CN113947250A - Urban fine-grained flow prediction method and system based on limited data resources - Google Patents
Urban fine-grained flow prediction method and system based on limited data resources Download PDFInfo
- Publication number
- CN113947250A CN113947250A CN202111235268.XA CN202111235268A CN113947250A CN 113947250 A CN113947250 A CN 113947250A CN 202111235268 A CN202111235268 A CN 202111235268A CN 113947250 A CN113947250 A CN 113947250A
- Authority
- CN
- China
- Prior art keywords
- grained
- flow
- coarse
- fine
- data
- 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 48
- 238000009826 distribution Methods 0.000 claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000005070 sampling Methods 0.000 claims abstract description 26
- 239000010419 fine particle Substances 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 12
- 230000009466 transformation Effects 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 230000036962 time dependent Effects 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- ZAKOWWREFLAJOT-CEFNRUSXSA-N D-alpha-tocopherylacetate Chemical compound CC(=O)OC1=C(C)C(C)=C2O[C@@](CCC[C@H](C)CCC[C@H](C)CCCC(C)C)(C)CCC2=C1C ZAKOWWREFLAJOT-CEFNRUSXSA-N 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- 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
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a city fine-grained flow prediction method and system based on limited data resources. The method comprises the following steps: acquiring flow distribution data and corresponding external factor data of a region to be predicted within a certain time; obtaining a fine-granularity flow distribution map and a coarse-granularity flow distribution map according to a set coarse-granularity scaling factor; down-sampling the coarse-grained flow distribution map to obtain a down-sampled coarse-grained flow map; and training a spatial inference coder from the downsampled coarse-grained flow chart to the coarse-grained flow distribution chart for predicting the regional fine-grained flow. Aiming at the condition of limited training resources, the invention provides an inference network to simplify the problem of urban fine-grained flow prediction and improve the flow prediction precision and reliability of the model.
Description
Technical Field
The invention belongs to the technical field of intelligent traffic, and particularly relates to a city fine-grained flow prediction method and system based on limited data resources.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
In recent years, with the advent of the big data age, artificial intelligence technology is widely applied to various levels of society: machine translation, intelligent customer service, smart cities and the like. The intelligent traffic is an important component in the field of smart cities, and needs a city planner to perform global monitoring on urban traffic and give a warning when traffic jam or an accident occurs on an urban road. However, in order to obtain global flow information, a large number of sensing devices need to be deployed on each street of a city, which is convenient for flow monitoring and correspondingly consumes a large amount of power resources. In order to reduce such energy consumption, researchers have proposed urban fine-grained traffic prediction methods that aim to predict fine-grained traffic changes from existing coarse-grained traffic.
The existing urban fine-grained traffic prediction method mostly depends on huge training data, and in practical application, the huge training data is not easy to obtain. If finite data resources are used for training the model, overfitting and poor generalization performance are easily caused, and the accuracy of a prediction result is influenced.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a city fine-grained flow prediction method and system based on limited data resources. A proper flow prediction auxiliary task is designed according to the characteristics of flow data, an inference network is provided to simplify the problem of urban fine-grained flow prediction under the condition of limited training resources, the data is fully utilized in a weight sampling mode by combining the change rule of the flow along with time and the influence of external factors, and the flow prediction precision and reliability of the model are improved.
In order to achieve the above object, one or more embodiments of the present invention provide the following technical solutions:
a city fine-grained flow prediction method based on limited data resources comprises the following steps:
acquiring flow distribution data and corresponding external factor data of a region to be predicted within a certain time;
according to the set coarse-and-fine granularity scaling factor, combining flow distribution data to obtain a fine-granularity flow distribution map and a coarse-granularity flow distribution map; processing the external factor data to obtain an external factor flow chart;
the coarse-grained flow distribution map is subjected to down-sampling according to the scaling factor to obtain a down-sampling coarse-grained flow map; and training a spatial inference coder from the downsampled coarse-grained flow chart to the coarse-grained flow distribution chart for predicting the regional fine-grained flow.
Further, the spatial inference encoder is configured to include a convolutional layer, a convolutional layer with a ReLU activation function, and an upsampling layer connected in sequence; and the training process is provided with structural constraint conditions, and the sum of the regional flows in the downsampling coarse-grained flow chart is required to be equal to the corresponding regional flow in the coarse-grained flow chart.
Further, the time encoder is trained according to the change rule of the flow data along with the time, specifically:
selecting a coarse-grained flow distribution map at a specific moment as an anchor point according to the change rule of flow data along with time, and obtaining a positive sample set and a negative sample set by calculating Euclidean distances between the anchor point and flow maps at other moments; and constructing a new positive sample set and a new negative sample set according to the weight sampling, and training the time encoder through triple comparison learning according to the anchor point, the new positive sample set and the new negative sample set.
Further, external factor data are obtained and processed to obtain an external factor flow chart.
Further, the external features include continuous features and discrete features; the processing the external factor data comprises:
carrying out nonlinear transformation on the discrete features, and carrying out tensor splicing on the discrete features and continuous features;
and carrying out nonlinear transformation and scale transformation on the spliced features to obtain an external factor flow chart.
Further, the method further comprises:
and adding the external factor flow chart and the coarse-grained flow chart to be used as the input of a combined model of the space encoder and the time encoder, and training to obtain a fine-grained flow prediction model for predicting the fine-grained flow of the area.
Further, the fine-grained flow prediction model is configured to include an operation layer, a convolution layer with a ReLu activation function and an upsampling layer which are connected in sequence and used for splicing the space encoder and the time encoder; and the training process is provided with structural constraint conditions, and the sum of the regional flows in the downsampling coarse-grained flow chart is required to be equal to the corresponding regional flow in the coarse-grained flow chart.
One or more embodiments provide a city fine-grained traffic prediction system based on limited data resources, comprising:
the data acquisition module is used for acquiring flow distribution data and corresponding external factor data within a certain time of the area to be predicted;
the coarse-fine particle size flow distribution diagram acquisition module is used for acquiring a fine-fine particle size flow distribution diagram and a coarse-fine particle size flow distribution diagram according to a set coarse-fine particle size scaling factor and by combining flow distribution data; processing the external factor data to obtain an external factor flow chart;
the spatial inference coder pre-training module is used for carrying out down-sampling on the coarse-granularity flow distribution map according to the scaling factor to obtain a down-sampling coarse-granularity flow map; and training a spatial inference coder from the downsampled coarse-grained flow chart to the coarse-grained flow distribution chart for predicting the regional fine-grained flow.
One or more embodiments provide an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the method for urban fine-grained traffic prediction based on limited data resources when executing the program.
One or more embodiments provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method for urban fine-grained traffic prediction based on limited data resources.
The above one or more technical solutions have the following beneficial effects:
the method and the device have the advantages that under the condition of limited training resources, a spatial reasoning network is provided to simplify the urban fine-grained flow prediction problem, the problem of predicting a fine-grained flow graph based on a coarse-grained flow graph is simplified into the problem of predicting a coarse-grained flow graph based on a downsampled coarse-grained flow graph, the required sample data volume is reduced on the premise of ensuring the prediction accuracy, and the operation amount is reduced.
According to the method, the change rule of the flow in the region along with the time is fully utilized, the flow graph and the corresponding positive and negative samples at the anchor point moment are selected, the positive and negative samples are reconstructed in a weight sampling mode, sample data are expanded, limited training resources are fully utilized, the time encoder which accords with the actual change trend of the flow is obtained, and the flow prediction precision and reliability of the model are improved.
In addition, external factors such as weather, temperature and the like are introduced to assist fine-grained flow prediction, a final flow prediction model is obtained by retraining on the basis of a spatial reasoning encoder and a time encoder, the network structure is simple, the model parameters are small, and good feature learning can be carried out on the premise that a training set is limited.
The method designs a suitable flow prediction auxiliary task according to the characteristics of flow data, and provides an inference network in a space angle, and the method focuses on simplifying the flow inference difficulty and learning a flow inference mode; the method provides weight sampling at a time angle for time comparison pre-training, fully utilizes limited data to generate feature data, and establishes a positive sample pair and a negative sample pair, so that features between an anchor point and a positive sample are similar, thereby being far away from the features of the negative sample and being beneficial to accurately predicting the time change rule of the flow. And in the final fine adjustment stage, external factors such as weather, temperature and the like are added, so that the flow prediction precision is greatly improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a flow chart of a city fine-grained traffic prediction method based on limited data resources according to one or more embodiments of the present invention;
FIG. 2 is a schematic illustration of fine-grained traffic prediction for a city according to one or more embodiments of the invention;
FIG. 3 is a graph illustrating traffic flow at different times in accordance with one or more embodiments of the invention.
Detailed Description
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example one
Due to the fact that training data are limited, the traditional method is easy to cause the characteristics of overfitting, poor generalization performance and the like. The embodiment discloses a city fine-grained flow prediction method based on limited data resources, which comprises the following steps:
step 1: and acquiring flow distribution data and corresponding external factor data of the area to be predicted within a certain time.
In this embodiment, the traffic distribution data includes pedestrian distribution data, bicycle distribution data, and vehicle distribution data of the area, and may be obtained from a public website.
In real life, flow prediction is indistinguishable from external factors (weather, temperature, holidays, etc.), such as: in rainy weather, people tend to be more indoors than outdoors. In order to make the urban fine-grained flow prediction accuracy higher, the embodiment acquires certain external factors, where the external factors include continuous features and discrete features, for example, the continuous features include temperature, wind speed, and the like; discrete features include holidays, weather (rainy, cloudy, sunny), and the like.
Step 2: and according to the set coarse-and-fine granularity scaling factor, combining the flow distribution data to obtain a fine-granularity flow distribution map and a coarse-granularity flow distribution map.
The scaling factor N belongs to Z+It is determined from the predicted fine grain requirement that a fine grain flow map of 64 × 64 picture size is predicted from a coarse grain flow map of 32 × 32 picture size, for example, with a scaling factor of 2, as shown in fig. 2. Urban fine-grained flow prediction has strict structural constraint, and requires that the sum of flows in a fine-grained region is strictly equal to the size of corresponding coarse-grained flow, the left graph in fig. 2 is a coarse-grained flow graph (the image size is 32 × 32), the right graph in fig. 2 is a fine-grained flow graph (the image size is 64 × 64), and the flow value in each grid of the coarse-grained flow graph is equal to the sum of flows in corresponding four grids in the fine-grained flow graph.
In order to ensure that the coarse-fine particle size flow distribution maps meet structural constraints, the method for obtaining the coarse-fine particle size flow distribution maps may be as follows: performing grid division on the regional map to obtain a fine-grained flow distribution map; and performing up-sampling on the fine-grained flow distribution map according to the set scaling factor to obtain a coarse-grained flow distribution map. In the fine-grained flow distribution map or the coarse-grained flow distribution map, one grid corresponds to one pixel in the image, and the pixel value is a flow value.
According to coarse grain flow diagram
And fine grain flow maps
And constructing an urban fine-grained flow prediction task:
in the embodiment, the urban fine-grained flow prediction is to infer a fine-grained flow graph according to a coarse-grained flow graph.
And step 3: the coarse-grained flow distribution map is subjected to down-sampling according to the scaling factor to obtain a down-sampling coarse-grained flow map; a spatial inference encoder is trained from a downsampled coarse-grained flow graph to a coarse-grained flow distribution graph.
Under the condition that a data set is limited, the overfitting problem is easily caused by directly predicting the urban fine-grained flow, and the difficulty of predicting the urban fine-grained flow is simplified. The original urban fine-grained flow prediction task is based on a coarse-grained flow chart XcAnd a scaling factor N to calculate a fine-grained flow chart XfBecause of structural constraint of the urban fine-grained flow prediction task, the method requires that a small area in the coarse-grained flow chart is divided into an NxN area in the fine-grained flow chart, and in order to simulate the division process, the embodiment provides a spatial reasoning network, and a 1 → NxN auxiliary task is established in such a way.
Suppose that the fine-grained traffic prediction of the city is based on Xc∈RI×JTo predict X using the coarse-grained flow chart and the scaling factor Nf∈RNI×NJFine grain flow chart, namely, a down-sampling coarse grain flow chart obtained by calculating a scaling factor and a coarse grain flow chart
Then constructing a downsampled coarse grainA spatial inference network from a velocity flow graph to a coarse-grained flow graph. The network simulates the prediction process of urban fine-grained flow, and the flow graph size of the network is smaller than that of the original task, so that the inference difficulty is reduced to a certain extent.
