithm model described in the test feature set pair is used to be assessed to obtain assessment result, and according to the assessment knot
Fruit is adjusted to obtain target algorithm model first algorithm model；
To in charge station transprovincially pending entry data and pending outlet data using the target algorithm model carry out
It is equipped with to obtain matching result；
The license plate number in license plate number and pending outlet data in pending entry data corresponding to the matching result into
Row similarity calculation optimizes the matching result based on the similarity result with obtaining similarity result.
2. highway goods stock according to claim 1 transprovincially matching process, which is characterized in that the acquisition prestores
The corresponding goods stock of charge station transprovincially entry data and outlet data, based in the outlet data license plate number with enter
The matching degree of license plate number in mouthful data generates the step of matched data and non-matched data and includes：
Charge station's information transprovincially is obtained according to freeway net topological structure, charge station location and provincial administrative area boundary, and
Obtain the goods stock outlet data corresponding with the information of charge station transprovincially and entry data to prestore；
There to be the corresponding outlet data of complete license plate number as target outlet data in outlet data, using preset algorithm from institute
State time and the entrance searched in entry data in and outlet data consistent with the complete license plate number in the target outlet data
The difference of time in data is located at the entry data of a setting time range as target entries data, and by the target entries
Data and with the target outlet data of the target entries Data Matching as matched data, using other data as mismatching number
According to.
3. highway goods stock according to claim 2 transprovincially matching process, which is characterized in that will be in outlet data
With the corresponding outlet data of complete license plate number as target outlet data, searched from the entry data using preset algorithm
The difference of the time in time and entry data with the complete license plate number in the target outlet data in consistent and outlet data
Value enters positioned at the entry data of a setting time range as target entries data, and by the target entries data and with the target
The target outlet data of mouthful Data Matching include using other data as the step of non-matched data as matched data：
License plate number L in outlet data is calculated using JaroWinklerDistance algorithmsOutletWith the license plate number in entry data
LEntranceSimilarity Slicense：
Slicense=Sj+(lp(1-Sj))
Wherein, m LOutletAnd LEntranceMatched number of characters, t are the number of transposition；
V kilometers per hour of standard must not be less than according to the minimum speed of highway, according to the distance D of charge station transprovincially, calculated
From outlet charge station to the running time T of entrance charge station：
T=(D/V) × 60
Entry data is screened based on the running time T, is subtracted out when screening obtains the entry time in entry data
The Outlet time of mouthful data is in time interval [- T, T] range and license plate number similarity SlicenseWhen more than a setting value, sentence
The corresponding vehicle of fixed corresponding outlet data and the corresponding vehicle of entry data are same vehicles, and by the entry data and with
The matched outlet data of the entry data is included in matched data respectively as the target entries data and target outlet data,
Other data are included in non-matched data.
4. highway goods stock according to claim 1 transprovincially matching process, which is characterized in that the entry data
Including entrance charge station coding, entrance license plate number, entry time, entrance vehicle, the total number of axle of entrance vehicle, entrance vehicle goods gross weight with
And entrance vehicle weight limitation, the outlet data include outlet charge station coding, outlet license plate number, Outlet time, outlet vehicle, go out
The total number of axle of mouthful vehicle, outlet vehicle goods gross weight and outlet vehicle weight limitation, by the sample data entry data and it is described go out
Mouth data perform mathematical calculations to obtain overall target, and are counted to the overall target to obtain the step of correlation metric
Suddenly include：
According to the entry time T in entry data in the sample dataEntrance, entrance vehicle CEntrance, the total number of axle A of entrance vehicleEntrance、
Entrance vehicle goods gross weight WEntranceAnd entrance vehicle weight limitation LWEntranceSubtract the Outlet time T in corresponding outlet dataOutlet, outlet vehicle
COutlet, outlet vehicle total number of axle AOutlet, outlet vehicle goods gross weight WOutletAnd outlet vehicle weight limitation LWEntranceObtain overall target：
Dtime=TEntrance-TOutlet
Dcar=CEntrance-COutlet
Daxis=AEntrance-AOutlet
Dweight=WEntrance-WOutlet
Dlimitweight=LWEntrance-LWOutlet
Wherein, DtimeFor entrance time difference, DcarFor entrance vehicle is poor, DaxisFor the total number of axle of entrance vehicle is poor, DweightFor
The total method of double differences of entrance vehicle goods, DlimitweightIt is poor for entrance vehicle weight limitation；
Statistic of classification goes out the entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, come in and go out
The total method of double differences D of mouth vehicle goodsweight, entrance vehicle weight limitation difference DlimitweightFeature distribution, and choose and judge outlet data and entrance
The whether matched correlation metric of data.
