CN102034351B - Short-term traffic flow prediction system - Google Patents
Short-term traffic flow prediction system Download PDFInfo
- Publication number
- CN102034351B CN102034351B CN2010102991975A CN201010299197A CN102034351B CN 102034351 B CN102034351 B CN 102034351B CN 2010102991975 A CN2010102991975 A CN 2010102991975A CN 201010299197 A CN201010299197 A CN 201010299197A CN 102034351 B CN102034351 B CN 102034351B
- Authority
- CN
- China
- Prior art keywords
- prediction
- predicted
- module
- data
- highway section
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term traffic flow prediction system, belonging to the field of traffic flow prediction system. A data processing subsystem receives traffic flow data from a traffic information collection platform through the Socket communication, and maintains a location table of the prediction subsystem; the data processing subsystem looks up the prediction subsystem that corresponds to the code in the location table of the prediction subsystem in accordance with the code of the road to be predicted and sends the calculating data to the prediction subsystem after calculating the data of the road to be predicted during the prediction cycle; the data processing subsystem receives the calculating data and maintains the location table of the prediction module on a mainframe; the data processing subsystem looks up in the location table of the prediction module in accordance with the code of the road to be predicted and predicts the road to be predicted; the data processing subsystem predicts the road to be predicted in different prediction duration and obtains the prediction results, and sends the results to a data analysis and display subsystem; and the data analysis and display subsystem receives the prediction results and analyses the prediction value and the actual value of each road to be predicted in each cycle, and calculates the prediction effects and shows the prediction effects.
Description
Technical field
The present invention relates to the system for forecasting traffic flow field, particularly a kind of traffic flow short-term prediction system.
Background technology
In the architecture of intelligent transportation system; Traffic flow short-term prediction system is important ingredient; The main effect of traffic flow short-term prediction system is to predict the following magnitude of traffic flow constantly in certain highway section to be predicted according to present flow rate according to the magnitude of traffic flow situation of history; And will predict the outcome and offer other users and program, comprise traffic control, traffic guidance and transportation information service systems etc.Because short-term traffic flow has very strong non-linear and uncertain; Even for a highway section to be predicted; The difficulty of traffic flow short-term prediction also is very big; And on the entire city road network; Have many highway sections to be predicted, therefore more harsh than the traffic flow short-term prediction method of using in the scientific research for the requirement of the traffic flow short-term prediction system of practicality, in the practical application of traffic flow short-term prediction system, there is following problem usually: 1, seriously rely on artificial data work of treatment and experience setting; 2, be difficult to satisfy the real-time estimate demand of big data quantity; 3, traffic flow short-term prediction system design underaction, algorithm and peripheral module close-coupled are difficult to expansion.
Summary of the invention
In order to address the above problem, the invention provides a kind of traffic flow short-term prediction system, said system comprises: data process subsystem, predicting subsystem and data analysis and display subsystem,
The traffic flow data that said data process subsystem transmits from the traffic information collection platform through the Socket communications reception; Safeguard the predicting subsystem location tables; When highway section to be predicted after the data statistics of predetermined period goes out; Code according to highway section to be predicted is searched the pairing said predicting subsystem of code in said predicting subsystem location tables, through Socket communication statistics is sent to said predicting subsystem; Said predicting subsystem receiving and counting data; Said predicting subsystem is safeguarded the prediction module location tables on main frame; Code according to highway section to be predicted is searched in said prediction module location tables, and highway section to be predicted is predicted, highway section to be predicted is predicted at difference prediction duration; Predicted the outcome, will predict the outcome through Socket communication sends to said data analysis and display subsystem; Said data analysis and display subsystem receive and predict the outcome, and predicted value and the actual value in each highway section to be predicted in each cycle are analyzed, and count prediction effect, and the demonstration prediction effect.
Said predicting subsystem comprises: prediction module manager, highway section prediction module, fundamental forecasting module and road section traffic volume properties of flow acquisition module,
Said prediction module manager is safeguarded said highway section prediction module; On main frame, safeguard the prediction module location tables; Code according to highway section to be predicted is searched in said prediction module location tables; Find the pointer of corresponding said highway section prediction module, use said highway section prediction module to predict; Said highway section prediction module is accomplished the prediction in highway section to be predicted; Said fundamental forecasting module is accomplished the prediction of highway section to be predicted at difference prediction duration; Said road section traffic volume properties of flow acquisition module adopts non parametric regression that predicted flow rate is converted into speed, realizes the prediction to average speed of traffic flow.
Said fundamental forecasting module comprises: library submodule, cache sub-module and configuration are provided with submodule.
Corresponding different predicting demand is set up said library submodule respectively, adopts memory, and as base class, library is inherited base class and realized the Virtual Function of base class regulation with the library class.
That stores in the said cache sub-module is recorded as a tlv triple < W, S, X >, and pattern X is made up of state vector and dependent variable; W is the wait value, and the value of dependent variable is W the predicted value behind predetermined period among the expression pattern X; S=0 representes that pattern X does not exist in said library submodule; S=1 representes that pattern X exists in said library submodule.
Said traffic flow short-term prediction system comprises:
The Data Identification, the forecast demand that are provided with in the submodule according to said configuration read from database and the corresponding field of Data Identification, form the state vector of the current traffic flow modes of expression of corresponding forecast demand;
In said library submodule, search for L pattern with pattern similarity to be matched according to pre-set criteria, wherein L is the quantity of pattern;
Travel through said cache sub-module the wait value of all records is subtracted 1, find out all wait values and be 0 record,, pattern is added said library submodule if S=0 constitutes a complete models with the dependent variable of actual flow value alternative patterns; If S=1 then check the precision that predicts the outcome and adjust accordingly.
Said data analysis and display subsystem comprise: second receiver module, second parse module, second memory module, second statistical module and data analysis and display interface module,
Said second receiver module receives predicting the outcome of each said predicting subsystem through the Socket communication mode; Said second parse module identifies complete data packet according to second preset protocol from predict the outcome; Said second memory module is set up predicted data tabulation and buffer memory in the internal memory to every highway section to be predicted; The predicted data in T cycle of highway section to be predicted and the actual flow data in T cycle are formed data to putting into said predicted data tabulation, with said data to sending to said second statistical module, said data analysis and display interface module and database; The predicted data in T+k cycle of highway section to be predicted is put into said buffer memory; Said second statistical module to adding up into average relative error, predicated error and impartial coefficient, obtains prediction effect with said data; Said data analysis and display interface module receive prediction effect prediction effect are analyzed, and prediction effect is shown.
The beneficial effect of technical scheme provided by the invention is:
Traffic flow short-term prediction provided by the invention system has adopted distributed frame, can bear the forecasting traffic flow task in a plurality of highway sections to be predicted simultaneously, and the fundamental forecasting module can increase and decrease as required; This traffic flow short-term prediction system is a data-driven, does not rely on artificial data to be handled the foundation with implementation pattern storehouse submodule, the operation of can starting from scratch; Fundamental forecasting module in this traffic flow short-term prediction system can conveniently be expanded; Adopt the traffic flow flow-speed-density relationship in the method match highway section to be predicted of non parametric regression, can average the prediction of travelling speed; And abundant view is provided through data analysis and display interface module, be convenient to analyze to predicting the outcome.
Description of drawings
Fig. 1 is the structural representation of traffic flow short-term prediction provided by the invention system;
Fig. 2 is the structural representation of data process subsystem provided by the invention;
Fig. 3 is the synoptic diagram of road section information provided by the invention;
Fig. 4 is the synoptic diagram of velocity information provided by the invention;
Fig. 5 is the synoptic diagram of flow information provided by the invention;
Fig. 6 is the distribution synoptic diagram of predicting subsystem provided by the invention;
Fig. 7 is the structural representation of predicting subsystem provided by the invention;
Fig. 8 is the class design diagram of library submodule provided by the invention;
Fig. 9 is the synoptic diagram that library submodule provided by the invention, cache sub-module and configuration are provided with the submodule cooperating;
Figure 10 is the structural representation of data analysis provided by the invention and display subsystem.
Embodiment
For making the object of the invention, technical scheme and advantage clearer, will combine accompanying drawing that embodiment of the present invention is done to describe in detail further below.
In order to address the above problem, the embodiment of the invention provides a kind of traffic flow short-term prediction system, and referring to Fig. 1, this traffic flow short-term prediction system comprises: data process subsystem, predicting subsystem and data analysis and display subsystem,
The traffic flow data that data process subsystem transmits from the traffic information collection platform through the Socket communications reception; Safeguard the predicting subsystem location tables; When highway section to be predicted after the data statistics of predetermined period goes out; Code according to highway section to be predicted is searched the pairing predicting subsystem of code in the predicting subsystem location tables, through Socket communication statistics is sent to predicting subsystem; Predicting subsystem receiving and counting data; Predicting subsystem is safeguarded the prediction module location tables on main frame; Code according to highway section to be predicted is searched in the prediction module location tables, and highway section to be predicted is predicted, highway section to be predicted is predicted at difference prediction duration; Predicted the outcome, will predict the outcome through Socket communication sends to data analysis and display subsystem; Data analysis and display subsystem receive and predict the outcome, and predicted value and the actual value in each highway section to be predicted in each cycle are analyzed, and count prediction effect, and the demonstration prediction effect.
