CN106022541A - Arrival time prediction method - Google Patents
Arrival time prediction method Download PDFInfo
- Publication number
- CN106022541A CN106022541A CN201610472352.6A CN201610472352A CN106022541A CN 106022541 A CN106022541 A CN 106022541A CN 201610472352 A CN201610472352 A CN 201610472352A CN 106022541 A CN106022541 A CN 106022541A
- Authority
- CN
- China
- Prior art keywords
- hourage
- cfs
- station
- connectionist model
- model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000013528 artificial neural network Methods 0.000 claims abstract description 46
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims description 44
- 238000012360 testing method Methods 0.000 claims description 30
- 238000001914 filtration Methods 0.000 claims description 15
- 230000001537 neural effect Effects 0.000 claims description 7
- 241001269238 Data Species 0.000 claims 1
- 230000009471 action Effects 0.000 abstract description 8
- 238000013530 stochastic neural network Methods 0.000 abstract 1
- 210000002569 neuron Anatomy 0.000 description 21
- 238000004364 calculation method Methods 0.000 description 15
- 230000001413 cellular effect Effects 0.000 description 12
- 230000008569 process Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 5
- 210000004218 nerve net Anatomy 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008447 perception Effects 0.000 description 3
- 230000032258 transport Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000013277 forecasting method Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 208000011580 syndromic disease Diseases 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/123—Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/123—Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams
- G08G1/127—Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams to a central station ; Indicators in a central station
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Theoretical Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
- Navigation (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
A system and a method for predicting arrival time based on a random neural network group are disclosed, wherein a vehicle-mounted action device periodically acquires positioning information (longitude and latitude coordinates), compares the positioning information with site information (a polygonal or circular area) in the vehicle-mounted action device to judge whether the vehicle-mounted action device arrives at the site (or leaves the site), and reports the arrival (or leaving) information and a time point to a cloud server. The cloud server is responsible for collecting arrival (or departure) information returned by the vehicle-mounted mobile device, analyzing travel time among all the stations, storing the data into the cloud database server, and using a travel time data set to train parameter values of the stochastic neural network group algorithm.
Description
The present patent application be filing date on March 3rd, 2015, Application No. 201510094269.5, invention entitled " one
Kind of arrival time prognoses system and method " the divisional application of patent application.
Technical field
The present invention relates to wireless communication technology field, particularly to a kind of arrival time Forecasting Methodology.
Background technology
At present, the prior art of public transport arrival time prediction is: use historical data to carry out statistics peace
All, obtain the average speed between each CFS to CFS and hourage, or the automobile's instant velocity applying vehicle instantly real-time is believed
Breath, estimates arrival time with this.But, these methods cannot react road condition change situation real-time between website, thus results in
Bigger arrival time information error.
The Taiwan Patent of Publication No. TW201137803, main proposition collects the information of arriving at a station that bus is returned in the past
Estimate the average speed between CFS to CFS and hourage, it is possible to add up according to different weeks and period, when
History average speed and hourage can be obtained during user inquiry.Estimate when arriving at a station although the method can provide rapidly
Between, but mainly use historical data meansigma methods, thus the prediction of arrival time cannot be carried out according to real-time road, therefore have
Bigger error may be caused in arrival time prediction.
The Taiwan Patent of Publication No. TW201344647, after prediction arrival time, according to the position letter that bus is real-time
Breath, provides speed to adjust suggestion to driver, improves, with this, the punctuality rate that arrives at a station.Although the method can estimate arrival time,
The punctuality control that arrives at a station can also be provided, but the method mainly uses the meansigma methods of historical data, and cannot be according in real time
Road conditions carry out the prediction of arrival time, therefore are likely to cause error bigger in arrival time prediction.
The Taiwan Patent of Taiwan Patent Publication No. TW201405497, main proposition works as vehicle-mounted action equipment through each
During section, by vehicle-mounted action equipment by each section hourage real-time repaying to the Surveillance center of rear end, then by monitoring
The shortest hourage and the longest hourage in each section are distributed to all vehicle-mounted action equipments by center.If vehicle-mounted action sets
Standby hourage, between the shortest hourage and the longest hourage, is the most no longer returned.Although the method can be effective
Grasp the hourage in each section and reduce transmission quantity, but not suggesting that the Forecasting Methodology of bus arrival time, therefore nothing
Method prediction bus arrival information.
The Taiwan Patent of Publication No. TW201117146, the main method that inquiring bus hourage is provided, can allow and make
User inquires real time position and the hourage of its bus to be taken.Although the method can allow user inquire public affairs
Hand over the real-time position of car and hourage, but do not suggest that the Forecasting Methodology of bus arrival time, therefore unpredictable bus
Arrive at a station information.
The Taiwan Patent of Publication No. TW200828190, main proposition utilizes the action equipment of user to receive and arrives at a station
Information, when user arrives at website, can give notice and remind user.Although the method can be reminded when arriving website
User, it is provided that real-time information of arriving at a station, but information of forecasting but cannot be provided.
Notification number is the Taiwan Patent of TWI252441, mainly proposes by bus reception satellite positioning signal, and in real time
Positional information is back to Surveillance center, then when being arrived at a station according to bus real time position by the prediction module of Surveillance center
Between predict.Although the method can be provided to time prediction of standing, but refers only in patent with reference to empirical value, and future specifically mentions
The Forecasting Methodology of bus arrival time.
Notification number is the Taiwan Patent of TWI341998, and the main real-time speed proposed according to bus and bus are to each
The distance of individual website, predicts hourage；And the distance according to the walking speed of user and user to each website,
Calculate the walking time.Finally further according to estimating hourage and walking time the website that is suitable for.Although the method can carry
For the Forecasting Methodology of bus hourage, but the method primary concern is that bus real-time speed instantly and to stop spacing
From, but the transport information between vehicle and website is not considered, therefore it is likely to cause error bigger in arrival time prediction.
The Taiwan Patent of Publication No. TW201232489, proposes to use the empirical modal of Hilbert-Huang conversion (HHT)
Decomposition method combines ash pattern and predicts road speed, is scaled hourage and arrival time further according to the speed estimated.Although
The method can effectively use mathematics and statistical model to carry out speed prediction, but be because the method and use all of data to carry out
Analyze, therefore the impact of extremum cannot be avoided, it would be possible to cause error bigger in arrival time prediction.
Summary of the invention
In view of above-mentioned problem of the prior art, it is an object of the invention to providing a kind of arrival time Forecasting Methodology, pass through
Collect the hourage between each section and CFS to CFS of period, and the random neural network group proposing novelty analyzes
Data acquisition system hourage stated, sets up multiple Connectionist model to avoid the impact of extremum, and considers many
Predicting the outcome of individual Connectionist model promotes the accuracy of prediction, predicts, with this, bus that user to be taken
Arrival time, will predict the outcome and be supplied to user as reference.
The arrival time prognoses system of the present invention includes multiple station stop board, multiple vehicle-mounted terminal equipment, multiple cytoreticulum
Network base station, high in the clouds calculation server, high in the clouds historical data base and multiple arrival time prognoses system client device.Its
In, each station stop board has a latitude and longitude coordinates information.When each described vehicle-mounted terminal equipment is close to station, the plurality of station
During board, each described vehicle-mounted terminal equipment senses the plurality of latitude and longitude coordinates information, and then produces information of arriving at a station.Arrive at a station letter
Breath is transmitted by the plurality of cellular network base station, and high in the clouds calculation server receives from cellular network base station
After information of arriving at a station, calculate hourage, remain hourage further according to hourage and inquiry station point prediction, and be converted to
Arrival time, then arrival time is transmitted by cellular network base station.High in the clouds historical data base stores longitude and latitude and sits
Hourage between mark information and station stop board.Arrival time prognoses system client device sends inquiry website, and connects
Receive the arrival time transmitted by cellular network base station, then show arrival time.
The arrival time Forecasting Methodology of the present invention comprises the following steps: to set random neural network group's algorithm parameter value；
Read the hourage between the CFS to CFS in historical data base；Randomly generate m Connectionist model；Filter out accuracy
After the Connectionist model of threshold value, remain k Connectionist model；Obtain the travelling between real-time CFS to CFS
Test data in time or test phase；Whilst on tour or test data are input to k Connectionist model after filtering
In, and predict the hourage of CFS to CFS；And after obtaining the hourage of CFS to CFS of prediction, it is scaled targeted sites
The time of advent.
In sum, the arrival time prognoses system of the present invention and method, have following in one or more advantages:
1. the hourage that the present invention collects between each real-time section and CFS to CFS of period estimates current vehicle
Position arrives the hourage of targeted sites.
2. the present invention proposes the random neural network group of novelty to analyze data acquisition system above-mentioned hourage, sets up multiple
Connectionist model, then consider predicting the outcome of multiple Connectionist model and promote prediction accuracy, come with this
The arrival time of the bus that prediction user to be taken, will predict the outcome and be supplied to user as reference.
3. the present invention is in the study stage of random neural network group's algorithm, for each Connectionist model respectively from
In data acquisition system, the many pen data of random taking-up are as training data, and using remaining data as the test number in the training stage
According to, then be input to training data in each Connectionist model learn, thus the impact of extremum can be avoided.
4. the present invention is at the test phase of random neural network group's algorithm with in the implementation stage, uses each class nerve net
The weight that the hourage of network model prediction and training stage are learnt to obtain is weighted averagely, finally by after weighted average
Hourage as the predicting travel time value of this random neural network group's algorithm, and will be scaled hourage when arriving at a station
Between, with this time prediction that carries out arriving at a station.
Accompanying drawing explanation
Fig. 1 is the structural representation of the embodiment of the present invention one arrival time prognoses system；
Fig. 2 is the schematic flow sheet of the embodiment of the present invention two arrival time Forecasting Methodology；
Fig. 3 is the schematic flow sheet of the embodiment of the present invention three arrival time Forecasting Methodology；
Fig. 4 is the schematic diagram of the embodiment of the present invention four Connectionist model；
Fig. 5 is the schematic diagram that the embodiment of the present invention five predicts hourage.
Detailed description of the invention
With reference to Fig. 1, the present invention is the system about the prediction of a kind of arrival time based on random neural network group.This is
System mainly can predict the arrival time of vehicle, it is adaptable to passenger traffic dealer, logistics dealer or other have arrival time forecast demand
Relevant dealer, and the arrival time of prediction is supplied to client device, allows client or user can grasp vehicle in real time
Information and arrive at a station information, saves waiting time, mainly comprises following six modules: (1) multiple station stop board 100: this station
Board equipment consists predominantly of one group of latitude and longitude coordinates information, and this information can have previously been stored in vehicle-mounted terminal equipment and high in the clouds fortune
Calculating in server, when vehicle-mounted terminal equipment is close to station stop board, vehicle-mounted terminal equipment can arrive at a station information with perception.Additionally, this
Stop board equipment can also embed RFID (Radio Frequency IDentification, radio frequency identification) label, works as car
Can be with perception station board when closing on, it is possible to this judges to arrive at a station.(2) multiple vehicle-mounted terminal equipments 101: this equipment mainly comprises
There are GPS (Global Positioning System, global positioning system) module, cellular network module and DBM
(drawing the most in FIG), can collect current vehicle position (comprising latitude and longitude coordinates), and judge whether current position faces
Nearly station stop board 100, if in the scope AT STATION near station board 100, judging to arrive at a station, and will arrive at a station information and time point passes through
Cellular network base station 102 is back to high in the clouds calculation server end 103.Additionally, at the part judged of arriving at a station, vehicle-mounted terminal equipment
101 can also embed RFID reader, can receive RFID volume from stop board equipment with perception station board when vehicle closes on
Mark signal, judges whether to arrive at a station.(3) multiple cellular network base stations 102: each cellular network base station 102 provides data
Transmitting function and the receive capabilities of data, be responsible for vehicle-mounted terminal equipment 101, high in the clouds calculation server 103 and arrival time
Data transmission between prognoses system client device 106.(4) high in the clouds calculation server 103: this server mainly can be collected
With the information of arriving at a station analyzed from vehicle-mounted terminal equipment 101, arrival time point, calculate each station according to each arrival time point
Trip between multiple CFSs to CFS before on hourage between arriving at a station, then the travel route of the targeted sites that user is inquired about
The data acquisition system of row time, is input to random neural network group's arrival time Forecasting Methodology proposed by the invention and has been trained
The neural network group become, is analyzed obtaining the residue predicting travel time arriving targeted sites with this with computing, then changes
Calculate the arrival time for arriving targeted sites.(5) high in the clouds historical data base 105: this data base mainly can store the every of history
Hourage between individual CFS to CFS, can be used to the training data set as random neural network group, be used for training often
Individual Connectionist model.(6) multiple arrival time prognoses system client devices 106: this equipment can be a mobile
Equipment, has human-computer interaction interface and network transmission module, and user can be allowed to pass through this equipment query and show what it to be obtained
The arrival time prediction of targeted sites.And can be by user pre-set its website to be taken and time, more thus equipment
Actively update and judge, actively sending prompting message and sound to user when vehicle will arrive.
Referring to figs. 2 and 3, the present invention more provides the side that a kind of arrival time based on random neural network group is predicted
Method.The method mainly will include 2 stages: (a) training stage and (b) carry out and test phase.Wherein, the training stage mainly wraps
Include 4 steps, be respectively as follows:
Step S201: set random neural network group's algorithm parameter value；
Step S202: read the hourage between each CFS to CFS in historical data base；
Step S203: randomly generate m Connectionist model；
Step S208: after filtering out the Connectionist model that accuracy is less than threshold value, remain k neural network mould
Type.
Carry out and test phase mainly includes 3 steps, be respectively as follows:
Step S301: obtain the hourage between real-time each CFS to CFS or the test data in test phase；
Step S302: enter data in k Connectionist model after filtration, and predict the trip between CFS to CFS
The row time；
Step S306: after obtaining CFS to CFS hourage of prediction, be scaled the time of advent of targeted sites.
In step s 201, first random neural network group's algorithm is set by arrival time prognoses system developer
Related parameter values, related parameter values includes Connectionist model quantity (follow-up will illustrate as a example by m), class nerve net
Each hidden layer in hidden layer maximum quantity in network model (follow-up will illustrate as a example by hmax), Connectionist model
Maximum neuronal quantity (follow-up will illustrate as a example by cmax), the training data number of training Connectionist model account for always
The ratio of training stage data number (follow-up will illustrate as a example by r%) and accuracy threshold value (follow-up will be with
Illustrate as a example by wthreshold).
In step S202, from high in the clouds historical data base 103, obtain vehicle arrive the time of each website, and convert
For the hourage between CFS to CFS, such as: the arrival time at station 1 is time point t1, and when the arrival time at station 2 is
Between put t2, then station 1 is | t2-t1 | to the hourage at station 2.Again this hourage is gathered as Connectionist model
Input and output data carry out follow-up study.As a example by Fig. 1, it is intended to when vehicle drives to station n-2 predict the n that gets to the station
Time (i.e. target output hourage be | tn-tn-2 |), input hourage data acquisition system can include | t2-t1 |,
|t3-t2|,...,|tn-2-tn-3|}。
In step S202, the random neural network group's algorithm ginseng set according to arrival time prognoses system developer
Numerical value, randomly generates m Connectionist model, and each Connectionist model is by the most random total instruction obtained
The data of the r% practicing data number use as training and study, and remaining data (i.e. the data volume of 100%-r%) are made
Checking for each Connectionist model uses, and the different data obtained are trained by each Connectionist model
And checking.Additionally, each Connectionist model will be according to pre-set parameter, generation 0～hmax hidden layer, Yi Jiwei
Each hidden layer generation 0～cmax neuron, the hidden layer of the most each Connectionist model and the combination of neuron
All will be different.It is input in Connectionist model be trained and learn by the data of aforesaid r% again, reaches convergence
After, then the data (i.e. test data in the training stage) of 100%-r% are input to the Connectionist model after training, and
Obtain the hourage of prediction, and compare with correct hourage, obtain each Connectionist model with this
Accuracy, and using this accuracy as the weighted value carried out and during test phase.
In step S208, after filtering out the Connectionist model that accuracy is less than threshold value, remain k class nerve net
Network model: the accuracy of m the Connectionist model randomly generated is compared with accuracy threshold value wthreshold,
After will be less than Connectionist model (i.e. accuracy the is the lowest) eliminating of this threshold value, remaining k Connectionist model；If
There is no the accuracy of any Connectionist model higher than threshold value, step S201 will be returned, arrival time prognoses system open
The personnel of sending out reset threshold value, and re-training random neural network group.
In step S301, carrying out and in test phase, first by between the CFS to CFS of the real-time vehicle first obtained
Hourage, such as: as station n-2 during vehicle moves to Fig. 1, user wants to inquire about the arrival time prediction of station n
(i.e. the hourage of target output is | tn-tn-2 |).Now, vehicle number hourage in this time distance will can be calculated
According to set | t2-t1 |, | t3-t2 | ..., | tn-2-tn-3 |, as the input data of Connectionist model.
In step s 302, real-time data acquisition system hourage of acquirement | t2-t1 |, | t3-t2 | ..., | tn-2-
Tn-3 | being input to k Connectionist model after filtering after }, each Connectionist model will dope one |
Tn-tn-2 | prediction hourage, then it is multiplied by respectively by the power of each Connectionist model acquired by the training stage
Weight values (i.e. the accuracy of each Connectionist model during the training stage), and by the summation of the value after weighting divided by weighted value
Summation (is i.e. weighted average).
In step S301, the prediction hourage obtained by acquirement considers k Connectionist model | tn-
Tn-2 | after, then time point tn-2 real-time for vehicle must be got to the station when arriving at a station of n plus prediction | tn-tn-2 | hourage
Between predict, and this predicted the outcome be supplied to user.
The present invention collect and analyze from vehicle-mounted terminal equipment 101 passback to (from) information of standing (comprise site information with
Time point etc.), after this data acquisition system is converted to the hourage between CFS to CFS, it is stored in high in the clouds historical data base 105,
And design and implementation one information prediction side of arriving at a station based on random neural network group's algorithm in calculation server 103 beyond the clouds
Method module, can access set hourage in high in the clouds historical data base 105, and be entered into based on random neural network
Group algorithm arrive at a station in information forecasting method module, carry out Connectionist model training predict hourage with this.When arriving at a station
When information prediction system client carries out the prediction of website arrival time, current vehicle-mounted terminal equipment 101 can be returned at this route
Front multiple site information to be input in the neural network group trained to carry out arrive hourage of targeted sites pre-
Surveying, reconvert and offer arrive the time of advent of targeted sites to the time prediction system client device 106 that arrives at a station.The present invention's
Technical characterstic essentially consists in proposition and one random neural network group's algorithm of design, and is applied to the information forecasting method that arrives at a station
In, below will illustrate by way of example.
The present invention provides the system that a kind of arrival time based on random neural network group is predicted, its system architecture such as figure
Shown in 1.This system includes multiple station stop board 100, multiple vehicle-mounted terminal equipment 101, multiple cellular network base station 102,
103, high in the clouds historical data base 105 of individual high in the clouds calculation server and multiple arrival time prognoses system client device
106.In the present embodiment as a example by the station stop board 100 of same route, this route there are n station, each station have tool
There is positional information (comprising longitude and latitude).As shown in Table 1, (i.e. n in Fig. 1 is altogether to include 12 stations at route 1
12), the longitude and latitude of its correspondence can be stored in vehicle-mounted terminal equipment；When car number 1 is travelled from station 1 toward station 2,
During 2014/4/1 14:53 the GPS module of vehicle-mounted terminal equipment detect vehicle place longitude be 120.97839, latitude be
24.808658, assessment vehicle closes on station 2 (such as: in air line distance 30 meters), then be judged as arriving at a station, and the letter that this arrived at a station
Breath (comprising Station XXX and time point) passes back to high in the clouds calculation server 103 by cellular network base station 102.
Additionally, station stop board 100 can also possess RFID label tag, and vehicle-mounted terminal equipment 101 can possess RFID reader,
May detect that the RFID label tag of this station stop board 100 when vehicle-mounted terminal equipment 101 closes on station stop board 100, and judge with this
For arriving at a station, then information of this being arrived at a station (comprising Station XXX and time point) passes back to high in the clouds by cellular network base station 102 and transports
Calculate server 103.Vehicle arrive at a station information report data acquisition system as shown in Table 2, mainly can note down route number, car number,
Station XXX and time point etc., and the vehicle information of arriving at a station can be converted to CFS to CFS hourage by high in the clouds calculation server 103
Information (as shown in Table 3), and information is stored in high in the clouds historical data base 105.Such as, car number 1 is dispatched a car by station 1
Time time be 2014/4/1 14:46:28, and arrive at station 2 at 2014/4/1 14:53:31, therefore station 1 is to station 2
Hourage is 423 seconds；And the time that car number 2 is when being dispatched a car by station 1 is 2014/4/1 19:32:22, and 2014/4/
1 19:40:13 arrives at station 2, and therefore station 1 is 471 seconds to the hourage at station 2.
When the vehicle of numbering 10001 drives to station 6 (between cloud server 103 its station 1 known～station 6
Hourage between CFS to CFS), and have an arrival time prognoses system client device to look into high in the clouds calculation server 103
The arrival time asking route number 1 station 12 (is i.e. predicted that the hourage at station 12 is arrived at station 6, and is converted to arriving of station 12
Reach the time).Now, data (the i.e. distance numbering 1 and 2 during high in the clouds calculation server 103 can use high in the clouds historical data base 105
CFS to CFS travel time information, as shown in Table 4) as random neural network group's algorithm in the data of training stage, build
Vertical random neural network group, and use this algorithm to carry out arriving at a station time prediction.
Table one stop location information
Table two vehicle arrives at a station information
Table three CFS to CFS travel time information
The training stage data of the random neural network group's algorithm of table four
The method of arrival time based on the random neural network group prediction of the present invention, its method flow such as Fig. 2 and Fig. 3
Shown in.The method mainly comprises 2 stages: (a) training stage and (b) carry out and test phase.
Training stage mainly comprises 4 steps, respectively steps S201: set random neural network group's algorithm parameter
Value；S202: read the hourage between each CFS to CFS in historical data base；S203: randomly generate m neural network
Model；And S208: after filtering out the Connectionist model that accuracy is less than threshold value, remain k Connectionist model.
Carry out and test phase mainly comprises 3 steps, respectively S301: obtain the trip between real-time each CFS to CFS
Test data in row time or test phase；S302: enter data in k Connectionist model after filtration, and
Hourage between prediction CFS to CFS；And S306: after obtaining CFS to CFS hourage of prediction, it is scaled targeted sites
The time of advent.
In the training stage, first will be set random neural network group's algorithm by arrival time prognoses system developer
Related parameter values (step S201).Such as, set and have 10 Connectionist models (i.e. m as 10), neural network mould
During in type, hidden layer maximum quantity is 5 (i.e. hmax is 5), Connectionist model, each hidden layer maximum neuronal quantity is 7
It is 60% (i.e. that (i.e. cmax is 7), the training data number of training Connectionist model account for the ratio of total training stage data number
R% is 60%) and accuracy threshold value be 0.945 (i.e. wthreshold is 0.945=94.5%), follow-up will be according to this
Parameter value produces 10 Connectionist models to be carried out arriving at a station time prediction.
In this S202 step, the vehicle obtaining history to high in the clouds historical data base is arrived the time of each website,
And the hourage being scaled between CFS to CFS, as shown in Table 4.Due in the present embodiment, vehicle to be predicted travels to car
Stand 6, and the time of advent at station to be predicted 12, and arrival time data acquisition system between known station 1～station 6 t1, t2,
T3, t4, t5, t6}, data acquisition system hourage being converted between CFS to CFS | t2-t1 |, | t3-t2 |, | t4-t3 |, | t5-
T4 |, | t6-t5 |, and in order to predict station 6 to station 12 hourage (i.e. the hourage of target output is | t12-t6
|).In the present embodiment by data acquisition system hourage { | t2-t1 |, | t3-t2 |, | t4-t3 |, | t5-t4 |, | t6-t5 | } point
The most named parameter name { x1, x2, x3, x4, x5}, and named parameter name y of | t12-t6 | hourage of target output.
Step S203 randomly generates in m Connectionist model, further includes step S204: produce training data and checking
Data.Specifically, random neural network group's algorithm parameter that the present invention sets according to arrival time prognoses system developer
Value, randomly generates 10 Connectionist models, and sets in Connectionist model hidden layer maximum quantity as 5, class nerve
In network model, each hidden layer maximum neuronal quantity is 7, and the hidden layer quantity of the most each Connectionist model will be between 0
～5 layers, the neuronal quantity of each hidden layer between 0～7, will produce the embodiment (step as shown in Table 5 of result
S205).The hidden layer of Connectionist model 1 is 1 layer, and the neuron number of this layer of hidden layer is 2 (as shown in Figure 4)；Class god
Be 2 layers through the hidden layer of network model 2, the neuron number of the 1st layer of hidden layer be the neuron number of 3, the 2nd layer hidden layer be 4
Individual；10 Connectionist models can be obtained by that analogy.Further, instruction is accounted for due to the training data number of training Connectionist model
Practicing the 60% of the total stroke count of phase data, as a example by table four, total stroke count of training stage data number is 10000, so each class
Neural network model will take out 6000 at random and use as training Connectionist model study, and remaining 4000
TDTRS (Testing Data in TRaining Stage, the test data in the training stage) will be respectively as the training stage
Time each Connectionist model checking use.In this step, 6000 pen data acquired by each Connectionist model
Set the most each randomly generate, each Connectionist model by obtain different data acquisition systems be trained and learn
Practise.
The random neural network group of table five
Step S206: Connectionist model training and study.In the present embodiment, 10 Connectionist models will divide
Do not input 6000 pen data be trained and learn, say following with as a example by Connectionist model 1 (as shown in Figure 4)
Bright, wherein 6000 pen data at Connectionist model 1 are one to comprise distance numbering 1 and do not comprise distance numbering 10000
Data combine, and with after the training of Connectionist model 1 and study explanation such as.
Step i: randomly generate the weight of each neuron, and the constant term of hidden layer and output layer neuron, such as table
Shown in six.
The weight of each neuron of table six Connectionist model 1, and the constant of hidden layer and output layer neuron
?
| w1,6 | w2,6 | w3,6 | w4,6 | w5,6 | w1,7 | w2,7 | w3,7 | w4,7 | w5,7 | w6,8 | w7,8 | 6 | 7 | 8 |
| 0.7 | 0.7 | 0.2 | 0.1 | 0.6 | 0.1 | 0.8 | 0.5 | 0.3 | 1.0 | 0.6 | 0.6 | 0.8 | 0.7 | 0.3 |
Step ii: 6000 pen data are inputted one by one to Connectionist model 1, below as a example by distance numbering 1.First
Being first the numerical value between 0～1 by data normalization, therefore the data in embodiment are smaller than 5000, therefore with divided by 5000
Carrying out normalization, result is as shown in Table 7.Further according to input signal, calculate the output signal of each hidden layer neuron, Qi Zhongben
Embodiment uses Logistic to distribute (i.e.) mode calculate output signal, calculation is as follows.
Distance numbering 1 numerical value after table seven normalization
Neuron 6:
Total input signal:
Converted output signal:
Neuron 7: total input signal:
Converted output signal:
Step iii: according to hidden layer output signal, calculate the output signal of output layer neuron.
Neuron 8:
Total input signal:
Converted output signal:
Step iv: compare the error term of output valve (i.e. 0.759554) and true value (i.e. 0.7796).
Neuron 8 error term:
Step v: error term is fed back to hidden layer, calculates the error term of hidden layer neuron respectively.
Neuron 6 error term:
Neuron 7 error term:
Step vi: according to neuron error item, updates each neuron weight and constant term, sets in the present embodiment
Practising speed σ is 0.8.
Step vii: repeat step ii～step vi, learns the input of each pen data to Connectionist model
Practise, until the difference of the output signal of the output signal of this bout and upper bout is less than threshold value othreshold (in this example
Middle othreshold is set to 0.01), then reach convergence and complete study, determining each neuron power of this Connectionist model
Weight and constant term.
The above-mentioned training for Connectionist model 1 and learning process, train other neural network mould the most simultaneously
Type (i.e. Connectionist model 2～Connectionist model 10), can support parallel calculation.After completing training, follow-up in prediction
Repeatable step ii～step iii when the hourage between station 12 is arrived at station 6, will test data or real time data as input
Signal, and output signal is whilst on tour predictive value.Wherein, by the predicting travel time value of Connectionist model output, need again
Carrying out normalized reduction, side can obtain number of seconds hourage, such as: output signal is 0.759554, need to be multiplied by 5000, take
Must it be 3797.769233 seconds hourage.
Step S207: Connectionist model checking and weight.When the training and that complete all Connectionist models
After habit, remaining 4000 pen data can be used to carry out the checking of each Connectionist model, and calculate average accuracy
Weight as each Connectionist model.As a example by Connectionist model 1, by whole for the test data in the training stage
It is input in the Connectionist model 1 after training repeat step ii～step iii, accuracy can be calculated.Such as, distance numbering
10000 when being input signal, and after its normalization, numerical value is as shown in Table 8, and obtaining predictive value is 0.75986369, then by predictive value
Being multiplied by 5000 is 3799.318449, can obtain accuracy for for 1-(| true value-predictive value |/true value)=1-(| 3939-
3799.318449 |/3939)=96.45%；By that analogy, the test data (TDTRS) in 4000 training stages can be calculated
Average accuracy, be 93.23% in this example.In the present embodiment, average correct corresponding to 10 Connectionist models
Rate is respectively 93.23%, 94.90%, 94.03%, 93.57%, 94.61%, 93.52%, 94.93%, 95.21%,
94.48%, 94.45%, as shown in Table 9.
Distance numbering 10000 numerical value after table eight normalization
The average accuracy of each Connectionist model of table nine
Step S208: after filtering out the Connectionist model that accuracy is less than threshold value, remain k neural network mould
Type.This step will analyze the average accuracy of each Connectionist model, and will be less than accuracy threshold value wthreshold
(i.e. 94.5% set by the present embodiment) filters out, wherein Connectionist model 1, Connectionist model 3, class nerve net
Network model 4, Connectionist model 6, Connectionist model 9, Connectionist model 10 etc. 6 will be filtered, and remaining 4
Individual Connectionist model and weighted value thereof are for carrying out and test phase.
Connectionist model after table ten, filtration and weighted value thereof
In step S301, carrying out and during test phase, the vehicle information of arriving at a station during treating excess syndrome is input to have trained
Random neural network group, carry out arriving at a station time prediction.Such as, arrival time prognoses system client device is 2014/5/3
It is intended to during 11:59:00 inquire about the time of advent arriving at station 12, will pick up the car between station 1～the arrival time at station 6 and CFS to CFS
Hourage (as shown in table 11), as the input data (as shown in table 12) of random neural network group, obtain target
Predictive value station 6 is to the hourage at station 12.
Table 11 vehicle arrives at a station information
Table 12 CFS to CFS travel time information
Additionally, arrival time prognoses system developer can also collect historical data as in test phase in this stage
Test data (TDTES), obtain each distance numbering each CFS to CFS between hourage as random class nerve net
Network group's input value, to analyze and optimization random neural network group.
In step S302, enter data into k Connectionist model after filtration, and predict the trip between CFS to CFS
The row time.As it is shown in figure 5, obtain input data after can using data as each filtration after Connectionist model
(i.e. Connectionist model 2, Connectionist model 5, Connectionist model 7, Connectionist model 8, such as table ten institute
Show) input signal, and neural by Connectionist model 2, Connectionist model 5, Connectionist model 7, class respectively
Network model 8 predict hourage be 3766.607 seconds, 3857.98 seconds, 3661.828 seconds, 3724.095 seconds (step S303),
As shown in table 13.Finally, then it is weighted averagely (step S304～S305) according to the weight of each Connectionist model to obtain
To 3752.516552 seconds (i.e. [94.90%*3766.607+94.61%*3857.98+94.93%* of predicting travel time value
3661.828+95.21%*3724.095]/[94.90%+94.61%+94.93%+95.21%]=3752.516552).
Connectionist model after table 13 filtration and weighted value thereof
In step S306, after obtaining CFS to CFS hourage of prediction, it is scaled the time of advent of targeted sites.Obtaining
After CFS to CFS predicting travel time value, can be according to current information of arriving at a station, and combining station predicting travel time value of arriving at a station is converted to
Arrive the time of advent of targeted sites.The distance numbering 10001 of the present embodiment get to the station 6 time point be 20,14/,5/3 11:
58:46, and station 6 is 3752.516552 seconds to the predicting travel time value at station 12, therefore station 12 is predicted arrival time and is
2014/5/3 13:01:19, then give, by this information back, the time prediction system client device that arrives at a station.
Practice, from the point of view of the example of passenger traffic dealer, carries out excess syndrome with the data of passenger traffic dealer A, altogether collects 2014
The data in whole month in March, comprises 2956 times the most altogether, contains 40 road sections altogether, and be respectively adopted in experimental situation
Different data prospect algorithm to test its accuracy, include Luo Jisi and return (Logistic Regression, LR), pass
The back propagation neural network (Back-Propagation Neural Network, BPNN) and proposed by the invention of system
Random neural network group (Random Neural Networks, RNN), it was demonstrated that the method is the most superior, experimental result
Shown in table 14.
Table 14 present invention is with other data prospecting method efficiency ratio relatively
| Method | Accuracy |
| Historical summary mean value method | 73.79% |
| Luo Jisi returns | 77.43% |
| Back propagation neural network | 77.88% |
| The present invention | 78.22% |
In sum, the arrival time prognoses system based on random neural network group of the present invention and method, through receiving
Collect the hourage between each section and CFS to CFS of period, and the random neural network group proposing novelty is above-mentioned to analyze
Data acquisition system hourage, set up multiple Connectionist model to avoid the impact of extremum, and consider multiple
Predicting the outcome of Connectionist model promotes prediction accuracy, predicts arriving at a station of bus that user to be taken with this
Time, it is provided that to user as reference.
The foregoing is only illustrative, rather than be restricted person.Any spirit and scope without departing from the present invention, and to it
The equivalent modifications carried out or change, in claim attached after being intended to be limited solely by.
[symbol description]
100: station stop board
101: vehicle-mounted terminal equipment
102: cellular network base station
103: high in the clouds calculation server
104: high in the clouds computing machine room
105: high in the clouds historical data base
106: arrival time prognoses system client device
S201～207, S301～S306: step
1～8: Connectionist model
Based on same inventive concept, the embodiment of the present invention additionally provides a kind of mobile terminal, due to the mobile terminal of Fig. 3
Corresponding method is the method that a kind of mobile terminal of the embodiment of the present invention starts, and therefore the enforcement of embodiment of the present invention method is permissible
See the enforcement of system, repeat no more in place of repetition.
Those skilled in the art are it should be appreciated that embodiments of the invention can be provided as method, system or computer program
Product.Therefore, the reality in terms of the present invention can use complete hardware embodiment, complete software implementation or combine software and hardware
Execute the form of example.And, the present invention can use at one or more computers wherein including computer usable program code
The upper computer program product implemented of usable storage medium (including but not limited to disk memory, CD-ROM, optical memory etc.)
The form of product.
The present invention is with reference to method, equipment (system) and the flow process of computer program according to embodiments of the present invention
Figure and/or block diagram describe.It should be understood that can the most first-class by computer program instructions flowchart and/or block diagram
Flow process in journey and/or square frame and flow chart and/or block diagram and/or the combination of square frame.These computer programs can be provided
Instruction arrives the processor of general purpose computer, special-purpose computer, Embedded Processor or other programmable data processing device to produce
A raw machine so that the instruction performed by the processor of computer or other programmable data processing device is produced for real
The device of the function specified in one flow process of flow chart or multiple flow process and/or one square frame of block diagram or multiple square frame now.
These computer program instructions may be alternatively stored in and computer or other programmable data processing device can be guided with spy
Determine in the computer-readable memory that mode works so that the instruction being stored in this computer-readable memory produces and includes referring to
Make the manufacture of device, this command device realize at one flow process of flow chart or multiple flow process and/or one square frame of block diagram or
The function specified in multiple square frames.
These computer program instructions also can be loaded in computer or other programmable data processing device so that at meter
Perform sequence of operations step on calculation machine or other programmable devices to produce computer implemented process, thus at computer or
The instruction performed on other programmable devices provides for realizing at one flow process of flow chart or multiple flow process and/or block diagram one
The step of the function specified in individual square frame or multiple square frame.
Although preferred embodiments of the present invention have been described, but those skilled in the art once know basic creation
Property concept, then can make other change and amendment to these embodiments.So, claims are intended to be construed to include excellent
Select embodiment and fall into all changes and the amendment of the scope of the invention.
Obviously, those skilled in the art can carry out various change and the modification essence without deviating from the present invention to the present invention
God and scope.So, if these amendments of the present invention and modification belong to the scope of the claims in the present invention and equivalent technologies thereof
Within, then the present invention is also intended to comprise these change and modification.
Claims (5)
1. an arrival time Forecasting Methodology, it is characterised in that the method includes:
Set multiple random neural network group's algorithm parameter value；
Read the multiple hourages between the CFS to CFS in historical data base；
Randomly generate m Connectionist model；
After filtering out the Connectionist model that accuracy is less than threshold value, remain k Connectionist model.
2. the method for claim 1, it is characterised in that the method also includes:
Obtain the plurality of hourage between real-time CFS to CFS or the multiple test datas in test phase；
The plurality of hourage or test data are input in k Connectionist model after filtering, and pre-survey station arrives
The multiple hourages stood；
After the plurality of hourage of the CFS to CFS obtaining prediction, it is scaled the time of advent of targeted sites.
3. the method for claim 1, it is characterised in that the plurality of random neural network group's algorithm parameter value includes
Each hidden in hidden layer maximum quantity in Connectionist model quantity, Connectionist model, Connectionist model
Hide layer maximum neuronal quantity, train the training data number of Connectionist model to account for the ratio of total training stage data number, with
And accuracy threshold value.
4. method as claimed in claim 3, it is characterised in that described in filter out the accuracy neural network less than threshold value
After model, the step of k Connectionist model of residue, including:
The accuracy of m the Connectionist model randomly generated being compared with described accuracy threshold value, eliminating is less than
The Connectionist model of described accuracy threshold value.
5. method as claimed in claim 2, it is characterised in that described be input to filter by multiple hourages or test data
After k Connectionist model, and predict multiple hourages of CFS to CFS, including:
Obtain real-time data acquisition system hourage；
Data acquisition system described hourage is input to k Connectionist model after filtering；
Predict multiple prediction hourage, then it is neural that hourage the plurality of prediction of prediction is multiplied by each described class respectively
The weighted value of network model；
The summation of the value after weighting is divided by the summation of weighted value.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW103134687A TWI522974B (en) | 2014-10-06 | 2014-10-06 | Arrival time prediction system and method |
| TW103134687 | 2014-10-06 | ||
| CN201510094269.5A CN104715630B (en) | 2014-10-06 | 2015-03-03 | Arrival time prediction system and method |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510094269.5A Division CN104715630B (en) | 2014-10-06 | 2015-03-03 | Arrival time prediction system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106022541A true CN106022541A (en) | 2016-10-12 |
| CN106022541B CN106022541B (en) | 2020-12-01 |
Family
ID=53414915
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510094269.5A Expired - Fee Related CN104715630B (en) | 2014-10-06 | 2015-03-03 | Arrival time prediction system and method |
| CN201610472352.6A Expired - Fee Related CN106022541B (en) | 2014-10-06 | 2015-03-03 | Arrival time prediction method |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510094269.5A Expired - Fee Related CN104715630B (en) | 2014-10-06 | 2015-03-03 | Arrival time prediction system and method |
Country Status (2)
| Country | Link |
|---|---|
| CN (2) | CN104715630B (en) |
| TW (1) | TWI522974B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018195847A1 (en) * | 2017-04-27 | 2018-11-01 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for determining an estimated time of arrival |
| CN109410580A (en) * | 2018-11-15 | 2019-03-01 | 山东管理学院 | A kind of real-time arrival time prediction technique of public transport and system |
| CN110361019A (en) * | 2018-04-11 | 2019-10-22 | 北京搜狗科技发展有限公司 | For predicting method, apparatus, electronic equipment and the readable medium of navigation time |
| CN110431544A (en) * | 2017-08-10 | 2019-11-08 | 北京嘀嘀无限科技发展有限公司 | Running time and apart from Prediction System and method |
| US11079244B2 (en) | 2017-06-22 | 2021-08-03 | Beijing Didi Infinity Technology And Development Co., Ltd. | Methods and systems for estimating time of arrival |
| CN113706874A (en) * | 2021-10-28 | 2021-11-26 | 华清科盛(北京)信息技术有限公司 | Method and device for predicting arrival time of logistics vehicles in factory in real time and electronic equipment |
| CN113865606A (en) * | 2017-11-23 | 2021-12-31 | 北京嘀嘀无限科技发展有限公司 | System and method for estimating time of arrival |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI591991B (en) * | 2016-01-29 | 2017-07-11 | Chunghwa Telecom Co Ltd | System and method for pre-signing vouchers for forecasting requests for traffic |
| CN105679021B (en) * | 2016-02-02 | 2018-11-06 | 招商局重庆交通科研设计院有限公司 | Journey time fusion forecasting and querying method based on traffic big data |
| CN105513362B (en) * | 2016-02-03 | 2017-11-03 | 东南大学 | A kind of bus platform adjacent area bus running state evaluation verification method |
| CN105489051A (en) * | 2016-02-15 | 2016-04-13 | 青岛海信电器股份有限公司 | Bus station arrival prompting method, mobile terminal and cloud server |
| CN109416878B (en) * | 2017-06-13 | 2022-04-12 | 北京嘀嘀无限科技发展有限公司 | System and method for recommending estimated time of arrival |
| CN109002722B (en) * | 2018-07-19 | 2021-06-29 | 郑州云海信息技术有限公司 | Method and system for estimating vulnerability scanning remaining time |
| CN109920248B (en) * | 2019-03-05 | 2021-09-17 | 南通大学 | Bus arrival time prediction method based on GRU neural network |
| CN110459056A (en) * | 2019-08-26 | 2019-11-15 | 南通大学 | A kind of public transport arrival time prediction technique based on LSTM neural network |
| CN111667689B (en) * | 2020-05-06 | 2022-06-03 | 浙江师范大学 | Method, device and computer device for predicting vehicle travel time |
| US20220051115A1 (en) * | 2020-08-12 | 2022-02-17 | Allstate Insurance Company | Control Platform Using Machine Learning for Remote Mobile Device Wake Up |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101630376A (en) * | 2009-08-12 | 2010-01-20 | 江苏大学 | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process |
| CN102081859A (en) * | 2009-11-26 | 2011-06-01 | 上海遥薇实业有限公司 | Control method of bus arrival time prediction model |
| US20110161261A1 (en) * | 2009-12-28 | 2011-06-30 | Nec(China) Co., Ltd. | Method and system for traffic prediction based on space-time relation |
| CN102854296A (en) * | 2012-08-30 | 2013-01-02 | 北京工业大学 | Sewage-disposal soft measurement method on basis of integrated neural network |
| CN102968670A (en) * | 2012-10-23 | 2013-03-13 | 北京京东世纪贸易有限公司 | Method and device for predicting data |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090062537A (en) * | 2007-12-13 | 2009-06-17 | (주)아이이지컨설팅 | A vehicles allocation control system |
| CN101794520B (en) * | 2010-03-15 | 2012-07-04 | 任兆玉 | Intelligent transportation system based on telecommunication technology and communication method thereof |
| CN102708701B (en) * | 2012-05-18 | 2015-01-28 | 中国科学院信息工程研究所 | System and method for predicting arrival time of buses in real time |
| CN103578267B (en) * | 2012-07-18 | 2015-12-16 | 北京掌行通信息技术有限公司 | Public transport based on public transport gps data is arrived at a station Forecasting Methodology and system |
| CN103885996A (en) * | 2012-12-24 | 2014-06-25 | 北京掌城科技有限公司 | Method of querying real-time bus arrival information |
| CN103236184B (en) * | 2013-04-03 | 2015-02-04 | 广州亿连达电子科技有限公司 | Method and device for automatic bus station report |
| TWM477647U (en) * | 2013-06-20 | 2014-05-01 | Trans Tech Integrate Company Ltd | RFID technique applied to bus stop for display of bus time estimation |
| CN103605725A (en) * | 2013-11-15 | 2014-02-26 | 中国联合网络通信集团有限公司 | Bus arrival time inquiring method, NFC terminal and server |
-
2014
- 2014-10-06 TW TW103134687A patent/TWI522974B/en not_active IP Right Cessation
-
2015
- 2015-03-03 CN CN201510094269.5A patent/CN104715630B/en not_active Expired - Fee Related
- 2015-03-03 CN CN201610472352.6A patent/CN106022541B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101630376A (en) * | 2009-08-12 | 2010-01-20 | 江苏大学 | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process |
| CN102081859A (en) * | 2009-11-26 | 2011-06-01 | 上海遥薇实业有限公司 | Control method of bus arrival time prediction model |
| US20110161261A1 (en) * | 2009-12-28 | 2011-06-30 | Nec(China) Co., Ltd. | Method and system for traffic prediction based on space-time relation |
| CN102854296A (en) * | 2012-08-30 | 2013-01-02 | 北京工业大学 | Sewage-disposal soft measurement method on basis of integrated neural network |
| CN102968670A (en) * | 2012-10-23 | 2013-03-13 | 北京京东世纪贸易有限公司 | Method and device for predicting data |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018195847A1 (en) * | 2017-04-27 | 2018-11-01 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for determining an estimated time of arrival |
| US11398002B2 (en) | 2017-04-27 | 2022-07-26 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for determining an estimated time of arrival |
| US11079244B2 (en) | 2017-06-22 | 2021-08-03 | Beijing Didi Infinity Technology And Development Co., Ltd. | Methods and systems for estimating time of arrival |
| CN110431544A (en) * | 2017-08-10 | 2019-11-08 | 北京嘀嘀无限科技发展有限公司 | Running time and apart from Prediction System and method |
| US11536582B2 (en) | 2017-08-10 | 2022-12-27 | Beijing Didi Infinity Technology And Development Co., Ltd. | System and method for estimating travel time and distance |
| CN110431544B (en) * | 2017-08-10 | 2023-04-18 | 北京嘀嘀无限科技发展有限公司 | Travel time and distance estimation system and method |
| CN113865606A (en) * | 2017-11-23 | 2021-12-31 | 北京嘀嘀无限科技发展有限公司 | System and method for estimating time of arrival |
| CN110361019A (en) * | 2018-04-11 | 2019-10-22 | 北京搜狗科技发展有限公司 | For predicting method, apparatus, electronic equipment and the readable medium of navigation time |
| CN110361019B (en) * | 2018-04-11 | 2022-01-11 | 北京搜狗科技发展有限公司 | Method, device, electronic equipment and readable medium for predicting navigation time |
| CN109410580A (en) * | 2018-11-15 | 2019-03-01 | 山东管理学院 | A kind of real-time arrival time prediction technique of public transport and system |
| CN113706874A (en) * | 2021-10-28 | 2021-11-26 | 华清科盛(北京)信息技术有限公司 | Method and device for predicting arrival time of logistics vehicles in factory in real time and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104715630A (en) | 2015-06-17 |
| TW201614607A (en) | 2016-04-16 |
| CN104715630B (en) | 2017-05-10 |
| TWI522974B (en) | 2016-02-21 |
| CN106022541B (en) | 2020-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104715630B (en) | Arrival time prediction system and method | |
| CN105493109B (en) | Inferred using the air quality of multiple data sources | |
| Chen et al. | A short-term traffic prediction model in the vehicular cyber–physical systems | |
| CN102708701B (en) | System and method for predicting arrival time of buses in real time | |
| CN104376716B (en) | Method for dynamically generating bus timetables on basis of Bayesian network models | |
| Akhavian et al. | Evaluation of queuing systems for knowledge-based simulation of construction processes | |
| Zhu et al. | An optimal estimation approach for the calibration of the car-following behavior of connected vehicles in a mixed traffic environment | |
| Daraghmi et al. | Negative binomial additive models for short-term traffic flow forecasting in urban areas | |
| CN105224992A (en) | To waiting for the method and system predicted of ridership and evaluation method and system | |
| Haghani et al. | An optimization model for real-time emergency vehicle dispatching and routing | |
| Wei et al. | Look-ahead insertion policy for a shared-taxi system based on reinforcement learning | |
| CN111507762B (en) | Urban taxi demand prediction method based on multitasking co-prediction neural network | |
| CN112288197B (en) | Intelligent scheduling method and device for station vehicles | |
| CN106503868A (en) | A kind of fleet's layout of roads method and relevant device | |
| Ram et al. | SMARTBUS: A web application for smart urban mobility and transportation | |
| CN108615391A (en) | A kind of intelligent bus dispatching method, system and medium | |
| CN107392389A (en) | Taxi dispatching processing method based on ARIMA models | |
| CN110348614A (en) | It is a kind of obtain passenger OD method and bus passenger flow prediction technique | |
| TWI604324B (en) | Performance evaluation system and method | |
| CN111415026A (en) | Unmanned equipment scheduling device, system and method | |
| Habtie et al. | Artificial neural network based real-time urban road traffic state estimation framework | |
| Zhu et al. | Estimating train choices of rail transit passengers with real timetable and automatic fare collection data | |
| Wilbur et al. | An online approach to solve the dynamic vehicle routing problem with stochastic trip requests for paratransit services | |
| CN105913654B (en) | A kind of Intelligent traffic management systems | |
| CN109409563A (en) | A kind of analysis method, system and the storage medium of the real-time number of bus operation vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20201201 |