5. highway goods stock according to claim 4 transprovincially matching process, which is characterized in that the correlation
Index is handled to obtain candidate characteristic set, and carries out significance level evaluation to the candidate characteristic set to obtain target signature
The step of collection includes：
When with nondimensionalization, qualitative features quantification, quantitative characteristic binaryzation and discrete features coding method to entrance
Between difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweight, entrance vehicle
Freight weight limit difference DlimitweightIt is handled, forms highway goods stock matching candidate feature set transprovincially；
The significance level of the candidate characteristic set is evaluated to obtain target signature using correlation coefficient process or variance back-and-forth method
Collection.
6. highway goods stock according to claim 5 transprovincially matching process, which is characterized in that use dimensionless
Change, qualitative features quantification, quantitative characteristic binaryzation, One-Hot coding and/or discrete features coding method are to the entrance time
Poor Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxis, the total method of double differences D of entrance vehicle goodsweightWith entrance vehicle
Freight weight limit difference DlimitweightHandled, formed highway goods stock transprovincially matching candidate feature set the step of include：
To the entrance time difference Dtime, entrance vehicle difference Dcar, the total number of axle difference D of entrance vehicleaxisIt is respectively adopted described
One-Hot codings are handled, and to the total method of double differences D of the entrance vehicle goodsweightWith entrance vehicle weight limitation difference Dlimitweight
Section is carried out respectively to zoom in the section of [- 1,1] to obtain highway goods stock matching candidate feature set transprovincially.
7. highway goods stock according to claim 1 transprovincially matching process, which is characterized in that special to the training
Collection use a variety of default machine learning algorithms to be trained study and obtains a variety of matching algorithm models transprovincially with correspondence, and from institute
Stating the step of a kind of the first algorithm model of conduct is chosen in a variety of models of matching algorithm transprovincially includes：
Value, support vector machines, naive Bayesian, decision tree, random forest and gradient is closed on using logistic regression, K respectively to carry
Liter machine learning algorithm is trained study to the training characteristics collection and obtains corresponding model, and calculates the accuracy rate of each model
Score；
Using a corresponding model of accuracy rate highest scoring in each model as the first algorithm model.
8. highway goods stock according to claim 1 transprovincially matching process, which is characterized in that use the test
Feature set carries out assessment to first algorithm model and is handled first algorithm model to obtain according to assessment result
Include to the step of target algorithm model：
It is tested using the first algorithm model described in the test feature set pair, and draws learning curve, ROC curve, calculated
AUC value；
According to learning curve, ROC curve and AUC value, the fitting state of first algorithm model is judged；
It is adjusted according to the parameter of first algorithm model of fitting state pair and characteristic variable to obtain target algorithm model.
9. highway goods stock according to claim 1 transprovincially matching process, which is characterized in that wait locating according to described
Manage entry data in license plate number and pending outlet data in license plate number carry out similarity calculation to obtain similarity result,
And the step of being optimized to the matching result based on the similarity result, includes：
It is the license plate number in the license plate number and pending outlet data in the corresponding pending entry data of matching to matching result
It carries out that similarity is calculated using JaroWinklerDistance algorithms；
When the similarity is more than a preset value, then the matching result is maintained, otherwise, the matching result is repaiied
Change.