One, data process subsystem
Referring to Fig. 2, data process subsystem comprises: first receiver module, first parse module, validation verification module, decoder module, first statistical module, packing data module, search the module and first memory module, wherein,
First receiver module, whether have traffic flow data arrive, if receive traffic flow data through Socket communication from the traffic information collection platform, and take out traffic flow data if monitoring; If, do not continue to monitor;
First parse module identifies complete data packet according to first preset protocol from traffic flow data;
Wherein, referring to table 1, the definition of first preset protocol that the embodiment of the invention adopts is specially: packet packet header is 0x68, and length is a byte; The IP address size is 4 bytes; Port numbers length is 2 bytes; Message identifier length is 1 byte; Data length is 1 byte; Packet bag tail is 0x18, and length is a byte.
The definition of table 1 first preset protocol
| Packet header | The IP address size | Port numbers | Message identifier | Data length | The bag tail |
| ?1byte、Ox68 | 4byte | 2byte | 1byte | 1byte | 1byte、Ox18 |
The validation verification module is verified processing according to the length of message header to the integrality of packet, and qualified packet is decoded; To underproof packet, carry out re-send request may or discard processing;
Wherein, the validity of packet mainly is meant the validation verification of packet in communications, comprises the integrity verification of packet and the verification of correctness of packet.Consider that the traffic data collection system has improved quality than higher data, only the integrality of packet has been carried out checking herein and handled according to the length of message header.Wherein, For defective packet; Carrying out re-send request may or discard processing is specially: carry out corresponding re-send request may or discard processing according to different type of messages; If twice re-send request may can not get response in general, think that then data have lost ageingly, directly abandon these data.
Decoder module extracts road section information, velocity information and flow information from packet;
Referring to Fig. 3, Fig. 4 and Fig. 5, provided the data layout of road section information, velocity information and flow information among above-mentioned 3 figure.
First statistical module is added up the traffic flow data that the traffic information collection platform receives according to predetermined period, and the data after the statistics are sent to the packing data module;
The packing data module reorganizes the requirement of the data after the statistics according to predicting subsystem and packing;
Search module, according to the codelookup in highway section to be predicted and the corresponding predicting subsystem of code in highway section to be predicted;
Because the structure of traffic flow short-term prediction system is distributed, the predicting subsystem of being responsible for certain prediction task in highway section to be predicted is distributed on several main frames, therefore must find corresponding predicting subsystem just can predict.Detailed process is following: data process subsystem is safeguarded a predicting subsystem location tables; When certain bar highway section to be predicted after the data of certain predetermined period are come out; In the predicting subsystem location tables, search the pairing predicting subsystem of code in highway section to be predicted according to the code in highway section to be predicted; Obtain the IP address of predicting subsystem place main frame, then with data send to the corresponding main frame of predicting subsystem on.Wherein, In order to obtain seek rate faster; The embodiment of the invention preferably adopts the Hash table structure to make up the predicting subsystem location tables, and with the code in highway section to be predicted key assignments as Hash table, the IP address of predicting subsystem place main frame is as data; Because the code in highway section to be predicted is continuously and do not have repetition, so the time of this Hash table and space consuming optimum all.During concrete the realization, can adopt sequential search method and binary search method etc., the embodiment of the invention does not limit this yet.
Data memory module is stored the deposit data after first statistical module counts in database.
Wherein, the deposit data after the statistics being become permanent storage in database, because the frequency of data storage is very frequent and relate to the I/O operation to database, is the bottleneck of traffic flow short-term prediction system performance therefore.
Two, predicting subsystem
Referring to Fig. 6; The forecast demand of traffic flow short-term prediction system is accomplished by predicting subsystem; In the traffic flow short-term prediction; The most basic forecast demand can be expressed as " flow of prediction highway section i after k predetermined period ", and " k predetermined period " is called the prediction duration, and its unit is exactly predetermined period.Thus it is clear that,, at first need predict the flow in several different highway sections to be predicted for a road net.And, have the demand of several different prediction durations again for a specific highway section to be predicted.Consider from the traffic flow data source; Data on flows is from fixed detecting device; Speed data comes from Floating Car, because Floating Car is being from the middle of constantly moving, therefore to each bar highway section to be predicted on the road network; Not necessarily each predetermined period can both obtain speed data, so the road traffic delay statistics that data process subsystem transmits has two types: i.e. speed-flow information or simple flow information.The highway section prediction module will be carried out different processing respectively according to the difference of data.If speed-flow information then sends to road section traffic volume properties of flow acquisition module with these data on the one hand, be used for obtaining the speed-density-discharge relation of road traffic delay; Data on flows that on the other hand will be wherein sends to each fundamental forecasting module and predicts.If simple flow information then only sends to the fundamental forecasting module with data on flows.Predicting subsystem is a data-driven, and predicting subsystem just is in waiting status usually, when receive a packet from main frame after, just from packet, extracts information, passes to relevant highway section prediction module to these information successively, accomplishes prediction.
Referring to Fig. 7, predicting subsystem comprises: prediction module manager, highway section prediction module, fundamental forecasting module and road section traffic volume properties of flow acquisition module,
The prediction module manager; The highway section prediction module is safeguarded, on main frame, safeguarded the prediction module location tables, in the prediction module location tables, search according to the code in highway section to be predicted; Find the pointer of corresponding road section prediction module, use the highway section prediction module to predict;
Wherein, In order to obtain seek rate faster; The embodiment of the invention preferably adopts the Hash table structure to make up the prediction module location tables; Code with highway section to be predicted is a key assignments, and the different pieces of information of this prediction module location tables partly is quoting of highway section prediction module, utilizes the time complexity O (1) of this highway section prediction module of codelookup in highway section to be predicted.Be interrupted because the code in the highway section to be predicted that the highway section prediction module on main frame is responsible for has, so use Hash table can waste some storage spaces.During concrete the realization, can adopt sequential search method and binary search method etc., the embodiment of the invention does not limit this yet.
The highway section prediction module is accomplished the prediction in highway section to be predicted;
Wherein, Comprise several highway section prediction module in the predicting subsystem; Each highway section prediction module is accomplished the prediction in a certain specific highway section to be predicted, and the fundamental forecasting task of each highway section prediction module is accomplished by different fundamental forecasting modules respectively according to the difference of prediction duration.When the highway section prediction module is to be called for the first time, then utilize configuration file and originate mode storehouse submodule to come this highway section prediction module of initialization.
The fundamental forecasting module is accomplished the prediction of highway section to be predicted at difference prediction duration;
Wherein, The fundamental forecasting module is to accomplish the elementary cell of forecasting traffic flow; Each fundamental forecasting module is accomplished the prediction of highway section to be predicted at difference prediction duration; Because the difference of prediction duration makes a highway section to be predicted that a plurality of forecast demands arranged, this highway section prediction module just has a plurality of fundamental forecasting modules corresponding with it.Each fundamental forecasting module comprises: library submodule, cache sub-module and configuration are provided with submodule, and this three sub-module cooperatively interacts, and accomplish a fundamental forecasting task.
The library submodule corresponding to the predictive mode set of different predicting demand, adopts memory, and as base class, library is inherited base class with the library class, and realizes the Virtual Function of base class regulation.
Wherein, when software is realized, adopt and inherit structure, base class only provides Virtual Function as interface, and the Virtual Function of stipulating in the derived class realization base class is accomplished required function.
Because the difference of forecast demand, the structure of pattern also is different in the library submodule, for example predicts 5 minutes and 10 minutes flows afterwards, and the meaning of the dimension of state vector and each component is inevitable different in the pattern.Therefore for the forecast demand of each different prediction duration, all must set up independent library submodule.Pattern in the library submodule adopts memory fully, at first is because it has directly influenced the speed of M nearest neighbor search, must adopt the fastest memory of access speed.Secondly can estimate the size of library submodule; Through methods such as data pre-service; It is feasible that the dimension of each state vector is controlled at 10 dimensions, adds the auxiliary storage spaces such as left and right sides subtree of dependent variable and KD tree, all takies 4 bytes calculating with each component and pointer; The size of every data is no more than 60byte, supposes that each library submodule has 10
6Individual pattern then occupies the memory headroom of 60MB altogether, and with existing level of hardware, every main frame is enough to support the library submodule of 20～30 this sizes in the left and right sides.And in fact, suppose that predetermined period is 3 minutes, and then 500 groups of measured datas can be collected every day in every highway section to be predicted, even these data can both be formed a brand-new pattern, and then 10
6Individual pattern also needs 2000 days, promptly about 6 years.Considering the pattern that in fact has many repetitions again, is can meet the requirements of fully so the library submodule adopts memory.Since use distribution-free regression procedure carry out forecasting traffic flow need in the library submodule, search for M individual with the most similar pattern of pattern to be predicted; For a kind of specific library sub modular structure; Just need a kind of specific M nearest neighbor search algorithm, therefore when software design, must consider expansion the library submodule.In the native system, regard the library submodule as one " virtual data type (ADT) ", the library submodule not only comprises the physics and the logical organization form of mode data, and has comprised and be based upon the pro forma various algorithms of this data organization.Adopt the design philosophy of facing interface,, at first set up a library class RangeQuery as base class referring to Fig. 8; The adding of pattern; Deletion operation such as is searched and all is presented as its Virtual Function, in library class RangeQuery; These Virtual Functions do not have the function of any reality, and actual functional capability need be in the derived class of RangeQuery rewrites Virtual Function according to concrete implementation to be accomplished.The fundamental forecasting module; Only need to safeguard the pointer of a base class RangeQuery; In fact this pointer points to the instance of certain derived class of RangeQuery, operates according to unified interface (Virtual Function), and need not know the realization details that the library submodule is concrete.KD tree and two kinds of concrete library submodules of linear list have been realized at present in the system; Realize using the library submodule of R tree or other data structures if desired; Only need to inherit RangeQuey and set up derived class Rtree; And several empty methods more than rewriteeing, then the base class in the fundamental forecasting module is quoted the pointer that points to the Rtree instance and get final product.Wherein, library class RegionQuery comprises five main empty methods, and its corresponding code is as follows:
Bool virtual AddData (Pattern Data Data): mode data is added the library submodule, return true after the success.
Bool virtual DeleteData (PatternData Data): mode data is deleted from the library submodule, returned true after the success.
PattemData [] virtual KNNSearch (float Range, int maxM, PatternData center): search is the center with supplemental characteristic center, and Range is a radius, maximum maxM arest neighbors data, and return the data acquisition that searches.
Bool Persistence (string Filename): mode data is stored as file, and defined file is called Filename, returns true after the success.
Bool DePersistence (string Filename): from schema file Filename, read historical pattern, set up library, return true after the success.
Wherein, the distribution-free regression procedure of above-mentioned employing is a universal method of the prior art, and when specifically realizing, the embodiment of the invention does not limit this.
That stores in the cache sub-module is recorded as a tlv triple < W, S, X >, and pattern X is made up of state vector and dependent variable; W is the wait value, and the value of dependent variable is W the predicted value behind predetermined period among the expression pattern X; S=0 representes that pattern X does not exist in the library submodule; S=1 representes that pattern X exists in the library submodule.
Wherein, S=0 representes that pattern X does not exist at present in the library submodule, and then the dependent variable part of X is meaningless, needs W all after date to receive real data and could constitute a complete models afterwards; If S=1 representes this pattern and in the library submodule, exists, still need checking whether accurate, then, W all after date utilize real data that its accuracy is tested after receiving real data.
Configuration is provided with submodule, is the set of data in EMS memory sign, and which data expression should use form the required state vector of certain forecast demand.
Each fundamental forecasting modules configured setting all is to generate through the configuration file that reads in the external memory.Configuration file exists for the form of xml file, the form that it is to the effect that following:
<Modes>
<Mode?Forcast＝″1″>
<road?ID＝″4″BackTrace＝″0″></road>
<road?ID＝″3″BackTrace＝″0″></road>
……
</Mode>
<Mode?Forcast＝″2″>
<road?ID＝″4″Target＝″1″BackTrace＝″1″></road>
……
</Mode>
</Modes>
Node wherein<mode></Mode>Between be a kind of particular prediction demand the state vector that will use form, with first<mode>Node is an example, the flow behind 1 predetermined period of Forcast=" 1 " expression prediction.<road ID=" 4 " BackTrace=" 0 "></road>One-component in the expression state vector is that ID is 4 highway section, pushes back the flow an of predetermined period.For each < Mode>node, all generate a fundamental forecasting module, and utilize this modules configured of content generation between < Mode>node that submodule is set.For a highway section prediction module, configuration file is necessary.And, therefore adopt memory because buffer data size is little and access is frequent.
Referring to Fig. 9, library submodule, cache sub-module and configuration are provided with flow process that the submodule cooperating carries out forecasting traffic flow and see hereinafter for details and describe:
1, the Data Identification, the forecast demand that are provided with in the submodule according to configuration read from database and the corresponding field of Data Identification, form the state vector of the current traffic flow modes of expression of corresponding forecast demand;
Wherein, data are formed the state vector of the current traffic flow modes of expression of corresponding certain forecast demand from reading corresponding field in the data process subsystem in the database.
2, in the library submodule, search for L pattern with pattern similarity to be matched according to pre-set criteria, wherein L is the quantity of pattern;
If in the library submodule, do not search similar pattern, though perhaps searched, at this moment existing icotype just need increase new model for the library submodule very little.At first pattern is added in the buffer memory, S=0, the independent variable unit in the pattern is left a blank.If in the library submodule, search similar pattern, need utilize actual flow value and this predicted value error to carry out feedback modifiers, at this moment need pattern be added in the buffer memory, S=1, the independent variable unit in the pattern is exactly this predicted value.Wherein, pre-set criteria is set according to the concrete applicable cases in the practical application, and when specifically realizing, the embodiment of the invention does not limit this.
3, the traversal cache sub-module subtracts 1 with the wait value of all records, finds out all wait values and be 0 record, if S=0, the complete models of dependent variable formation with actual flow value alternative patterns adds the library submodule with pattern; If S=1, the then precision that predicts the outcome of check and adjusting accordingly.
Wherein, the precision that check predicts the outcome adopts universal method of the prior art to get final product with adjustment, promptly according to the applicable cases of reality the precision that predicts the outcome is adjusted to design requirement, and the embodiment of the invention does not limit this when specifically realizing.
Each highway section prediction module also comprises a road section traffic volume properties of flow acquisition module, and road section traffic volume properties of flow acquisition module adopts non parametric regression that predicted flow rate is converted into speed, realizes the prediction to average speed of traffic flow.Wherein, state vector is expressed as < W, B, Q, O >, and W representes weather conditions, and B representes the ratio of the medium-and-large-sized car of traffic flow, and Q representes flow, and O representes occupation rate.
For practical applications such as traffic guidances, be concerned about it more is the Average Travel Speed on the highway section, and the information that directly dopes is flow, therefore also must possess the function that calculates Average Travel Speed according to data on flows.The basis that this function realizes is the relation between flow-speed in the traffic flow theory-density three; In traffic flow; Flow=this relation of speed * density is permanent the establishment; For wherein also existing the relation that certain is confirmed between two parameters, still the concrete form of this relation is but because difference to some extent such as highway section physical arrangement, pavement behavior, physical environment and vehicle composition must just can be found out concrete relation through gathering real data.Wherein, the ratio of weather conditions and large car is mathematical to the influence of flow-length velocity relation, and because each highway section to be predicted all has independent road section traffic volume properties of flow acquisition module, therefore the situation in highway section to be predicted self has also just implied wherein.Because in traffic flow; Generally, the relation of flow-average velocity is not unique, therefore only depends on data on flows can't obtain unique velocity amplitude; So in state vector, also must add occupation rate O, these four components have just determined a unique velocity amplitude.System is in operation; Can obtain flow, large car ratio and the occupation rate data in highway section to be predicted through the traffic information collection platform; Can obtain the Average Travel Speed in highway section to be predicted through Floating Car; Weather conditions can obtain from some website through checkout equipment artificial, that install in the outfield or web service manner owing to be not the amount of real-time change.If comprise two kinds of information of speed and flow in the data that data process subsystem transmits; Then these information can be given self-contained road section traffic volume properties of flow acquisition module by the highway section prediction module; Road section traffic volume properties of flow acquisition module is formed a traffic stream characteristics pattern with Weather information after receiving these information, just can carry out through distribution-free regression procedure when flow-speed transforms.
Three, data analysis and display subsystem
Data analysis and display subsystem are collected predicting the outcome of each predicting subsystem; Presented with statistical graph is given traffic administration person; And calculate some important statistical indicators; In data analysis and display subsystem, mainly use three statistical indicators, reflected the performance of the different aspect of prediction respectively:
1, average relative error has reflected the average effect of long-time prediction, is the direct indicator of estimating the prediction effect quality, and computing formula is following:
Wherein, F
iThe predicted flow rate of representing i cycle, f
iThe actual flow of representing i cycle, N are represented predetermined period number.
2, predict variance, reflected the stability of prediction effect, computing formula is following:
Wherein, N representes predetermined period number, e
iThe predicated error of representing i the predetermined period in highway section to be predicted,
The predicated error mean value of N predetermined period of expression, wherein e
i=| F
i-f
i|.
3, impartial coefficient has reflected the satisfaction of predicting, impartial coefficient can be thought better prediction greater than 0.9 prediction, and computing formula is following:
Referring to Figure 10, data analysis and display subsystem comprise: second receiver module, second parse module, second memory module, second statistical module and data analysis and display interface module,
Second receiver module receives predicting the outcome of each predicting subsystem through the Socket communication mode;
Wherein, data analysis and display subsystem are less than data process subsystem from the pressure of data throughout, can use fully with data process subsystem in similar techniques realize that the detailed description situation repeats no more at this referring to data process subsystem.
Second parse module identifies complete data packet according to second preset protocol from predict the outcome;
Promptly from packet, extract highway section to be predicted, present flow rate, predicted flow rate and predicted flow rate representative be the flow in which cycle.Wherein, referring to table 2, the definition of second preset protocol that the embodiment of the invention adopts is specially: the packet header of packet is 0x78, and length is 1 byte; Packet length is 1 byte; The code in highway section to be predicted is 2 bytes; Current period is 4 bytes; This cycle flow float; The prediction duration is 1 byte; Predicted flow rate float; The bag tail of packet is 0x18, and length is 1 byte.
The definition of table 2 second preset protocol
Second memory module; Predicted data tabulation and buffer memory in the internal memory are set up in every highway section to be predicted; The predicted data
and the actual flow data
in T cycle in T cycle of highway section R to be predicted are formed data to putting into the predicted data tabulation, data are sent to second statistical module, data analysis and display interface module and database to
and
; The predicted data
in T+k cycle of highway section R to be predicted is put into buffer memory;
Second statistical module; Data are added up into average relative error, predicated error and impartial coefficient to
with
, obtain prediction effect;
Data analysis and display interface module receive prediction effect prediction effect are analyzed, and prediction effect is shown.
Wherein, deposit database in and can be used as permanent storage.The embodiment of the invention provides real-time display mode and user inquiring display mode, and real-time display mode is: when new prediction effect is made, and the real-time display interface that refreshes; The user inquiring display mode is: specify certain time period and concrete zone the user, in database, inquire about and demonstration, the display interface module provides 3 kinds of views:
(1) contrast view: actual value and the predicted value of highway section to be predicted at T predetermined period shown with broken line graph.
(2) error view:, show with broken line graph in actual value of T predetermined period and the relative error between the predicted value with highway section to be predicted.
(3) statistical indicator view: specify the viewing area, in the viewing area, the prediction effect in T predetermined period is shown.
In sum; The embodiment of the invention provides a kind of traffic flow short-term prediction system; This traffic flow short-term prediction system has adopted distributed frame, can bear the forecasting traffic flow task in a plurality of highway sections to be predicted simultaneously, and the fundamental forecasting module can increase and decrease as required; This traffic flow short-term prediction system is a data-driven, does not rely on artificial data to be handled the foundation with implementation pattern storehouse submodule, the operation of can starting from scratch; Fundamental forecasting module in this traffic flow short-term prediction system can conveniently be expanded; Adopt the traffic flow flow-speed-density relationship in the method match highway section to be predicted of non parametric regression, can average the prediction of travelling speed; And abundant view is provided through data analysis and display interface module, be convenient to prediction effect is analyzed.
It will be appreciated by those skilled in the art that accompanying drawing is the synoptic diagram of a preferred embodiment, the invention described above embodiment sequence number is not represented the quality of embodiment just to description.
The above is merely preferred embodiment of the present invention, and is in order to restriction the present invention, not all within spirit of the present invention and principle, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (1)
1. a traffic flow short-term prediction system is characterized in that, comprising: data process subsystem, predicting subsystem and data analysis and display subsystem,
The traffic flow data that said data process subsystem transmits from the traffic information collection platform through the Socket communications reception; Safeguard the predicting subsystem location tables; When highway section to be predicted after the data statistics of predetermined period goes out; Code according to highway section to be predicted is searched the pairing said predicting subsystem of code in said predicting subsystem location tables, through Socket communication statistics is sent to said predicting subsystem; Said predicting subsystem receiving and counting data; On main frame, safeguard the prediction module location tables; Code according to highway section to be predicted is searched in said prediction module location tables, and highway section to be predicted is predicted, highway section to be predicted is predicted at difference prediction duration; Predicted the outcome, will predict the outcome through Socket communication sends to said data analysis and display subsystem; Said data analysis and display subsystem receive and predict the outcome, and predicted value and the actual value in each highway section to be predicted in each cycle are analyzed, and count prediction effect, and the demonstration prediction effect;
Said predicting subsystem comprises: prediction module manager, highway section prediction module, fundamental forecasting module and road section traffic volume properties of flow acquisition module,
Said prediction module manager is safeguarded said highway section prediction module; On main frame, safeguard the prediction module location tables; Code according to highway section to be predicted is searched in said prediction module location tables; Find the pointer of corresponding said highway section prediction module, use said highway section prediction module to predict; Said highway section prediction module is accomplished the prediction in highway section to be predicted; Said fundamental forecasting module is accomplished the prediction of highway section to be predicted at difference prediction duration; Said road section traffic volume properties of flow acquisition module adopts non parametric regression that predicted flow rate is converted into speed, realizes the prediction to average speed of traffic flow;
Said fundamental forecasting module comprises: library submodule, cache sub-module and configuration are provided with submodule;
Corresponding different predicting demand is set up said library submodule respectively, adopts memory, and as base class, library is inherited base class and realized the Virtual Function of base class regulation with the library class;
That stores in the said cache sub-module is recorded as a tlv triple < W, S, X >, and pattern X is made up of state vector and dependent variable; W is the wait value, and the value of dependent variable is W the predicted value behind predetermined period among the expression pattern X; S=0 representes that pattern X does not exist in said library submodule; S=1 representes that pattern X exists in said library submodule;
Said traffic flow short-term prediction system comprises:
The Data Identification, the forecast demand that are provided with in the submodule according to said configuration read from database and the corresponding field of Data Identification, form the state vector of the current traffic flow modes of expression of corresponding forecast demand;
In said library submodule, search for L pattern with pattern similarity to be matched according to pre-set criteria, wherein L is the quantity of pattern;
Travel through said cache sub-module the wait value of all records is subtracted 1, find out all wait values and be 0 record,, pattern is added said library submodule if S=0 constitutes a complete models with the dependent variable of actual flow value alternative patterns; If S=1 then check the precision that predicts the outcome and adjust accordingly;
Said data analysis and display subsystem comprise: second receiver module, second parse module, second memory module, second statistical module and data analysis and display interface module,
Said second receiver module receives predicting the outcome of each said predicting subsystem through the Socket communication mode; Said second parse module identifies complete data packet according to second preset protocol from predict the outcome; Said second memory module is set up predicted data tabulation and buffer memory in the internal memory to every highway section to be predicted; The predicted data in T cycle of highway section to be predicted and the actual flow data in T cycle are formed data to putting into said predicted data tabulation, with said data to sending to said second statistical module, said data analysis and display interface module and database; The predicted data in T+k cycle of highway section to be predicted is put into said buffer memory; Said second statistical module to adding up into average relative error, predicated error and impartial coefficient, obtains prediction effect with said data; Said data analysis and display interface module receive prediction effect prediction effect are analyzed, and prediction effect is shown.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010102991975A CN102034351B (en) | 2010-09-30 | 2010-09-30 | Short-term traffic flow prediction system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010102991975A CN102034351B (en) | 2010-09-30 | 2010-09-30 | Short-term traffic flow prediction system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102034351A CN102034351A (en) | 2011-04-27 |
| CN102034351B true CN102034351B (en) | 2012-07-25 |
Family
ID=43887189
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010102991975A Expired - Fee Related CN102034351B (en) | 2010-09-30 | 2010-09-30 | Short-term traffic flow prediction system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102034351B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021665B (en) * | 2014-06-06 | 2016-09-07 | 河南理工大学 | Genetic search Short-time Traffic Flow Forecasting Methods |
| CN106128101B (en) * | 2016-06-29 | 2018-09-04 | 南京城建隧桥经营管理有限责任公司 | The visualization device of magnitude of traffic flow short-term prediction result |
| CN107086989B (en) * | 2017-03-30 | 2020-08-21 | 努比亚技术有限公司 | Data packet identification system and method |
| CN112397146B (en) * | 2020-12-02 | 2021-08-24 | 广东美格基因科技有限公司 | Microbial omics data interaction analysis system based on cloud platform |
| CN112802331A (en) * | 2020-12-30 | 2021-05-14 | 青岛中兴智能交通有限公司 | System and method for judging urban traffic travel rule according to parking data |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101488284A (en) * | 2008-01-16 | 2009-07-22 | 闵万里 | Intelligent management system for road traffic condition instant prediction |
| CN101286269A (en) * | 2008-05-26 | 2008-10-15 | 北京捷讯畅达科技发展有限公司 | Traffic flow forecasting system combining dynamic real time traffic data |
| CN101593424B (en) * | 2009-07-10 | 2010-09-29 | 浙江大学 | Intelligent combination forecasting method for short-term traffic flow |
| CN101739819A (en) * | 2009-11-19 | 2010-06-16 | 北京世纪高通科技有限公司 | Method and device for predicting traffic flow |
| CN101789176B (en) * | 2010-01-22 | 2013-09-04 | 天津市市政工程设计研究院 | Forecasting method for port area short-time traffic flow under model of reservation cargo concentration in port |
-
2010
- 2010-09-30 CN CN2010102991975A patent/CN102034351B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN102034351A (en) | 2011-04-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102034351B (en) | Short-term traffic flow prediction system | |
| Cervero | Road expansion, urban growth, and induced travel: A path analysis | |
| CN102163225A (en) | A fusion evaluation method of traffic information collected based on micro blogs | |
| CN105868861A (en) | Bus passenger flow evolution analysis method based on time-space data fusion | |
| Balakrishna et al. | Time-dependent origin-destination estimation without assignment matrices | |
| CN101794507A (en) | Method for evaluating macroscopic road network traffic state based on floating car data | |
| CN103747523A (en) | User position predicating system and method based on wireless network | |
| CN106447119A (en) | Short-term traffic flow prediction method and system based on convolutional neural network | |
| CN109859495A (en) | A method of overall travel speed is obtained based on RFID data | |
| Anwar et al. | Capturing the spatiotemporal evolution in road traffic networks | |
| CN111160867A (en) | Large-scale regional parking lot big data analysis system | |
| Zou et al. | Passenger flow prediction using smart card data from connected bus system based on interpretable xgboost | |
| Kucukyilmaz et al. | A machine learning approach for result caching in web search engines | |
| Chen et al. | Spatio-temporal top-k term search over sliding window | |
| CN109003128A (en) | Based on improved random forest public bicycles website Demand Forecast method | |
| CN104899702A (en) | Big data-based decoration budget quota management system and method | |
| Willumsen | An entropy maximising model for estimating trip matrices from traffic counts | |
| CN104008150A (en) | Method and system for predicting social network information transmission trend | |
| CN105956816A (en) | Cargo transportation information intelligent processing method | |
| CN106781508B (en) | Short-time Traffic Flow Forecasting Methods based on multiple phase space under a kind of Spark environment | |
| CN115691140B (en) | Analysis and prediction method for space-time distribution of automobile charging demand | |
| Mousa et al. | Enhancing vanet connectivity through a realistic model for rsu deployment on highway | |
| CN113343358B (en) | Electric vehicle charging load space-time distribution modeling method considering road information | |
| CN101968780A (en) | Nonparametric regression method | |
| CN114881430A (en) | Community life convenience evaluation method based on network map service |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120725 Termination date: 20120930 |