For a downsampled coarse grain flow chart
Firstly, it is coded by 3X 3 convolution with two channels as C to obtain shallow semantic representation
The above operation is named as a spatial inference encoder EncinfThe method comprises the following specific operations:
Hinf＝Encinf(Xmc) (1)
the shallow semantics are nonlinearly changed by using another convolution operation with a ReLU activation function
Hdinf＝ReLu(Conv(Hinf)) (2)
Then the up-sampling operation is used for carrying out size expansion and feature dimension reduction on the data:
Uif＝Upsampling(Hdinf) (3)
wherein U isif∈RH×W×C，
Because urban fine-grained flow prediction has structural constraint, the sum of fine-grained flow areas is required to be strictly equal to the corresponding coarse-grained flow. Therefore, a normalization constraint is designed:
wherein
Is that
And the ith row and the jth column in the same row, and
representing a probability value. Predicted coarse grain flow diagram
Wherein |, indicates element multiplication. The mean square error is calculated as a loss function:
wherein FinfIs an inference network, theta represents a parameter that can be learned in the inference network model.
And 4, step 4: selecting anchor point time according to the change trend of the flow data along with time, calculating corresponding positive samples and negative samples, constructing a new positive sample set and a new negative sample set according to weight sampling, and training a time encoder through triple comparison learning.
The existing urban fine-grained flow prediction method only infers a fine-grained flow graph according to a current coarse-grained flow graph, and ignores the influence on flow inference at different moments. Because the traffic data has proximity and periodicity, similar traffic conditions help in learning the traffic characteristics, and a time weight sampling strategy is designed in the embodiment to assist in traffic inference.
As shown in fig. 3, a curve is drawn according to the total flow of the flow chart at each moment, and the flow data is found to have periodicity and proximity, for example, the moment of point a in the graph (denoted as an anchor point) and its total flow is similar to the moment of point B (denoted as a positive sample), but different from the moments of points C and D (denoted as a negative sample).
By calculating Euclidean distances between anchor points and flow charts at other moments, a positive sample set is easily obtained
Sum negative sample set
The calculation method is as follows:
due to limited training resources, the obtained positive sample set K1And negative sample set K2The number is limited, the single selection of the most similar positive samples and the least similar negative samples cannot fully and effectively utilize data, and a new sample is constructed by selecting a weight sampling mode:
by using the method, the positive sample set is obtained
Constructed as a new positive sample X+Set negative samples
Constructed as a new negative sample X-. Their flow distributions follow respective sample set laws and have positive and negative sample characteristics, given a set of coarse-grained views
Time of flightEncoder EnctcnAnd obtaining high-level semantic representation information, wherein the triple Loss is introduced to enable the anchor point to be more similar to the high-level semantic representation of the positive sample and enable the anchor point to be far away from the high-level semantic representation of the negative sample. Triple Loss is calculated as follows:
where f (-) is a learnable nonlinear transformation and α is a positive boundary parameter.
And 5: processing the external factor data to obtain an external factor flow chart; and adding the external factor flow chart and the coarse-grained flow chart to be used as input of a combined model of the space coder and the time coder, and training to obtain a fine-grained flow prediction model.
Through the pre-training of the step 3 and the step 4, a spatial inference encoder Enc is obtained respectivelyinfAnd a time encoder EnctcnThey contain rich spatio-temporal data features. In the step, the model is corrected by adopting external factors, so that a final prediction model is obtained.
External factors can be divided into continuous and discrete characteristics, for example, continuous characteristics including temperature, wind speed, etc.; discrete features include holidays, weather (rainy, cloudy, sunny), and the like. The processing the external factor data comprises:
non-linear transformation of external discrete features into ecatAnd with the continuation feature econCarrying out tensor splicing operation:
e＝[econ；ecat] (12)
carrying out nonlinear transformation and scale transformation on the external characteristic e to obtain an external factor flow chart Xe∈RI×J。
Training the fine-grained flow prediction model comprises:
mapping external factors to flowXeAnd coarse grain flow diagram XcAdd as model input:
Xce＝Xc+Xe (13)
two pre-trained encoders are used for fine adjustment of a downstream fine-grained flow prediction task, and firstly, splicing operation is carried out on two encoder features:
Hinf＝Encinf(Xce) (14)
Htcn＝Enctcn(Xce) (15)
Ha＝Concat(Hinf,Htcn) (16)
where Concat is the tensor splicing operation yielding Ha∈RH×W×2CThen using convolution with ReLu activation function to make nonlinear transformation to obtain HD∈RH×W×C：
HD＝ReLu(Conv(Ha)) (17)
For H obtained by decoderDAnd performing size expansion and feature dimension reduction on the data by using an upsampling operation:
Uf＝Upsampling(HD) (18)
wherein U isf∈RNH×NW×C，
Because urban fine-grained flow prediction has structural constraint, the sum of fine-grained flow areas is required to be strictly equal to the corresponding coarse-grained flow. Therefore, a normalization constraint is designed:
wherein
Is that
And the ith row and the jth column in the same row, and
representing a probability value. Final prediction fine-grained flow chart
Wherein |, indicates element multiplication. The mean square error is calculated as a loss function:
where F represents the fine tuning model and θ represents a parameter that can be learned in the model.
Through continuous iterative training of the model, parameters in the model are also continuously improved, so that the optimal model on the verification set is saved for later model testing.
Example two
The present embodiment aims to provide a system for predicting fine-grained traffic in a city based on limited data resources, where the system includes:
the data acquisition module is used for acquiring flow distribution data of a region to be predicted within a certain time;
the coarse-fine particle size flow distribution diagram acquisition module is used for acquiring a fine-fine particle size flow distribution diagram and a coarse-fine particle size flow distribution diagram according to a set coarse-fine particle size scaling factor and by combining flow distribution data; processing the external factor data to obtain an external factor flow chart;
the spatial inference coder pre-training module is used for carrying out down-sampling on the coarse-granularity flow distribution map according to the scaling factor to obtain a down-sampling coarse-granularity flow map; and training a spatial inference coder from the downsampled coarse-grained flow chart to the coarse-grained flow distribution chart for predicting the regional fine-grained flow.
As a further optimized technical solution, the system further includes:
the time encoder pre-training module is used for training a time encoder according to the change rule of the flow data along with time;
the forecasting model fine-tuning module is used for acquiring external factor data and processing the external factor data to obtain an external factor flow chart; and adding the external factor flow chart and the coarse-grained flow chart to be used as the input of a combined model of the space encoder and the time encoder, and training to obtain a fine-grained flow prediction model for predicting the fine-grained flow of the area.
EXAMPLE III
The embodiment aims at providing an electronic device.
An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method as described in embodiment one when executing the program.
Example four
An object of the present embodiment is to provide a computer-readable storage medium.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method as set forth in the first embodiment.
The steps involved in the apparatuses of the above second, third and fourth embodiments correspond to the first embodiment of the method, and the detailed description thereof can be found in the relevant description of the first embodiment. The term "computer-readable storage medium" should be taken to include a single medium or multiple media containing one or more sets of instructions; it should also be understood to include any medium that is capable of storing, encoding or carrying a set of instructions for execution by a processor and that cause the processor to perform any of the methods of the present invention.
The invention adopts a mode of pre-training and fine-tuning to predict urban fine-grained flow, and provides two pre-training methods aiming at the time-space characteristics of flow data: in a space angle, a reasoning network is provided, wherein the reasoning network is a simplified version of urban fine-grained flow prediction and simulates the reasoning process of the urban fine-grained flow prediction but simplifies the reasoning difficulty; in a time angle, weight sampling is provided for time comparison pre-training, limited data can be fully utilized by using a weight sampling mode, some similar flow data are generated according to the distribution characteristics of the data, and then a triple comparison pre-training method is used for training. In the fine adjustment stage, external information such as weather, temperature and the like is added to help the fine-grained flow prediction of the city.
Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using general purpose computer means, or alternatively, they can be implemented using program code that is executable by computing means, such that they are stored in memory means for execution by the computing means, or they are separately fabricated into individual integrated circuit modules, or multiple modules or steps of them are fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A city fine-grained flow prediction method based on limited data resources is characterized by comprising the following steps:
acquiring flow distribution data of a region to be predicted within a certain time;
obtaining a fine-granularity flow distribution map according to the flow distribution data, and obtaining a coarse-granularity flow distribution map according to a set coarse-granularity scaling factor;
the coarse-grained flow distribution map is subjected to down-sampling according to the scaling factor to obtain a down-sampling coarse-grained flow map; and training a spatial inference coder from the downsampled coarse-grained flow chart to the coarse-grained flow distribution chart for predicting the regional fine-grained flow.
2. The method of claim 1, wherein the spatial inference encoder is configured to include a convolutional layer, a convolutional layer with a ReLU activation function, and an upsampled layer, which are connected in sequence; and the training process is provided with structural constraint conditions, and the sum of the regional flows in the downsampling coarse-grained flow chart is required to be equal to the corresponding regional flow in the coarse-grained flow chart.
3. The urban fine-grained traffic prediction method based on limited data resources according to claim 1, characterized in that a time encoder is trained according to a time-dependent change rule of traffic data, specifically:
selecting a coarse-grained flow distribution map at a specific moment as an anchor point according to the change rule of flow data along with time, and obtaining a positive sample set and a negative sample set by calculating Euclidean distances between the anchor point and flow maps at other moments; and constructing a new positive sample set and a new negative sample set according to the weight sampling, and training the time encoder through triple comparison learning according to the anchor point, the new positive sample set and the new negative sample set.
4. The urban fine-grained traffic prediction method based on limited data resources as claimed in claim 3, characterized in that external factor data is also obtained and processed to obtain an external factor traffic map.
5. The urban fine-grained traffic prediction method based on limited data resources according to claim 4, characterized in that the external features comprise continuous features and discrete features; the processing the external factor data comprises:
carrying out nonlinear transformation on the discrete features, and carrying out tensor splicing on the discrete features and continuous features;
and carrying out nonlinear transformation and scale transformation on the spliced features to obtain an external factor flow chart.
6. The method for urban fine-grained traffic prediction based on limited data resources according to claim 5, characterized in that the method further comprises:
and adding the external factor flow chart and the coarse-grained flow chart to be used as the input of a combined model of the space encoder and the time encoder, and training to obtain a fine-grained flow prediction model for predicting the fine-grained flow of the area.
7. The urban fine-grained traffic prediction method based on limited data resources of claim 6, wherein the fine-grained traffic prediction model is configured to comprise an operation layer, a convolutional layer with a ReLu activation function and an upsampling layer which are connected in sequence and are used for splicing a spatial encoder and a time encoder; and the training process is provided with structural constraint conditions, and the sum of the regional flows in the downsampling coarse-grained flow chart is required to be equal to the corresponding regional flow in the coarse-grained flow chart.
8. A city fine-grained flow prediction system based on limited data resources is characterized by comprising the following components:
the data acquisition module is used for acquiring flow distribution data and corresponding external factor data within a certain time of the area to be predicted;
the coarse-fine particle size flow distribution diagram acquisition module is used for acquiring a fine-fine particle size flow distribution diagram and a coarse-fine particle size flow distribution diagram according to a set coarse-fine particle size scaling factor and by combining flow distribution data; processing the external factor data to obtain an external factor flow chart;
the spatial inference coder pre-training module is used for carrying out down-sampling on the coarse-granularity flow distribution map according to the scaling factor to obtain a down-sampling coarse-granularity flow map; and training a spatial inference coder from the downsampled coarse-grained flow chart to the coarse-grained flow distribution chart for predicting the regional fine-grained flow.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method for urban fine-grained traffic prediction based on limited data resources according to any one of claims 1 to 7 when executing the program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the method for urban fine-grained traffic prediction based on limited data resources according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111235268.XA CN113947250B (en) | 2021-10-22 | 2021-10-22 | Urban fine granularity flow prediction method and system based on limited data resources |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111235268.XA CN113947250B (en) | 2021-10-22 | 2021-10-22 | Urban fine granularity flow prediction method and system based on limited data resources |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113947250A true CN113947250A (en) | 2022-01-18 |
| CN113947250B CN113947250B (en) | 2024-07-26 |
Family
ID=79332532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111235268.XA Active CN113947250B (en) | 2021-10-22 | 2021-10-22 | Urban fine granularity flow prediction method and system based on limited data resources |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113947250B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662805A (en) * | 2022-05-26 | 2022-06-24 | 山东融瓴科技集团有限公司 | Traffic flow prediction method based on similar time sequence comparison |