10. a kind of highway goods stock transprovincially coalignment, which is characterized in that including：
Acquisition module, entry data and outlet data for obtaining the corresponding goods stock of charge station transprovincially to prestore, is based on
The matching degree generation of the license plate number in license plate number and entry data in the outlet data includes matched data and mismatch number
According to sample data；
Computing module, for by the sample data entry data and the outlet data perform mathematical calculations it is comprehensive to obtain
Index is closed, and the overall target is counted to obtain correlation metric；
Processing module, for being handled the correlation metric to obtain candidate characteristic set, and to the candidate characteristic set
Significance level evaluation is carried out to obtain target signature collection, and to the target signature collection be allocated to obtain training characteristics collection and
Test feature collection；
Training module, for using a variety of default machine learning algorithms to be trained study with to deserved the training characteristics collection
To a variety of matching algorithm models transprovincially, and a kind of the first algorithm mould of conduct is chosen from a variety of models of matching algorithm transprovincially
Type；
Evaluation module, for using the first algorithm model described in the test feature set pair to be assessed to obtain assessment result,
And first algorithm model is adjusted to obtain target algorithm model according to the assessment result；
Matching module, for in charge station transprovincially pending entry data and pending outlet data using the target calculation
Method model is matched to obtain matching result；
Optimization module, for being carried out to the license plate number in the license plate number and pending outlet data in the pending entry data
Similarity calculation is to obtain the license plate number in similarity result pending entry data corresponding to the matching result and wait locating
License plate number in reason outlet data carries out similarity calculation to obtain similarity result, and based on the similarity result to described
Matching result optimizes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810664921.6A CN108764375B (en) | 2018-06-25 | 2018-06-25 | Highway goods stock transprovincially matching process and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810664921.6A CN108764375B (en) | 2018-06-25 | 2018-06-25 | Highway goods stock transprovincially matching process and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764375A true CN108764375A (en) | 2018-11-06 |
| CN108764375B CN108764375B (en) | 2019-05-03 |
Family
ID=63977506
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810664921.6A Active CN108764375B (en) | 2018-06-25 | 2018-06-25 | Highway goods stock transprovincially matching process and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764375B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109871513A (en) * | 2019-01-28 | 2019-06-11 | 重庆首讯科技股份有限公司 | A kind of current behavior similarity calculating method of vehicle on highway and system |
| CN110335473A (en) * | 2019-08-14 | 2019-10-15 | 中国联合网络通信集团有限公司 | Block the personal identification method and system of license plate vehicle |
| CN112419721A (en) * | 2020-11-18 | 2021-02-26 | 交通运输部科学研究院 | Method and device for calculating key indexes of transportation of vehicles on highway and electronic equipment |
| CN112560074A (en) * | 2021-02-20 | 2021-03-26 | 支付宝(杭州)信息技术有限公司 | Vehicle passing data processing method, device, equipment and system |
| CN112991134A (en) * | 2021-05-11 | 2021-06-18 | 交通运输部科学研究院 | Driving path reduction measuring and calculating method and device and electronic equipment |
| CN114446045A (en) * | 2021-09-14 | 2022-05-06 | 武汉长江通信智联技术有限公司 | Method for studying and judging illegal transportation behaviors of vehicles on highway in epidemic situation period |
| CN116821721A (en) * | 2023-07-03 | 2023-09-29 | 上海金润联汇数字科技有限公司 | Method, device, equipment and medium for identifying cross-city network about car |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593879A (en) * | 2013-10-08 | 2014-02-19 | 南京爱沓信息技术有限公司 | Vehicle positioning integrated intelligent charge management system and method applied to same |
| CN105844721A (en) * | 2016-01-27 | 2016-08-10 | 吴加强 | Non-stop charging method for driving in and out of expressway |
| US20170254660A1 (en) * | 2016-03-04 | 2017-09-07 | Volvo Car Corporation | Method and system for utilizing a trip history |
| CN107146292A (en) * | 2017-04-28 | 2017-09-08 | 成都通甲优博科技有限责任公司 | A kind of freeway toll station vehicle management server, system and method |
| CN107564294A (en) * | 2017-08-25 | 2018-01-09 | 深圳前海华夏智信数据科技有限公司 | Unlicensed car recognition methods and device based on virtual car plate |
| CN107730640A (en) * | 2016-08-10 | 2018-02-23 | 西安艾润物联网技术服务有限责任公司 | Bill settlement method and device |
-
2018
- 2018-06-25 CN CN201810664921.6A patent/CN108764375B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593879A (en) * | 2013-10-08 | 2014-02-19 | 南京爱沓信息技术有限公司 | Vehicle positioning integrated intelligent charge management system and method applied to same |
| CN105844721A (en) * | 2016-01-27 | 2016-08-10 | 吴加强 | Non-stop charging method for driving in and out of expressway |
| US20170254660A1 (en) * | 2016-03-04 | 2017-09-07 | Volvo Car Corporation | Method and system for utilizing a trip history |
| CN107730640A (en) * | 2016-08-10 | 2018-02-23 | 西安艾润物联网技术服务有限责任公司 | Bill settlement method and device |
| CN107146292A (en) * | 2017-04-28 | 2017-09-08 | 成都通甲优博科技有限责任公司 | A kind of freeway toll station vehicle management server, system and method |
| CN107564294A (en) * | 2017-08-25 | 2018-01-09 | 深圳前海华夏智信数据科技有限公司 | Unlicensed car recognition methods and device based on virtual car plate |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109871513A (en) * | 2019-01-28 | 2019-06-11 | 重庆首讯科技股份有限公司 | A kind of current behavior similarity calculating method of vehicle on highway and system |
| CN109871513B (en) * | 2019-01-28 | 2023-03-31 | 重庆首讯科技股份有限公司 | Method and system for calculating similarity of vehicle passing behaviors on highway |
| CN110335473A (en) * | 2019-08-14 | 2019-10-15 | 中国联合网络通信集团有限公司 | Block the personal identification method and system of license plate vehicle |
| CN110335473B (en) * | 2019-08-14 | 2021-05-11 | 中国联合网络通信集团有限公司 | Identity recognition method and system for vehicle with license plate covered |
| CN112419721A (en) * | 2020-11-18 | 2021-02-26 | 交通运输部科学研究院 | Method and device for calculating key indexes of transportation of vehicles on highway and electronic equipment |
| CN112419721B (en) * | 2020-11-18 | 2021-06-08 | 交通运输部科学研究院 | Method and device for calculating key indexes of transportation of vehicles on highway and electronic equipment |
| CN112560074A (en) * | 2021-02-20 | 2021-03-26 | 支付宝(杭州)信息技术有限公司 | Vehicle passing data processing method, device, equipment and system |
| CN112991134A (en) * | 2021-05-11 | 2021-06-18 | 交通运输部科学研究院 | Driving path reduction measuring and calculating method and device and electronic equipment |
| CN114446045A (en) * | 2021-09-14 | 2022-05-06 | 武汉长江通信智联技术有限公司 | Method for studying and judging illegal transportation behaviors of vehicles on highway in epidemic situation period |
| CN116821721A (en) * | 2023-07-03 | 2023-09-29 | 上海金润联汇数字科技有限公司 | Method, device, equipment and medium for identifying cross-city network about car |
| CN116821721B (en) * | 2023-07-03 | 2024-04-02 | 上海金润联汇数字科技有限公司 | Method, device, equipment and medium for identifying cross-city network about car |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764375B (en) | 2019-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764375B (en) | Highway goods stock transprovincially matching process and device | |
| CN108777004B (en) | Expressway coach transportation vehicle transprovincially matching process and device | |
| CN108648074A (en) | Loan valuation method, apparatus based on support vector machines and equipment | |
| WO2017143919A1 (en) | Method and apparatus for establishing data identification model | |
| CN111028016A (en) | Sales data prediction method and device and related equipment | |
| CN108062674A (en) | Order fraud recognition methods, system, storage medium and electronic equipment based on GPS | |
| CN107679734A (en) | It is a kind of to be used for the method and system without label data classification prediction | |
| CN103970747A (en) | Data processing method for network side computer to order search results | |
| CN107767273A (en) | Asset Allocation method, electronic installation and medium based on social data | |
| CN114386856A (en) | Method, device and equipment for identifying empty-shell enterprise and computer storage medium | |
| CN114566052A (en) | Method for judging rotation of highway traffic flow monitoring equipment based on traffic flow direction | |
| CN113704389A (en) | Data evaluation method and device, computer equipment and storage medium | |
| CN113379334B (en) | Road section bicycle riding quality identification method based on noisy track data | |
| CN115206102B (en) | Method, device, electronic equipment and medium for determining traffic path | |
| CN107291722B (en) | Descriptor classification method and device | |
| CN113360845A (en) | Vehicle source transaction probability prediction method and device, electronic device and storage medium | |
| CN106781482A (en) | Novel intelligent traffic data processing system | |
| CN112991134B (en) | Driving path reduction measuring and calculating method and device and electronic equipment | |
| CN116823069B (en) | Intelligent customer service quality inspection method based on text analysis and related equipment | |
| CN113837764B (en) | Risk early warning method, risk early warning device, electronic equipment and storage medium | |
| CN112784789B (en) | Method, device, electronic equipment and medium for identifying traffic flow of road | |
| CN107798446A (en) | The evaluation process method and device of rich ore route | |
| CN113673595A (en) | Data processing method, device and equipment | |
| CN115936282A (en) | Method and device for optimizing score model, electronic equipment and storage medium | |
| Jin et al. | Inspection Strategy Design of Toll-Free Logistics Vehicle in China Freeway: Case Study in Shaanxi Province |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |