CN113935510B - Crowd distribution prediction method, device, equipment and storage medium - Google Patents
Crowd distribution prediction method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN113935510B CN113935510B CN202110975975.6A CN202110975975A CN113935510B CN 113935510 B CN113935510 B CN 113935510B CN 202110975975 A CN202110975975 A CN 202110975975A CN 113935510 B CN113935510 B CN 113935510B
- Authority
- CN
- China
- Prior art keywords
- crowd
- space
- unit
- chain
- obtaining
- 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.)
- Active
Links
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
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06Q10/063—Operations research, analysis or management
- G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
- Y02A30/60—Planning or developing urban green infrastructure
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the field of crowd distribution prediction, in particular to a crowd distribution prediction method, a device, equipment and a storage medium. According to the invention, each space unit on the trip chain is adopted, each space unit forms a multi-stage space unit, and the multi-stage space unit can better reflect the crowd distribution in the space where the space unit is located. The multi-order spatial units are adopted to construct the crowd distribution prediction model, crowd distribution data among all spatial units in the space can be comprehensively reflected, the prediction model with higher accuracy is obtained, and the accuracy of crowd distribution prediction is further improved on the basis of obtaining the prediction model with higher accuracy.
Description
Technical Field
The invention relates to the field of crowd distribution prediction, in particular to a crowd distribution prediction method, a device, equipment and a storage medium.
Background
The change in the number of people on a geospatial unit is a macroscopic manifestation of human movement over a spatial region. The statistical data of the population distribution is widely applied to the fields of urban planning, road traffic, environmental protection, medical treatment and health and the like. The urban population movement characteristics and the evolution law thereof are explored, and important decision basis can be provided for urban land planning and traffic planning, public safety incident risk assessment and emergency command, commercial site selection, advertisement putting and the like. This makes population distribution prediction research a necessary prerequisite for many location-based service applications.
The current crowd distribution prediction method has the following two defects. For one, limited by the high data sampling cost and the rise of privacy protection awareness, we can often obtain only limited sampling spatial interaction stream data for crowd distribution prediction (for example, 10% sampling data). Due to the fact that systematic sampling design is often lacked in the acquisition of passively collected trip chain big data, a space interactive flow network constructed by the data often shows different data characteristics due to different sampling strategies, and the influence is particularly obvious in the fluctuation of low crowd density space units with deficient data. The problem affects the accuracy and stability of constructing the space unit interactive flow network, so that the accuracy of crowd distribution prediction is reduced. Second, the flow of people between the spatial cells can cause a change in the distribution of people. This crowd flow can be macroscopically generalized as an interactive flow between spatial units. The generation of such interactive streams is related to the selection of human motor behavior, and the selection process is directly reflected in the individual trip chain. Existing research tends to generalize an individual trip chain into a series of interactive flows (starting point-end point) between every two space units, and a multi-order interactive process between the space units is omitted. However, the change in population distribution is not only affected by adjacent spatial units in the individual trip chain (first order interaction flows), but also affected by non-adjacent places in the individual trip chain (multi-order interaction flows). For example, when a person goes to work from home via a bus stop, if this travel chain is summarized as a first order interaction flow between home and bus stop and a first order interaction flow between bus stop and work site, then this important multi-order spatial interaction flow between home and work will be ignored. Therefore, in the prior art, only the crowd distribution of two adjacent space units on one trip chain is considered to predict the crowd distribution of the space area where the space units are located, and the influence of a plurality of non-adjacent space units on one trip chain on the crowd distribution in the space is ignored.
In summary, the accuracy of predicting the population distribution in the prior art is low.
Thus, there is a need for improvements and enhancements in the art.
Disclosure of Invention
In order to solve the technical problems, the invention provides a crowd distribution prediction method, a device, equipment and a storage medium, and solves the problem of low accuracy in predicting crowd distribution in the prior art.
In order to achieve the purpose, the invention adopts the following technical scheme:
in a first aspect, the present invention provides a method for predicting a population distribution, including:
acquiring crowd distribution data of each space unit, wherein each space unit is positioned on a trip chain, and the trip chain is used for recording a track formed by the visit of each space unit by a crowd;
according to the crowd distribution data and the trip chain, obtaining a crowd distribution prediction model corresponding to a space area, wherein the space area is composed of each space unit;
and obtaining the crowd distribution prediction data corresponding to the space region according to the crowd distribution prediction model.
In an implementation manner, the obtaining, according to the crowd distribution data and the trip chain, a crowd distribution prediction model corresponding to a spatial area, where the spatial area is formed by each spatial unit, includes:
obtaining a low crowd density unit in the space unit according to the space unit, wherein the low crowd density unit is a space unit with the number of people lower than a set value;
according to the trip chain corresponding to the low crowd density unit, obtaining a trip history chain in the trip chain corresponding to the low crowd density unit;
obtaining a historical space interaction matrix according to the travel historical chain and the crowd distribution data corresponding to the low crowd density unit;
and obtaining a crowd distribution prediction model corresponding to the space region according to the historical space interaction matrix and the crowd distribution data of each low crowd density unit on the trip chain.
In one implementation, the obtaining a crowd distribution prediction model corresponding to the spatial region according to the historical spatial interaction matrix and the crowd distribution data of each low crowd density unit on the trip chain includes:
obtaining a trip simulation chain in the trip chain according to the historical space interaction matrix;
and obtaining a crowd distribution prediction model corresponding to the space region according to the crowd distribution data of each low crowd density unit in the travel simulation chain and the crowd distribution data of each low crowd density unit in the travel history chain.
In an implementation manner, the obtaining a trip simulation chain in the trip chain according to the historical space interaction matrix includes:
obtaining the probability of each low crowd density unit being visited according to the number of times that each low crowd density unit in the historical space interaction matrix is visited by the crowd;
and obtaining the travel simulation chain formed among the low crowd density units according to the visited probability of the low crowd density units.
In one implementation, the obtaining a crowd distribution prediction model corresponding to the spatial region according to the crowd distribution data of each low crowd density unit in the travel simulation chain and the crowd distribution data of each low crowd density unit in the travel history chain includes:
taking each low population density unit as a node in a word vector model, and taking the interaction flow strength among the low population density units as the weight of an edge in the word vector model to obtain a preset word vector model;
training a preset word vector model through the crowd distribution data of each low crowd density unit on the travel simulation chain and the crowd distribution data of each low crowd density unit on the travel history chain to obtain the trained word vector model;
and obtaining a crowd distribution prediction model corresponding to the space region according to the trained word vector model.
In one implementation, the obtaining, according to the trained word vector model, a crowd distribution prediction model corresponding to the spatial region includes:
and inputting the trained word vector model into a graph volume model, and training the graph volume model to obtain a crowd spatial distribution prediction model in the crowd distribution prediction model.
In an implementation manner, the obtaining, according to the trained word vector model, a crowd distribution prediction model corresponding to the spatial region further includes:
and inputting the matrix corresponding to the crowd spatial distribution prediction model into a long-time memory neural network model, and training the long-time memory neural network model to obtain a crowd time change prediction model in the crowd distribution prediction model.
In a second aspect, an embodiment of the present invention further provides a device for a crowd distribution prediction method, where the device includes the following components:
the data acquisition module is used for acquiring crowd distribution data of each space unit, each space unit is positioned on a trip chain, and the trip chain is used for recording a track formed by the crowd visiting each space unit;
the prediction model generation module is used for obtaining a crowd distribution prediction model corresponding to a space area according to the crowd distribution data and the trip chain, wherein the space area is composed of all the space units;
and the prediction data generation module is used for obtaining the crowd distribution prediction data corresponding to the space area according to the crowd distribution prediction model.
In a third aspect, an embodiment of the present invention further provides a terminal device, where the terminal device includes a memory, a processor, and a crowd distribution prediction program that is stored in the memory and is executable on the processor, and when the processor executes the crowd distribution prediction program, the steps of the crowd distribution prediction method described above are implemented.
In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, where a crowd distribution prediction program is stored on the computer-readable storage medium, and when the crowd distribution prediction program is executed by a processor, the steps of the crowd distribution prediction method are implemented.
Has the advantages that: according to the invention, each space unit on the trip chain is adopted, each space unit forms a multi-order space unit, and the multi-order space units can better reflect the crowd distribution in the space where the space units are located. The multi-order spatial units are adopted to construct the crowd distribution prediction model, crowd distribution data among all spatial units in the space can be comprehensively reflected, the prediction model with higher accuracy is obtained, and the accuracy of crowd distribution prediction is further improved on the basis of obtaining the prediction model with higher accuracy.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a detailed flow chart of the present invention.
Detailed Description
The technical scheme of the invention is clearly and completely described below by combining the embodiment and the attached drawings of the specification. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The research shows that the current population distribution prediction method has the following two defects. For one, limited by the high data sampling cost and the rising of privacy protection awareness, we can often obtain only limited sampled spatial interaction stream data for crowd distribution prediction (e.g. 10% sampled data). Due to the fact that systematic sampling design is often lacked in the acquisition of passively collected trip chain big data, a space interactive flow network constructed by the data often shows different data characteristics due to different sampling strategies, and the influence is particularly obvious in the fluctuation of low crowd density space units with deficient data. The problem affects the accuracy and stability of constructing the space unit interactive flow network, so that the accuracy of crowd distribution prediction is reduced. Second, the flow of people between the spatial cells can cause a change in the distribution of people. This crowd flow can be macroscopically generalized as an interactive flow between spatial units. The generation of such interactive streams is related to the selection of human motor behavior, and the selection process is directly reflected in the individual trip chain. Existing research tends to generalize an individual trip chain into a series of interactive flows (starting point-end point) between every two space units, and a multi-order interactive process between the space units is omitted. However, the change in population distribution is affected not only by adjacent spatial units in the individual trip chain (first order interaction flows), but also by non-adjacent places in the individual trip chain (multi-order interaction flows). For example, when a person goes to work from home via a bus stop, if this travel chain is summarized as a first order interaction flow between home and bus stop and a first order interaction flow between bus stop and work site, this important multi-order spatial interaction flow between home and work will be ignored. Therefore, in the prior art, only the crowd distribution of two adjacent space units on one trip chain is considered to predict the crowd distribution of the space area where the space units are located, and the influence of a plurality of non-adjacent space units on one trip chain on the crowd distribution in the space is ignored.
In order to solve the technical problems, the invention provides a crowd distribution prediction method, a device, equipment and a storage medium, and solves the problem of low accuracy of crowd distribution prediction in the prior art. When the crowd distribution at the next moment of the space area needs to be predicted, the crowd distribution data at the current moment of the space area is input into the prediction model, and then the crowd distribution at the next moment of the space area can be predicted. The multi-order spatial units are adopted to construct the crowd distribution prediction model, crowd distribution data among all spatial units in the space can be comprehensively reflected, the prediction model with higher accuracy is obtained, and the accuracy of crowd distribution prediction is further improved on the basis of obtaining the prediction model with higher accuracy.
For example, block a (space block) has many bus stops (space units), and a person arrives at bus stop g from bus stop a via bus stops b, c, d, e, f. a. b, c, d, e, f and g form a trip chain, and the person only influences the crowd distribution of the bus stop a at the starting station and the bus stop g at the terminal station and influences the crowd distribution of the bus stops b, c, d, e and f between the bus stop a and the bus stop g. In the embodiment, when the crowd distribution of the block a is modeled, the crowd distribution data of the a space unit, the b space unit, the c space unit, the d space unit, the e space unit, the f space unit and the g space unit on the trip chain are comprehensively considered, so that when the crowd distribution prediction model of the embodiment is used for predicting the crowd distribution of the block a, the crowd distribution data of each bus stop can be predicted.
Exemplary method
The crowd distribution prediction method can be applied to terminal equipment. In this embodiment, as shown in fig. 1, the method for predicting the crowd distribution specifically includes the following steps:
s100, acquiring crowd distribution data of each space unit, wherein each space unit is located on a trip chain, and the trip chain is used for recording a track formed by the crowd visiting each space unit.
The spatial units in this embodiment are obtained by dividing the spatial regions, and the trip chain is a track formed by the movement of an individual between two spatial units or between several spatial units.
For example, the following steps are carried out: city C includes streets C1 and C2, and C1 and C2 are the spatial zones in this embodiment, but this embodiment may also partition city C according to traffic and regular grids. C1 includes C11 space unit, C12 space unit, C13 space unit, C2 includes C21 space unit, C22 space unit, C23 space unit, and the individuals go from C11 space unit to C12 space unit to C13 space unit, so as to form a C11 to C13 travel chain, and when the individuals go from C12 to C13 to C21, in addition to forming a travel chain from C12 to C21, a travel chain from C1 to C2 is derived, so as to increase the data amount of the embodiment.
And S200, obtaining a crowd distribution prediction model corresponding to the space region according to the crowd distribution data and the trip chain.
The embodiment is directed to a low crowd density spatial unit, and due to the lack of sampling design in passive trip chain data acquisition, the extracted spatial interaction mode may have great difference. This makes the spatial interaction characteristics in less crowded places insufficient, thereby reducing the accuracy of the crowd distribution prediction. In order to solve the problem of data shortage, the embodiment adopts a weighted random walk strategy to enhance the data of the trip chain. Step S200 includes steps S201, S202, S203, and S204 as follows:
s201, obtaining a low crowd density unit in the space unit according to the space unit, wherein the low crowd density unit is a space unit with the number of people lower than a set value.
In this embodiment, the low population density unit is determined when the spatial unit personnel data is lower than a certain value, and the trip chain generated by the individual in the spatial unit is limited, which is not beneficial to subsequent modeling.
S202, obtaining a travel history chain in the travel chain corresponding to the low crowd density unit according to the travel chain corresponding to the low crowd density unit.
The individual travel history chain of the low crowd density unit is comprehensively considered in the embodiment, and the individual travel history chain can reflect the influence of individuals on the crowd distribution data of each space unit.
S203, obtaining a historical space interaction matrix according to the travel history chain and the crowd distribution data corresponding to the low crowd density unit.
The embodiment forms a pedestrian volume matrix G according to the historical pedestrian volume of the space unit reflected by the finite sampling dataMConstructing physical nullsThe inter-movement interaction map GM is used for modeling the spatial mode of the distribution change of the crowd. The pedestrian volume matrix is generated by deducing individual trip chains, and when the space units of two adjacent trip chains of the same individual are different, the movement is identified as one movement of different space units (the individual historical trip chain only has two trip chains, one of the two trip chains is from the space unit a to the space unit b of the space area A, and the other is from the space unit C to the space unit d of the space area C. since the individual has only two historical trip chains, the data is too little to be beneficial to subsequent prediction, the trip chain from the space area A to the space area C of the third trip chain is derived on the basis of the two trip chainsi,jRepresenting the historical flow of people after the convergence between two spatial units i, j.
And S204, obtaining a crowd distribution prediction model corresponding to the space region according to the historical space interaction matrix and the crowd distribution data of each low crowd density unit on the trip chain.
The historical interaction matrix covers the crowd distribution data brought by a plurality of individual trip chains, so that the crowd distribution data can be used for obtaining a crowd distribution prediction model. For low-density crowds, an existing trip chain of the low-density crowds is still insufficient to obtain a high-precision crowd distribution prediction model, in order to solve the above problem, in this embodiment, a simulation chain is derived from the trip chain, so as to increase the data amount of the trip chain, and the process of obtaining the simulation chain is to obtain the probability of each low-crowd-density unit being visited according to the number of times that each low-crowd-density unit in the historical space interaction matrix is visited by crowds, and obtain the trip simulation chain formed among the low-crowd-density units according to the probability of each low-crowd-density unit being visited. The following describes the process of acquiring the trip simulation chain in detail:
in combination with the historical spatial interaction matrix, the trip chain simulation can be regarded as a weighted random walking process. Given a current spatial unit SmAnd the space unit S that the individual has previously visitedlAccess the next space unit SnThe probability of (c) can be defined as follows.
P(St+1＝Sn|St＝Sm)＝α(l,n)*wm,n/Z
Where p denotes the access probability, wm,nRepresenting the corresponding edge weights in the physical space motion interaction diagram, Z representing the normalization coefficient, dl,nRepresenting a space element SlAnd SnThe minimum number of edges in between. In the implementation, the parameter p and the parameter q are distributed as constants 0.25 and 2, wherein the larger the value of the parameter p is, the lower the probability of accessing the accessed space unit is; the parameter q controls movement in an "inward" direction, i.e. the individual moves from a cell adjacent to a spatial cell towards the spatial cell, or an "outward" direction, i.e. the individual moves from a cell adjacent to a spatial cell away from the spatial cell, the higher the parameter is such that access is close to SlThe lower the probability of the spatial cell of (a). And selecting a low population density space unit as an initial node based on the proposed trip chain simulation strategy, and generating a simulation trip chain.
After the simulated trip chain is obtained, the individual trip chain data is actually enhanced on the basis of the trip history chain so as to construct a word vector model (multi-order interactive flow network), and the simulated trip chain and the trip history chain are used for constructing the multi-order interactive flow network. Constructing the multi-order interactive flow network comprises constructing a spatial unit embedded expression model and a multi-order spatial interactive flow intensity matrix, which are introduced respectively as follows:
constructing a spatial unit embedded expression model: giving an individual trip chain Trajk＝S1,S2,…,SHThe core idea of the embedded expression is to maximize the average log probability of correctly accessing the next spatial unit (i.e., taking the maximum value as follows):
where K represents the order of the multi-order spatial interaction to be considered in the model, P (S)t+k|St) Is represented in the current spatial unit StIn the case of (2), the visited place S is correctly predictedt+kIs defined as follows:
where V represents the total number of spatial cells,
the initial N-dimensional vector of the space unit adopts a back propagation rule in the training process.
Constructing a multi-order spatial interaction flow intensity matrix: and embedding the spatial units based on the spatial units trained in the previous step into an expression model, and expressing each spatial unit as an N-dimensional vector. The intensity of the interactive flow of the multi-order space unit can be calculated through the similarity between vectors, and the calculation formula is as follows:
wherein, MSI(i,j)Representing a space element SiAnd SjThe strength of the inter-flow of the multi-level space unit (when the individual is in the space unit S)iWill continue to access the space unit SjThe possibility of (d),
and
respectively represent the trained spatial unit vectors
And
the die of (1). MSI(i,j)The higher the value is, the two spatial units have more frequent co-occurrence relationship or common upstream and downstream relationship in the massive individual trip chain, that is, the strength of the multi-order spatial interaction flows at the two places is higher.
After the multi-order space interactive flow intensity matrix is constructed, the alternating intensity among all the space units can be obtained, and the alternating intensity is used as the edge weight of the multi-order interactive flow network and the space units are used as the nodes of the multi-order interactive flow network to complete the construction of the multi-order interactive flow network.
In this embodiment, a crowd distribution prediction model starts to be constructed after a multi-order interaction flow network is constructed, the crowd distribution prediction model of this embodiment includes a crowd spatial distribution prediction model and a crowd temporal change prediction model, and the two models are respectively described below:
the crowd spatial distribution prediction model method adopts the graph convolution model to model the crowd change spatial mode, and uses the current crowd distribution and the multi-order interaction flow network constructed in the step S200 as the input of the graph convolution module:
wherein, l represents the number of convolutions,
representing an activation function, F representing a multi-level interaction flow network, H(l+1)Representing a population distribution matrix, W(l)Representing an initial random weight matrix for model training.
The crowd time change prediction model models a crowd change time mode by using a long-short time memory network, and a space mode modeling result matrix is used as the input of the long-short time memory neural network. The formula is as follows:
wherein f ist,it,otRespectively represent a forgetting layer, an input layer, an output layer, CtRepresents the state of the cell, and represents the state of the cell,
represents a candidate value, Wf，Wi，Wc，WoRespectively representing the weight matrix of the corresponding layer, bf，bi，bc，boRepresents the corresponding layer residual, H(l+1)Is a population distribution matrix, htAnd (4) presetting a prediction result for the crowd distribution.
S300, obtaining the crowd distribution prediction data corresponding to the space area according to the crowd distribution prediction model.
The obtained crowd distribution data of the space area at the current moment is input into the obtained crowd distribution prediction model, so that the crowd distribution data of each space unit in the space area and the crowd distribution data of the next moment can be obtained, and the data provide technical means for mastering the law of human movement behaviors, thereby providing substantive data support for applications such as short-time traffic prediction, epidemiological investigation and the like. Meanwhile, the prediction method provided by the invention can also provide method support for sociological research such as human movement behavior characteristics and group behavior rules, and the like, help to understand the complex relationship between the urban environment and human activities, and further improve the service quality based on the position such as path optimization and personalized recommendation, so that the smart city has more 'set' wisdom.
As can be seen from the above steps, the prediction method provided by the present invention includes data collection as shown in fig. 2, where the data collection is to collect population distribution data in a spatial region, and construct an individual trip chain from the data, and for low-density people, because the trip chain data is less, the trip chain of the part of people is enhanced by the present invention, so-called trip chain enhancement is to increase the number of trip chains by increasing the trajectory, and construct an interactive flow network after obtaining enough trip chains, and then construct a prediction model according to the interactive flow network.
In summary, the invention adopts each space unit on the trip chain, each space unit constitutes a multi-level space unit, and the multi-level space unit can better reflect the crowd distribution in the space where the space unit is located. The multi-order spatial units are adopted to construct the crowd distribution prediction model, crowd distribution data among all spatial units in the space can be comprehensively reflected, the prediction model with higher accuracy is obtained, and the accuracy of crowd distribution prediction is further improved on the basis of obtaining the prediction model with higher accuracy. In addition, the method is based on a crowd prediction framework of a graph convolution-long-time memory neural network, and is used for fitting the high-dimensional nonlinear relation between the multi-order space interaction flow network and crowd distribution change, so that the future crowd distribution in the city can be accurately predicted.
Exemplary devices
The embodiment also provides a device of the crowd distribution prediction method, which comprises the following components:
the data acquisition module is used for acquiring crowd distribution data of each space unit, each space unit is positioned on a trip chain, and the trip chain is used for recording a track formed by the crowd visiting each space unit;
the prediction model generation module is used for obtaining a crowd distribution prediction model corresponding to a space area according to the crowd distribution data and the trip chain, wherein the space area is composed of all the space units;
a prediction data generation module for obtaining the crowd distribution prediction data corresponding to the space region according to the crowd distribution prediction model
Based on the above embodiment, the present invention further provides a terminal device, which includes.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, databases, or other media used in embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
In summary, the present invention discloses a method, an apparatus, a device and a storage medium for predicting a crowd distribution, wherein the method comprises: acquiring crowd distribution data of each space unit, wherein each space unit is positioned on a trip chain, and the trip chain is used for recording a track formed by the visit of each space unit by a crowd; according to the crowd distribution data and the trip chain, obtaining a crowd distribution prediction model corresponding to a space area, wherein the space area is composed of each space unit; and obtaining the crowd distribution prediction data corresponding to the space region according to the crowd distribution prediction model.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (7)
1. A method for predicting a population distribution, comprising:
acquiring crowd distribution data of each space unit, wherein each space unit is positioned on a trip chain, and the trip chain is used for recording a track formed by the visit of each space unit by a crowd;
according to the crowd distribution data and the trip chain, obtaining a crowd distribution prediction model corresponding to a space area, wherein the space area is composed of each space unit;
according to the crowd distribution prediction model, obtaining crowd distribution prediction data corresponding to the space area;
obtaining a crowd distribution prediction model corresponding to a space region according to the crowd distribution data and the trip chain, wherein the space region is composed of each space unit and comprises:
obtaining a low crowd density unit in the space units according to the space units, wherein the low crowd density unit is a space unit with the number of people lower than a set value;
obtaining a trip history chain contained in the trip chain according to the trip chain;
obtaining a historical space interaction matrix according to the travel historical chain corresponding to the low crowd density unit and the crowd distribution data corresponding to the low crowd density unit;
obtaining the probability of each low crowd density unit being visited according to the number of times that each low crowd density unit in the historical space interaction matrix is visited by the crowd;
obtaining a travel simulation chain formed among the low population density units according to the access probability of the low population density units;
and obtaining a crowd distribution prediction model corresponding to the space region according to the crowd distribution data of each low crowd density unit in the travel simulation chain and the crowd distribution data of each low crowd density unit in the travel history chain.
2. The method of predicting the distribution of people according to claim 1, wherein the obtaining the model of the distribution of people corresponding to the spatial region according to the distribution of people of each of the low population density units in the travel simulation chain and the distribution of people of each of the low population density units in the travel history chain comprises:
taking each low population density unit as a node in a word vector model, and taking the interaction flow strength among the low population density units as the weight of an edge in the word vector model to obtain a preset word vector model;
training a preset word vector model through the crowd distribution data of each low crowd density unit on the travel simulation chain and the crowd distribution data of each low crowd density unit on the travel history chain to obtain the trained word vector model;
and obtaining a crowd distribution prediction model corresponding to the space region according to the trained word vector model.
3. The method according to claim 2, wherein the obtaining a prediction model of the population distribution corresponding to the spatial region according to the trained word vector model comprises:
and inputting the trained word vector model into a graph volume model, and training the graph volume model to obtain a crowd spatial distribution prediction model in the crowd distribution prediction model.
4. The method according to claim 3, wherein the obtaining a prediction model of the population distribution corresponding to the spatial region according to the trained word vector model further comprises:
and inputting the matrix corresponding to the crowd spatial distribution prediction model into a long-time memory neural network model, and training the long-time memory neural network model to obtain a crowd time change prediction model in the crowd distribution prediction model.
5. An apparatus for a method of predicting a distribution of people, the apparatus comprising:
the data acquisition module is used for acquiring crowd distribution data of each space unit, each space unit is positioned on a trip chain, and the trip chain is used for recording a track formed by the crowd visiting each space unit;
the prediction model generation module is used for obtaining a crowd distribution prediction model corresponding to a space area according to the crowd distribution data and the trip chain, wherein the space area is composed of all the space units;
the prediction data generation module is used for obtaining the crowd distribution prediction data corresponding to the space area according to the crowd distribution prediction model;
obtaining a crowd distribution prediction model corresponding to a space region according to the crowd distribution data and the trip chain, wherein the space region is composed of each space unit and comprises:
obtaining a low crowd density unit in the space units according to the space units, wherein the low crowd density unit is a space unit with the number of people lower than a set value;
obtaining a trip history chain contained in the trip chain according to the trip chain;
obtaining a historical space interaction matrix according to the travel historical chain corresponding to the low crowd density unit and the crowd distribution data corresponding to the low crowd density unit;
obtaining the probability of each low crowd density unit being visited according to the number of times that each low crowd density unit in the historical space interaction matrix is visited by the crowd;
obtaining a travel simulation chain formed among the low population density units according to the probability of the low population density units being visited;
and obtaining a crowd distribution prediction model corresponding to the spatial region according to the crowd distribution data of each low crowd density unit in the travel simulation chain and the crowd distribution data of each low crowd density unit in the travel history chain.
6. A terminal device, comprising a memory, a processor, and a crowd distribution prediction program stored in the memory and executable on the processor, wherein the processor, when executing the crowd distribution prediction program, performs the steps of the crowd distribution prediction method according to any one of claims 1-4.
7. A computer-readable storage medium, having a crowd distribution prediction program stored thereon, which when executed by a processor, performs the steps of the crowd distribution prediction method according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110975975.6A CN113935510B (en) | 2021-08-24 | 2021-08-24 | Crowd distribution prediction method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110975975.6A CN113935510B (en) | 2021-08-24 | 2021-08-24 | Crowd distribution prediction method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113935510A CN113935510A (en) | 2022-01-14 |
| CN113935510B true CN113935510B (en) | 2022-05-31 |
Family
ID=79274504
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110975975.6A Active CN113935510B (en) | 2021-08-24 | 2021-08-24 | Crowd distribution prediction method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113935510B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019188458A1 (en) * | 2018-03-29 | 2019-10-03 | Nec Corporation | Method, system, and computer readable medium for performance modeling of crowd estimation techniques |
| CN112418508A (en) * | 2020-11-19 | 2021-02-26 | 中国科学院地理科学与资源研究所 | Population distribution prediction method based on interaction between physical space and social network space |
| CN112418567A (en) * | 2021-01-22 | 2021-02-26 | 中航信移动科技有限公司 | People flow prediction method and device, computer equipment and storage medium |
| CN112766600A (en) * | 2021-01-29 | 2021-05-07 | 武汉大学 | Urban area crowd flow prediction method and system |
| CN113254934A (en) * | 2021-06-29 | 2021-08-13 | 湖南大学 | Binary code similarity detection method and system based on graph matching network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104463364B (en) * | 2014-12-04 | 2018-03-20 | 中国科学院深圳先进技术研究院 | A kind of Metro Passenger real-time distribution and subway real-time density Forecasting Methodology and system |
-
2021
- 2021-08-24 CN CN202110975975.6A patent/CN113935510B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019188458A1 (en) * | 2018-03-29 | 2019-10-03 | Nec Corporation | Method, system, and computer readable medium for performance modeling of crowd estimation techniques |
| CN112418508A (en) * | 2020-11-19 | 2021-02-26 | 中国科学院地理科学与资源研究所 | Population distribution prediction method based on interaction between physical space and social network space |
| CN112418567A (en) * | 2021-01-22 | 2021-02-26 | 中航信移动科技有限公司 | People flow prediction method and device, computer equipment and storage medium |
| CN112766600A (en) * | 2021-01-29 | 2021-05-07 | 武汉大学 | Urban area crowd flow prediction method and system |
| CN113254934A (en) * | 2021-06-29 | 2021-08-13 | 湖南大学 | Binary code similarity detection method and system based on graph matching network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113935510A (en) | 2022-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492830B (en) | Mobile pollution source emission concentration prediction method based on time-space deep learning | |
| Lee et al. | A computerized feature selection method using genetic algorithms to forecast freeway accident duration times | |
| CN111091196B (en) | Passenger flow data determination method and device, computer equipment and storage medium | |
| CN110210656B (en) | Shared bicycle flow prediction method and system based on station behavior analysis | |
| Esquivel et al. | Spatio-temporal prediction of Baltimore crime events using CLSTM neural networks | |
| CN110570035B (en) | People flow prediction system for simultaneously modeling space-time dependency and daily flow dependency | |
| Zéphyr et al. | Approximate stochastic dynamic programming for hydroelectric production planning | |
| Lee et al. | An efficient prediction for heavy rain from big weather data using genetic algorithm | |
| Massaoudi et al. | Performance evaluation of deep recurrent neural networks architectures: Application to PV power forecasting | |
| Zheng et al. | Hybrid deep learning models for traffic prediction in large-scale road networks | |
| Chen et al. | Pedestrian behavior prediction model with a convolutional LSTM encoder–decoder | |
| Ma et al. | Fuzzy hybrid framework with dynamic weights for short‐term traffic flow prediction by mining spatio‐temporal correlations | |
| Moreno et al. | Use of computational intelligence techniques to predict flooding in places adjacent to the Magdalena River | |
| Turan et al. | Predicting monthly river flows by genetic fuzzy systems | |
| Perea et al. | New memory-based hybrid model for middle-term water demand forecasting in irrigated areas | |
| CN113935510B (en) | Crowd distribution prediction method, device, equipment and storage medium | |
| Indra et al. | Modeling of optimal deep learning based flood forecasting model using twitter data | |
| Bak et al. | Deep learning algorithm development for river flow prediction: PNP algorithm | |
| CN112418508B (en) | Population distribution prediction method based on interaction between physical space and social network space | |
| CN109102698B (en) | Method for predicting short-term traffic flow in road network based on integrated LSSVR model | |
| CN116822722A (en) | Water level prediction method, system, device, electronic equipment and medium | |
| CN112766240B (en) | Residual multi-graph convolution crowd distribution prediction method and system based on space-time relationship | |
| Talebi et al. | A framework for real-time operation of urban detention reservoirs: Application of the cellular automata and rainfall nowcasting | |
| Lee et al. | Demand modelling for emergency medical service system with multiple casualties cases: k-inflated mixture regression model | |
| CN114065996A (en) | Traffic flow prediction method based on variational self-coding learning |
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 |