CN107038858B - Commute private car dynamic share-car recommended method - Google Patents

Commute private car dynamic share-car recommended method Download PDF

Info

Publication number
CN107038858B
CN107038858B CN201710380104.3A CN201710380104A CN107038858B CN 107038858 B CN107038858 B CN 107038858B CN 201710380104 A CN201710380104 A CN 201710380104A CN 107038858 B CN107038858 B CN 107038858B
Authority
CN
China
Prior art keywords
car
vehicle
commuting
share
private car
Prior art date
Application number
CN201710380104.3A
Other languages
Chinese (zh)
Other versions
CN107038858A (en
Inventor
范晓亮
唐方
李军
王程
陈龙彪
臧彧
程明
Original Assignee
厦门大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 厦门大学 filed Critical 厦门大学
Priority to CN201710380104.3A priority Critical patent/CN107038858B/en
Publication of CN107038858A publication Critical patent/CN107038858A/en
Application granted granted Critical
Publication of CN107038858B publication Critical patent/CN107038858B/en

Links

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06QDATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/30Transportation; Communications

Abstract

The invention discloses a kind of commuting private car dynamic share-car recommended methods, including, license auto-recognition system data are inputted, it is pre-processed, generate track of vehicle collection;A kind of working day commuting private car recognizer is designed in conjunction with the high reproducibility and periodicity of commuting private car trip time and space idea based on city vehicle track collection;In conjunction with family-work address, commuting time window and the working day dynamic effects factor (weather, accident and traffic) of commuting private car, realizes and the key feature for influencing commuting private car share-car is extracted;Based on commuting vehicle space-time characteristic, match the commuting vehicle with similar commuting time and space idea, and combine influence of the dynamic effects factor to driver's share-car wish, these dynamic factors are effectively combined with user in the share-car demand of Spatial dimensionality, realize the dynamic share-car matching based on user intention, a kind of share-car mechanism steady in a long-term is provided for commuting private car, and significantly reduces private car vehicle number in peak period morning and evening.

Description

Commute private car dynamic share-car recommended method

Technical field

The present invention relates to intelligent transport system field, in particular to a kind of commuting private car dynamic share-car recommended method.

Background technique

With the rapid development based on location-based service and development of Mobile Internet technology, Ride-share service has become alleviation urban transportation Economic new model is shared in congestion.

Recommend field in share-car, traditional algorithm has two: first, and existing method is mostly to meet the real-time of passenger Share-car demand is simple target, i.e., how to reduce the waiting time of passenger, and is not yet effectively solved between share-car driver because same The problem of industry competes and increases private car usage amount.Therefore, existing share-car method is possible to cause more in peak period morning and evening Serious congestion in road.Second, existing share-car scheme depends on the mobile number of the large-scale crowds such as GPS track, mobile phone signaling mostly According to, track, abundant but semantic information is lacked, and not yet effectively considers the dynamics such as weather, festivals or holidays and occasion, traffic accident Great influence of the factor to share-car scheme.Therefore, it is difficult to provide personalized dynamic share-car scheme for user.

Summary of the invention

It is an object of the invention to overcome above-mentioned the deficiencies in the prior art, a kind of commuting private car dynamic share-car recommendation is provided Method.

To achieve the above object, the invention adopts the following technical scheme:

Commute private car dynamic share-car recommended method, comprising the following steps:

S1, input VLPR data set, pre-process VLPR data set, generate city vehicle track data collection;

S2, the high reproducibility based on commuting private car trip time and space idea, filter out from city vehicle track data concentration The low non-commuting vehicle of liveness carries out the analysis of family address and work address to remaining vehicle, will possess family-work simultaneously The vehicle of address is determined as the private car that commutes;

S3, family-work address in conjunction with the private car that commutes, commuting time window and working day dynamic effects factor difference are real Now space characteristics, temporal characteristics and the dynamic effects factor feature extraction to commuting private car share-car is influenced, the dynamic effects Factor includes weather, accident and traffic flow conditions;

S4, space characteristics and temporal characteristics based on commuting vehicle match the scheduled bus with similar commuting time and space idea , and influence of the dynamic effects factor to driver's share-car wish is combined, by these dynamic factors and user in Spatial dimensionality Share-car demand effectively combines, and realizes the dynamic share-car matching based on user intention.

Further, step S1 specifically includes the following steps:

S11, local private car is extracted, by license plate number and license plate color feature, is extracted from VLPR data set local private Family's car data;

S12, field are refined, and extract required field from local private savings car data, generate vehicle record data;

S13, redundancy cross vehicle record cleaning, deleted vehicle record data in repetition vehicle record and incorrect recording data;

S14, track of vehicle data set is generated, crosses vehicle record to after cleaning, classifies according to license plate number, identical license plate number note Vehicle time-sequencing was pressed in record, and identical license plate is then crossed vehicle record, and first place is connected in chronological order, formed track of vehicle data.

Further field needed for step S12 includes VLPR number, VLPR type, the direction VLPR, license plate number, license plate Color, lane number and mistake vehicle time.

Further, step S2 specifically includes the following steps:

S21, low liveness vehicle filtering define labourer hired by the month based on the periodicity and repeatability of commuting private car trip rule Make to enliven private car of the number of days less than 15 days in day to be the too low vehicle of liveness, the judgement too low vehicle of liveness is non-commuting vehicle And it is excluded;

S22, potential address for determining potential commuting private vehicle and potential work address, to determine the potential " family of vehicle Address ", analysis is come off duty to the track of vehicle in work hours next day section, L daily1, L2..., LnFor track VLPR equipment Address sequence, t1, t2..., tn-1Pass through the time interval sequence per continuous two VLPR equipment, t for vehiclem=max (ti), i =1,2 ..., n-1, works as tmL is determined greater than given thresholdm+1For its potential address;To determine the potential " place of working of vehicle Location ", working daily is analyzed to the track of vehicle in next time interval, L1, L2..., LnFor the track device address VLPR sequence Column, t1, t2..., tn-1Pass through the time interval sequence per continuous two VLPR equipment for vehicle, works as tmIt is greater than given threshold Determine LmIt is potential " work address " for the vehicle;

S23, vehicle man address and work address are determined, for potential address of each potential commuting private car, calculated Frequency freq (the PH of each potential addressi), the potential address of mark frequency highest and frequency greater than 0.8 is family address, For potential work address, its frequency freq (PW from fixed address is calculatedi), while calculating it apart from family Distance dist (the PW of locationi), determination possesses maximum value max (freq (PWi)×dist(PWi)) potential work address be real Work address;

S24, it determines commuting private car, determine while possessing family-work address private car for the private car that commutes.

Further, step S3 specifically includes the following steps:

S31, space characteristics extract, and the longitude and latitude of the family address of each commuting private car and work address is private as commuting Family's vehicle space characteristics;

S32, temporal characteristics extract, using multiple working days be used as observation time range, by departure time every workday with Arrival time respectively as one-dimensional characteristic, in order to which by time quantization, the departure time is converted to its hour apart from same day 00:00 Number, arrival time is converted to the hourage of its distance 06:00, if the same day morning peak vehicle without commuting track, when setting out Between with arrival time with 0 replace;

S33, dynamic effects factor feature extraction, it is contemplated that the dynamic effects such as traffic flow, traffic accident and weather condition because The influence that element recommends share-car, the dynamic effects factor feature for extracting every workday morning peak respectively will be for traffic flow No Monday, as one-dimensional characteristic otherwise it was 0 that being, which is 1,;For traffic accident, the traffic thing that same day morning peak Xiamen City is occurred Therefore sum is used as Accident Characteristic;For weather, consider that the visibility of same day morning peak and rainfall are special as bidimensional weather respectively Sign.

Further, in step S31, consider close starting point and the importance of terminal when share-car, prevent other feature poly- Excessive interference is generated in class algorithm, and longitude is converted to the difference of itself and all VLPR equipment minimum longitudes, latitude is converted For the difference of itself and all VLPR equipment minimum latitudes, by the longitude and latitude after conversion simultaneously multiplied by 1000, to increase space characteristics Weight.

Further, step S4 specifically includes the following steps:

S41, be by the space characteristics of each commuting private car and temporal characteristics description commuting private car commuting it is regular when Empty feature vector carries out coarseness cluster to commuting private car with k-means algorithm, makes the commuting with similar commuting rule Vehicle gathers for one kind.

S42, according to dynamic effects because usually determine driver to passenger provide Ride-share service willingness factor IF, specifically For,

IF=ω1×isMonday+ω2×(-visibility)+ω3×rainfall+ω4× accident wherein, IsMonday indicates whether Monday, and being is 1, and no is 0;Visibility indicates morning peak visibility;Rainfall indicates early high Peak rainfall;Accident indicates morning peak traffic accident sum;ω1, ω2, ω3, with ω4Four factors are respectively represented to logical The influence degree of diligent private car trip rule,Wherein visibility visibili and impact factor are negatively correlated, therefore Coefficient is negative;

S43, under different willingness factor effects, in conjunction with different shared seating capacity seats ∈ [Isosorbide-5-Nitrae], with recurrence AGNES algorithm generates the commuting private car share-car combination under different sharing seating capacity.

Further, one is defined first to acquire most preferably shared seating capacity under Different Dynamic influence factor in step S43 Energy function E, formula is as follows,

E (IF, seats)=D (IF, seats)+α S (IF, seats)

In formula, IF is willingness factor, and seats is to be ready shared amount of seats, and data item D (IF, seats) is to subtract after share-car Few vehicle number, smooth item S (IF, seats) are share-car accuracy and influence of passenger's average latency to final share-car effect, α is balance factor, α ∈ [0,1];

The calculation formula of data item D (IF, seats) and smooth item S (IF, seats) are as follows,

D (IF, seats)=N-M (IF, seats);

In formula, N is total vehicle number, and M (IF, seats) is vehicle combination number, and WT (IF, seats) is passenger's average waiting Time, RA (IF, seats) are average share-car accuracy,It is adjustable parameter with β, for determining share-car accuracy and passenger etc. Weight to time effects share-car effect;

To each willingness factor IF, taking seats is so that energy function E (if, seats) is the shared seat of maximum value Number, then have,

Seats=argmaxE (if, seats), seats ∈ [Isosorbide-5-Nitrae].

After adopting the above technical scheme, compared with the background technology, the present invention, having the advantages that and on the one hand improving passenger's On the other hand share-car satisfaction, i.e. reduction passenger waiting time and guarantee share-car accuracy guarantee the comprehensive society of Ride-share service Benefit significantly reduces share-car vehicle number in peak period morning and evening.

Detailed description of the invention

Fig. 1 is flow chart of the present invention;

Fig. 2 is VLPR data set, and wherein Fig. 2 (a) is Xiamen City VLPR device distribution situation, and Fig. 2 (b) is Xiamen City VLPR Facility information exemplary diagram, Fig. 2 (c) are that VLPR crosses vehicle record exemplary diagram, and Fig. 2 (d) is that VLPR equipment saves example of fields figure;

Fig. 3 is Xiamen City's weather history and casualty data exemplary diagram, and wherein Fig. 3 (a) is Xiamen City's 1 day to 5 May in 2016 Month weather condition exemplary diagram on the 4th, Fig. 3 (b) is Xiamen City's part traffic accident data instance figure in 2016;

Fig. 4 is original VLPR data set by pretreatment generation track of vehicle exemplary diagram;

Fig. 5 is part commuting vehicle " family address " and " work address " exemplary diagram;

Fig. 6 is part commuting private car key feature exemplary diagram;

Fig. 7 is to carry out coarseness to part commuting private car to cluster exemplary diagram;

Fig. 8 is the private car cluster that commutes to some coarseness, presses and does not carry out the thin cluster instance graph of dynamic on the same day;

Fig. 9 is commuting private car dynamic share-car recommended method analysis of experimental results figure, and wherein Fig. 9 (a) is that each working day is flat Time distribution map is waited, Fig. 9 (b) is each working day share-car accuracy distribution map, and Fig. 9 (c) is to reduce vehicle number on each working day Distribution map.

Specific embodiment

In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and It is not used in the restriction present invention.

Embodiment

As shown in Fig. 1 flow chart of the invention mainly includes following four step:

S1, input license auto-recognition system (Vehicle License Plate Recognition, abbreviation VLPR) data Collection, pre-processes VLPR data, generates city vehicle track data collection;

S2, the high reproducibility based on commuting private car trip time and space idea, filter out from city vehicle track data concentration Apparent non-commuting vehicle, the analysis of family address and work address is carried out to remaining vehicle, will possess family-work address simultaneously Vehicle be determined as the private car that commutes;

S3, family-work address in conjunction with the private car that commutes, commuting time window and working day dynamic effects factor difference are real Now space characteristics, temporal characteristics and the dynamic effects factor feature extraction to commuting private car share-car is influenced, the dynamic effects Factor includes weather, accident and traffic flow conditions;

S4, space characteristics and temporal characteristics based on commuting vehicle match the scheduled bus with similar commuting time and space idea , and influence of the dynamic effects factor to driver's share-car wish is combined, by these dynamic factors and user in Spatial dimensionality Share-car demand effectively combines, and realizes the dynamic share-car matching based on user intention.

Wherein, step S1 specifically includes the following steps:

S11, local private car is extracted.By license plate number and license plate color feature, extracted from all VLPR record data We need type of vehicle, i.e., local private car;

S12, field are refined.Original VLPR record includes 22 fields, 7 fields needed for refining this method, including, VLPR number, the direction VLPR, license plate number, license plate color, lane number, spends the vehicle time at VLPR type;

S13, redundancy cross vehicle record cleaning.Due to the defect of license auto-recognition system equipment, VLPR equipment, which exists, to be repeated to record The case where license board information, be subject to for the first time shooting cross vehicle information, delete repeat record.In addition, incorrect recording data also can It is cleaned;

S14, track of vehicle data set is generated.Vehicle record is crossed to after cleaning, is classified according to license plate number, identical license plate number note Vehicle time-sequencing was pressed in record, and identical license plate is then crossed vehicle record, and first place is connected in chronological order, formed track of vehicle data.

Fig. 2 is VLPR data set exemplary diagram, and wherein Fig. 2 a is Xiamen City VLPR device distribution situation, and Fig. 2 b is Xiamen City VLPR facility information exemplary diagram, Fig. 2 c are that VLPR crosses vehicle record exemplary diagram, and Fig. 2 d is that VLPR equipment saves example of fields figure;Fig. 4 Track of vehicle exemplary diagram is generated by pretreatment for original VLPR data set.

On the basis of step S1 obtains track of vehicle data set, step S2 is carried out, to identify city commuting private car, In include:

S21, low liveness vehicle filtering.In view of the periodicity and repeatability of commuting private car trip rule, liveness Too low vehicle can be excluded, therefore filter the private of low liveness (i.e. month 20 working days enlivened number of days less than 15 days) Family's vehicle, to exclude apparent non-commuting vehicle.

S22, potential address and work address are determined.In order to determine potential address of vehicle, (afternoon every night is analyzed 6 points to 9 time intervals in morning next day) track of vehicle, L1, L2..., LnFor the track device address VLPR sequence, t1, t2..., tn-1Pass through the time interval sequence per continuous two VLPR equipment, t for vehiclem=max (ti), i=1,2 ..., n-1 work as tm > 6h, then it is assumed that Lm+1For its potential address;Work address potential for vehicle analyzes (6:00 AM to afternoon 9 on daily daytime Point time interval) track of vehicle, L1, L2..., LnFor the track device address VLPR sequence, t1, t2..., tn-1For vehicle process In view of a part commuting vehicle Work break (ratio can occur at noon for the time interval sequence per continuous two VLPR equipment Such as eat out, go home halfway), if tm> 2h, and LmThe vehicle record moment is spent earlier than 11 points of the morning, it is believed that LmFor the vehicle Potential work address.

S23, family address and work address are determined.For potential address of each potential commuting private car, calculate each Frequency freq (the PH of potential addressi), the potential address of mark frequency highest and frequency greater than 0.8 is family address.For Potential " work address " calculates it from the frequency freq (PW of determining family addressi), while it is calculated apart from family address Distance dist (PWi), i.e. oneself address to VLPR number of devices between potential work address track, finally, determination possesses maximum Value max (freq (PWi)×dist(PWi)) potential work address be real work address.

S24, commuting private car is determined.Commuting private car is the private car for possessing " family-work address " pair simultaneously, oneself Location to the track between work address is the track that commutes.

Fig. 5 is part commuting vehicle man address and work address exemplary diagram.

On the basis of step S2 acquisition commuting private car, step S3 is carried out, to realize to influence commuting private car share-car Key feature extract, specifically includes the following steps:

S31, space characteristics extract.By " family address " and " work address " longitude and latitude of each commuting private car, as logical Diligent private car space characteristics, it is contemplated that the importance of close terminus when share-car prevents other feature from generating in clustering algorithm Excessive interference, this method increase the weight of space characteristics: longitude is converted to the difference of itself and all VLPR equipment minimum longitudes Latitude, is converted to the difference of itself and all VLPR equipment minimum latitudes by value, by the longitude and latitude after conversion simultaneously multiplied by 1000, with Increase weight.

S32, temporal characteristics extract.In order to reflect the commuting rule of commuting private car comprehensively, using multiple working days as sight Time range is surveyed, using departure time every workday and arrival time as one-dimensional characteristic, in order to by time quantization, we Departure time is converted to its hourage apart from same day 00:00 by method, and arrival time is converted to the hour of its distance 06:00 Number, if the same day morning peak vehicle is replaced without commuting track, departure time and arrival time with 0.

S33, dynamic effects factor feature extraction.In view of the dynamic effects such as traffic flow, traffic accident and weather condition because The influence that element recommends share-car, this method extract the dynamic effects factor feature of every workday morning peak respectively.For traffic Stream, since all a pair of of traffic impacts are larger, will whether Monday is as one-dimensional characteristic, otherwise it is 0 that being, which is 1,;For traffic thing Therefore using same day morning peak, (the traffic accident sum that 7:00 to the Xiamen City 9:00) occurs is as Accident Characteristic;For weather, respectively The visibility and rainfall for considering same day morning peak are as bidimensional weather characteristics.

Fig. 6 is part commuting private car key feature exemplary diagram.

On the basis of step S3 is extracted commuting private car key feature, step S4 is carried out, is anticipated with realizing based on user The dynamic share-car of hope matches, specifically includes the following steps:

S41, be by the space characteristics of each commuting private car and temporal characteristics description commuting private car commuting it is regular when Empty feature vector (if observing time range is d working day, space-time characteristic total (4+2d) dimension), with k-means algorithm to logical Diligent private car carries out coarseness cluster, gathers the commuting vehicle with similar commuting rule for one kind.

S42, for the private car that commutes, different dynamic effects factors (weather, the magnitude of traffic flow, traffic accident etc.) item Under part, size is not to the wish (being ready shared amount of seats) of passenger (Riders) offer Ride-share service by driver (Drivers) Together, the share-car combination finally generated is also different.First, it would be desirable to it defines driver and the willingness factor of Ride-share service is provided, and According to dynamic effects because usually determining that willingness factor IF, formula are as follows:

IF=ω1×isMonday+ω2×(-visibility)+ω3×rainfall+ω4× accident wherein, IsMonday indicates whether Monday, and being is 1, and no is 0;Visibility indicates morning peak visibility;Rainfall indicates early high Peak rainfall;Accident indicates morning peak traffic accident sum.ω1, ω2, ω3, with ω4Four factors are respectively represented to logical The influence degree of diligent private car trip rule,Wherein visibility visibility and impact factor are in negative It closes, therefore coefficient is negative.

S43, under different willingness factor effects, in conjunction with different shared seating capacity seats ∈ [Isosorbide-5-Nitrae], with recurrence AGNES algorithm generates the commuting private car share-car combination under different sharing seating capacity.

To acquire most preferably shared seating capacity under Different Dynamic influence factor in step S43, an energy function is defined first E, formula is as follows,

E (IF, seats)=D (IF, seats)+α S (IF, seats)

In formula, IF is willingness factor, and seats is to be ready shared amount of seats, and data item D (IF, seats) is to subtract after share-car Few vehicle number, smooth item S (IF, seats) are share-car accuracy and influence of passenger's average latency to final share-car effect, α is balance factor, α ∈ [0,1];

The calculation formula of data item D (IF, seats) and smooth item S (IF, seats) are as follows,

D (IF, seats)=N-M (IF, seats);

In formula, N is total vehicle number, and M (IF, seats) is vehicle combination number, and WT (IF, seats) is passenger's average waiting Time, RA (IF, seats) are average share-car accuracy,It is adjustable parameter with β, for determining share-car accuracy and passenger etc. Weight to time effects share-car effect;Share-car target is to guarantee that D (if, seats) and RA (IF, seats) is big as far as possible, WT (IF, Seats) small as far as possible, therefore WT (IF, seats) coefficient is made to be negative.

To each willingness factor IF, taking seats is so that energy function E (if, seats) is the shared seat of maximum value Number, then have,

Seats=argmaxE (if, seats), seats ∈ [Isosorbide-5-Nitrae].

Fig. 3 is Xiamen City's weather history and casualty data exemplary diagram, and wherein Fig. 3 a is Xiamen City's May 1 to May in 2016 Weather condition exemplary diagram on the 4th, Fig. 3 b are Xiamen City's part traffic accident data instance figures in 2016.

Fig. 7 is to carry out coarseness to part commuting private car to cluster exemplary diagram.

Fig. 8 is the private car cluster that commutes to some coarseness, presses and does not carry out the thin cluster instance graph of dynamic on the same day.

Fig. 9 is commuting private car dynamic share-car recommended method analysis of experimental results figure, and wherein Fig. 9 a is that each work is per day Serving-time distribution figure, average daily waiting time are 7.45 minutes, and Fig. 9 b is each working day share-car accuracy distribution map, average daily share-car Accuracy is that 83%, Fig. 9 c is to reduce vehicle number distribution map each working day, average daily to reduce 5478, accounts for about commuting private car total amount 30%.

From being can be seen that in above-mentioned figure using after share-car recommended method proposed by the present invention, performance has been obtained significantly Promotion: on the one hand ensure that the share-car satisfaction (lesser waiting time and higher share-car accuracy) of user, another party Face significantly reduces the commuting private car usage quantity of peak period morning and evening.

More than, it is merely preferred embodiments of the present invention, but scope of protection of the present invention is not limited thereto, it is any In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by those familiar with the art, all answers It is included within the scope of the present invention.Therefore, protection scope of the present invention should be subject to the protection scope in claims.

Claims (6)

1. commute private car dynamic share-car recommended method, which comprises the following steps:
S1, input VLPR data set, pre-process VLPR data set, generate city vehicle track data collection;
S2, the high reproducibility based on commuting private car trip time and space idea, filter out active from city vehicle track data concentration Low non-commuting vehicle is spent, the analysis of family address and work address is carried out to remaining vehicle, family-work address will be possessed simultaneously Vehicle be determined as the private car that commutes;
S3, it is realized respectively pair in conjunction with family-work address, commuting time window and the working day dynamic effects factor of commuting private car Influence space characteristics, temporal characteristics and the dynamic effects factor feature extraction of commuting private car share-car, the dynamic effects factor Including weather, accident and traffic flow conditions;
S4, space characteristics and temporal characteristics based on commuting vehicle match the commuting vehicle with similar commuting time and space idea, and Influence in conjunction with dynamic effects factor to driver's share-car wish needs these dynamic factors and user in the share-car of Spatial dimensionality Effective combination is asked, realizes the dynamic share-car matching based on user intention;
Wherein, step S4 specifically includes the following steps:
It S41, is that the regular space-time of description commuting private car commuting is special by the space characteristics of each commuting private car and temporal characteristics Vector is levied, coarseness cluster is carried out to commuting private car with k-means algorithm, makes the commuting vehicle with similar commuting rule Gather for one kind;
S42, according to dynamic effects because usually determine driver to passenger provide Ride-share service willingness factor IF, specifically,
IF=ω1×isMonday+ω2×(-visibility)+ω3×rainfall+ω4× accident wherein, IsMonday indicates whether Monday, and being is 1, and no is 0;Visibility indicates morning peak visibility;Rainfall indicates early high Peak rainfall;Accident indicates morning peak traffic accident sum;ω1, ω2, ω3, with ω4Four factors are respectively represented to logical The influence degree of diligent private car trip rule,Wherein visibility visibility and impact factor are negatively correlated, Therefore coefficient is negative;
S43, under different willingness factor effects, in conjunction with different shared seating capacity seats ∈ [Isosorbide-5-Nitrae], with recurrence AGNES Algorithm generates the commuting private car share-car combination under different sharing seating capacity;
Wherein, an energy function is defined first to acquire most preferably shared seating capacity under Different Dynamic influence factor in step S43 E, formula is as follows,
E (IF, seats)=D (IF, seats)+α S (IF, seats)
In formula, IF is willingness factor, and seats is to be ready shared amount of seats, and data item D (IF, seats) is to reduce vehicle after share-car Number, smooth item S (IF, seats) are share-car accuracy and influence of passenger's average latency to final share-car effect, and α is Balance factor, α ∈ [0,1];
The calculation formula of data item D (IF, seats) and smooth item S (IF, seats) are as follows,
D (IF, seats)=N-M (IF, seats);
In formula, N is total vehicle number, and M (IF, seats) is vehicle combination number, when WT (IF, seats) is passenger's average waiting Between, RA (IF, seats) is average share-car accuracy,It is adjustable parameter with β, for determining that share-car accuracy and passenger wait The weight of time effects share-car effect;
To each willingness factor IF, take seats be so that energy function E (IF, seats) is the shared seating capacity of maximum value, Then have,
Seats=argmaxE (IF, seats), seats ∈ [Isosorbide-5-Nitrae].
2. commuting private car dynamic share-car recommended method according to claim 1, it is characterised in that: step S1 is specifically included Following steps:
S11, local private car is extracted, by license plate number and license plate color feature, local private car is extracted from VLPR data set Data;
S12, field are refined, and extract required field from local private savings car data, generate vehicle record data;
S13, redundancy cross vehicle record cleaning, deleted vehicle record data in repetition vehicle record and incorrect recording data;
S14, track of vehicle is generated, crosses vehicle record to after cleaning, classify according to license plate number, when identical license plate number record pressed vehicle Between sort, then identical license plate crossed to vehicle record is the first in chronological order to be connected, form track of vehicle data.
3. commuting private car dynamic share-car recommended method according to claim 2, it is characterised in that: needed in step S12 Field include VLPR number, VLPR type, the direction VLPR, license plate number, license plate color, lane number and cross the vehicle time.
4. commuting private car dynamic share-car recommended method according to claim 1, which is characterized in that step S2 is specifically included Following steps:
S21, low liveness vehicle filtering define moon working day based on the periodicity and repeatability of commuting private car trip rule In to enliven private car of the number of days less than 15 days be the too low vehicle of liveness, determine that the too low vehicle of liveness is non-commuting vehicle and to give To exclude;
S22, potential address for determining potential commuting private vehicle and potential work address, to determine the potential " family of vehicle Location ", analysis are come off duty to the track of vehicle in work hours next day section, L daily1, L2..., LnFor track VLPR equipment Location sequence, t1, t2..., tn-1Pass through the time interval sequence per continuous two VLPR equipment, t for vehiclem=max (ti), i= 1,2 ..., n-1, works as tmL is determined greater than given thresholdm+1For its potential address;To determine that vehicle is potential " work address ", Analysis is gone to work daily to the track of vehicle in next time interval, L1, L2..., LnFor the track device address VLPR sequence, t1, t2..., tn-1Pass through the time interval sequence per continuous two VLPR equipment for vehicle, works as tmDetermine greater than given threshold LmIt is potential " work address " for the vehicle;
S23, vehicle man address and work address are determined, for potential address of each potential commuting private car, calculated each Frequency freq (the PH of potential addressi), the potential address of mark frequency highest and frequency greater than 0.8 is family address, for Potential work address calculates its frequency freq (PW from fixed addressi), while it is calculated apart from family address Distance dist (PWi), determination possesses maximum value max (freq (PWi)×dist(PWi)) potential work address be real work Make address;
S24, it determines commuting private car, determine while possessing family-work address private car for the private car that commutes.
5. commuting private car dynamic share-car recommended method according to claim 1, which is characterized in that step S3 is specifically included Following steps:
S31, space characteristics extract, using the longitude and latitude of the family address of each commuting private car and work address as commuting private car Space characteristics;
S32, temporal characteristics extract, using multiple working days as observation time range, by departure time every workday and arrival Time respectively as one-dimensional characteristic, in order to which by time quantization, the departure time is converted to its hourage apart from same day 00:00, Arrival time is converted to the hourage of its distance 06:00, if the same day morning peak vehicle without commuting track, the departure time with Arrival time is replaced with 0;
S33, dynamic effects factor feature extraction consider traffic flow, traffic accident and weather condition dynamic effects factor to share-car The dynamic effects factor feature of every workday morning peak is extracted in the influence of recommendation respectively, will whether Monday is made for traffic flow For one-dimensional characteristic, otherwise it is 0 that being, which is 1,;For traffic accident, the traffic accident sum that same day morning peak is occurred is as accident Feature;For weather, the visibility and rainfall for considering same day morning peak respectively are as bidimensional weather characteristics.
6. commuting private car dynamic share-car recommended method according to claim 5, it is characterised in that: in step S31, consider Close starting point and the importance of terminal, prevent other feature from generating excessive interference in clustering algorithm when share-car, and longitude is turned It is changed to the difference of itself and all VLPR equipment minimum longitudes, latitude is converted to the difference of itself and all VLPR equipment minimum latitudes Value, by the longitude and latitude after conversion simultaneously multiplied by 1000, to increase the weight of space characteristics.
CN201710380104.3A 2017-05-25 2017-05-25 Commute private car dynamic share-car recommended method CN107038858B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710380104.3A CN107038858B (en) 2017-05-25 2017-05-25 Commute private car dynamic share-car recommended method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710380104.3A CN107038858B (en) 2017-05-25 2017-05-25 Commute private car dynamic share-car recommended method

Publications (2)

Publication Number Publication Date
CN107038858A CN107038858A (en) 2017-08-11
CN107038858B true CN107038858B (en) 2019-05-28

Family

ID=59539366

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710380104.3A CN107038858B (en) 2017-05-25 2017-05-25 Commute private car dynamic share-car recommended method

Country Status (1)

Country Link
CN (1) CN107038858B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108320503A (en) * 2018-01-19 2018-07-24 江苏本能科技有限公司 Vehicle traveling querying method and system based on point identification

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104751625A (en) * 2013-12-25 2015-07-01 上海博泰悦臻网络技术服务有限公司 Carpooling method and carpooling system based on trajectory analysis
CN105674995A (en) * 2015-12-31 2016-06-15 百度在线网络技术(北京)有限公司 Method for acquiring commuting route based on user's travel locus, and apparatus thereof
CN105701560A (en) * 2015-12-31 2016-06-22 百度在线网络技术(北京)有限公司 Method and device for determining commuting route information
CN106027637A (en) * 2016-05-18 2016-10-12 福建工程学院 Car-pooling method and system based on trajectory information
KR20160150212A (en) * 2015-06-19 2016-12-29 (주)연결해 Techniques for matching the movement path of the carpool driver and rider

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9612127B2 (en) * 2014-07-25 2017-04-04 GM Global Technology Operations LLC Carpool finder assistance
US9904900B2 (en) * 2015-06-11 2018-02-27 Bao Tran Systems and methods for on-demand transportation

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104751625A (en) * 2013-12-25 2015-07-01 上海博泰悦臻网络技术服务有限公司 Carpooling method and carpooling system based on trajectory analysis
KR20160150212A (en) * 2015-06-19 2016-12-29 (주)연결해 Techniques for matching the movement path of the carpool driver and rider
CN105674995A (en) * 2015-12-31 2016-06-15 百度在线网络技术(北京)有限公司 Method for acquiring commuting route based on user's travel locus, and apparatus thereof
CN105701560A (en) * 2015-12-31 2016-06-22 百度在线网络技术(北京)有限公司 Method and device for determining commuting route information
CN106027637A (en) * 2016-05-18 2016-10-12 福建工程学院 Car-pooling method and system based on trajectory information

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于时间序列聚类方法分析北京出租车出行量的时空特征;程静 等;《地球信息科学学报》;20160930;第18卷(第9期);第1227-1239页

Also Published As

Publication number Publication date
CN107038858A (en) 2017-08-11

Similar Documents

Publication Publication Date Title
Alexander et al. Origin–destination trips by purpose and time of day inferred from mobile phone data
Çolak et al. Analyzing cell phone location data for urban travel: current methods, limitations, and opportunities
Wang et al. When will you arrive? estimating travel time based on deep neural networks
Kumar et al. Short-term traffic flow prediction using seasonal ARIMA model with limited input data
EP2779138B1 (en) Method for reading vehicle license plates
Balan et al. Real-time trip information service for a large taxi fleet
Chen et al. Nlpmm: A next location predictor with markov modeling
CN102460534B (en) Computer implementation method of predicting expected road traffic conditions based on historical and current data and computing system
US20150294422A1 (en) Assessing asynchronous authenticated data sources for use in driver risk management
CN102289935B (en) Use the data estimation road traffic condition from Mobile data source
Paefgen et al. Multivariate exposure modeling of accident risk: Insights from Pay-as-you-drive insurance data
CN103325245B (en) Method for predicting space-time traveling track of blacklisted vehicle
CN101710448B (en) Road traffic state detecting device based on omnibearing computer vision
US20140114556A1 (en) Traffic prediction using real-world transportation data
US20130325629A1 (en) Interactive Advertising Based On License Plate Recognition
Malleson et al. Exploring the impact of ambient population measures on London crime hotspots
Liao et al. Deep sequence learning with auxiliary information for traffic prediction
Polzin The case for moderate growth in vehicle miles of travel: a critical juncture in US travel behavior trends
JP2015125775A (en) System and method for multi-task learning system for prediction of demand on system
Lee et al. Discovering traffic bottlenecks in an urban network by spatiotemporal data mining on location-based services
US8090523B2 (en) Travel-time prediction apparatus, travel-time prediction method, traffic information providing system and program
Liu et al. Urban traffic prediction from mobility data using deep learning
US20060178811A1 (en) Method and apparatus for end-to-end travel time estimation using dynamic traffic data
Wang et al. Citywide traffic congestion estimation with social media
DE102012201472A1 (en) Method for providing parking information on free parking spaces

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