| CN117332886A (en) * | 2023-08-29 | 2024-01-02 | 齐鲁工业大学(山东省科学院) | Urban flow prediction method based on contrast learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106447119A (en) * | 2016-10-11 | 2017-02-22 | 济南观澜数据技术有限公司 | Short-term traffic flow prediction method and system based on convolutional neural network |
| CN112101682A (en) * | 2020-09-25 | 2020-12-18 | 北京百度网讯科技有限公司 | Traffic pattern prediction method, traffic pattern prediction device, server, and readable medium |
| CN112766600A (en) * | 2021-01-29 | 2021-05-07 | 武汉大学 | Urban area crowd flow prediction method and system |
| CN112767682A (en) * | 2020-12-18 | 2021-05-07 | 南京航空航天大学 | Multi-scale traffic flow prediction method based on graph convolution neural network |
-
2021
- 2021-10-22 CN CN202111235268.XA patent/CN113947250B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106447119A (en) * | 2016-10-11 | 2017-02-22 | 济南观澜数据技术有限公司 | Short-term traffic flow prediction method and system based on convolutional neural network |
| CN112101682A (en) * | 2020-09-25 | 2020-12-18 | 北京百度网讯科技有限公司 | Traffic pattern prediction method, traffic pattern prediction device, server, and readable medium |
| CN112767682A (en) * | 2020-12-18 | 2021-05-07 | 南京航空航天大学 | Multi-scale traffic flow prediction method based on graph convolution neural network |
| CN112766600A (en) * | 2021-01-29 | 2021-05-07 | 武汉大学 | Urban area crowd flow prediction method and system |
Non-Patent Citations (2)
| Title |
|---|
| GONG YONGSHUN: "Network-wide Crowd Flow Prediction of Sydney Trains via customized Online Non-negative Matrix Factorization", PROCEEDING OF THE 27TH ACM INTERNATIOMAL CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT, 31 December 2019 (2019-12-31) * |
| 陈勐: ""轨迹预测与意图挖掘问题研究"", 《中国优秀博士学位论文全文数据库信息科技辑》, no. 10, 15 October 2016 (2016-10-15) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662805A (en) * | 2022-05-26 | 2022-06-24 | 山东融瓴科技集团有限公司 | Traffic flow prediction method based on similar time sequence comparison |
| CN117332886A (en) * | 2023-08-29 | 2024-01-02 | 齐鲁工业大学(山东省科学院) | Urban flow prediction method based on contrast learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113947250B (en) | 2024-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109658695B (en) | Multi-factor short-term traffic flow prediction method | |
| CN113947250A (en) | Urban fine-grained flow prediction method and system based on limited data resources | |
| CN111062395B (en) | Real-time video semantic segmentation method | |
| CN113159428A (en) | Traffic flow prediction method, electronic device, and storage medium | |
| US20240143999A1 (en) | Multi-modal data prediction method based on causal markov model | |
| CN114418606B (en) | Network vehicle order demand prediction method based on space-time convolution network | |
| CN113962460B (en) | Urban fine granularity flow prediction method and system based on space-time comparison self-supervision | |
| Fernandes et al. | Long short-term memory networks for traffic flow forecasting: exploring input variables, time frames and multi-step approaches | |
| CN115393289A (en) | Tumor image semi-supervised segmentation method based on integrated cross pseudo label | |
| CN116305967A (en) | Sea surface wind speed inversion method based on convolution neural network and multi-modal feature fusion | |
| CN114627441A (en) | Unstructured road recognition network training method, application method and storage medium | |
| CN116663742A (en) | Regional capacity prediction method based on multi-factor and model fusion | |
| CN117477539A (en) | Short-term load prediction method based on time depth convolution network | |
| CN116453343A (en) | Intelligent traffic signal control optimization algorithm, software and system based on flow prediction in intelligent networking environment | |
| CN117079148A (en) | Urban functional area identification method, device, equipment and medium | |
| CN114170519A (en) | High-resolution remote sensing road extraction method based on deep learning and multidimensional attention | |
| CN116861262B (en) | Perception model training method and device, electronic equipment and storage medium | |
| CN118035777A (en) | Air pollutant concentration prediction method and device based on space-time diagram data | |
| CN117649526A (en) | High-precision semantic segmentation method for automatic driving road scene | |
| CN116542391B (en) | Urban area passenger flow volume prediction method based on big data | |
| CN117275238A (en) | Short-time traffic flow prediction method for dynamic graph structure attention mechanism | |
| CN116933931A (en) | Cloud computing double-flow feature interaction electric vehicle charging pile occupation prediction method | |
| CN116977976A (en) | Traffic sign detection method and system based on YOLOv5 | |
| CN115204489A (en) | Urban vehicle speed prediction method and system based on graph attention network and weather weight | |
| CN115995002A (en) | Network construction method and urban scene real-time semantic segmentation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |