CN110322064B - Urban trip demand prediction method - Google Patents
Urban trip demand prediction method Download PDFInfo
- Publication number
- CN110322064B CN110322064B CN201910570152.8A CN201910570152A CN110322064B CN 110322064 B CN110322064 B CN 110322064B CN 201910570152 A CN201910570152 A CN 201910570152A CN 110322064 B CN110322064 B CN 110322064B
- Authority
- CN
- China
- Prior art keywords
- time period
- travel demand
- area
- region
- city
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000003062 neural network model Methods 0.000 claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 40
- 239000011159 matrix material Substances 0.000 claims description 31
- 238000004458 analytical method Methods 0.000 claims description 11
- 230000003595 spectral effect Effects 0.000 claims description 9
- 230000015654 memory Effects 0.000 claims description 7
- 230000000737 periodic effect Effects 0.000 claims description 7
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 2
- 238000013277 forecasting method Methods 0.000 claims 1
- 239000013598 vector Substances 0.000 description 7
- 230000001419 dependent effect Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000000977 initiatory effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000001413 cellular effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005314 correlation function Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- 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
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Game Theory and Decision Science (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Educational Administration (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Operations Research (AREA)
- Quality & Reliability (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 provides a method for forecasting urban travel demand, which comprises the following steps: for any region in a city, acquiring the travel demand characteristics and the crowd flow characteristics of the region in the previous time period of the current time period; inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period; the graph convolution neural network model is obtained after training based on training sample data and a predetermined travel demand prediction result. The method and the system can accurately predict the urban travel demand.
Description
Technical Field
The invention belongs to the technical field of intelligent transportation, and particularly relates to a city travel demand prediction method.
Background
With the popularization of the mobile internet and the continuous improvement of the urbanization development level, China has made a leap-type development in the shared economy field in recent years. The sharing economy integrates originally dispersed goods or services through the internet platform and provides the users with more flexible market prices, thereby realizing best-of-things and demand allocation.
At present, China has achieved outstanding achievements in the field of shared economy, particularly in the field of shared travel. However, the problem of difficulty in taxi taking is still severe, taxi taking efficiency and taxi taking response rate are reduced, and taxi taking difficulty is increased more obviously in a main people-intensive area in a city. In extreme weather such as heavy rain and high temperature, the response time of taxi taking is prolonged. In order to continuously deal with the problem of difficulty in taxi taking, the existing large data platform resources need to be fully utilized, vehicle resources need to be finely scheduled, and intelligent traffic supply is realized.
Meanwhile, with the rapid development of wireless communication and sensor technologies, wireless mobile terminal devices are spreading explosively. The huge user group provides possibility for researching refined intelligent travel: the method comprises the steps that rich user client position information can be provided through user connection base station information, and population density distribution and travel change rules of different areas can be obtained in time through a big data technology; and secondly, when the user uses the internet trip application, the used data flow can completely record the taxi taking behavior information of the user, so that almost complete trip demand information of the user can be obtained. The position distribution information of the user and the taxi taking behavior information are combined, and a user travel demand model with finer granularity and higher precision can be established.
Although methods for analyzing and predicting urban travel needs exist in the prior art, the studies all face the problems of sparse data coverage and the inability to effectively model spatial relationships between areas. Aiming at the research of data coverage sparseness and based on the data of the traditional taxi company in the early stage, the position of a user for getting on or off the taxi is mainly extracted from GPS track data, so that the taxi taking requirement of the user is analyzed. However, the amount of taxi taking demands inferred from historical GPS data may deviate from the actual taxi taking demands of the user. For the effective modeling of the spatial relationship between the areas, the prior art includes a prediction model based on autoregressive integrated moving average (ARIMA), a prediction model based on long and short term memory network (LSTM), a prediction model based on spatial autocorrelation and the like, but the above schemes either ignore the influence of spatial factors or use an approximate estimation model to model the travel demand, so that the models cannot well reflect the long-distance spatial influence caused by the movement of the user, and thus the high-precision urban travel demand prediction cannot be realized.
Disclosure of Invention
In order to overcome the problem of low prediction accuracy of the existing urban travel demand prediction method or at least partially solve the problem, embodiments of the present invention provide an urban travel demand prediction method.
According to a first aspect of the embodiments of the present invention, there is provided a method for predicting urban travel demand, including:
for any region in a city, acquiring the travel demand characteristics and the crowd flow characteristics of the region in the previous time period of the current time period;
inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period;
the graph convolution neural network model is obtained after training based on training sample data and a predetermined travel demand prediction result.
According to a second aspect of the embodiments of the present invention, there is also provided an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor calls the program instructions to be able to execute the method for predicting urban travel demand provided in any one of the various possible implementations of the first aspect.
The embodiment of the invention provides a city travel demand prediction method, which is used for predicting travel demands by utilizing a graph convolution neural network model established based on a region space dependency relationship according to the time correlation of the travel demands of various regions and the correlation of region crowd flow and the travel demands, and fully considering the influence of the flow characteristics and the historical data of the occurrence demand characteristics of the region crowd on the region travel demands, thereby accurately predicting the city travel demands.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic overall flow chart of a method for predicting urban travel demand according to an embodiment of the present invention;
fig. 2 is a schematic diagram of the degree of cross-correlation between the population outflow and the travel demand characteristics of each area in the urban travel demand prediction method provided by the embodiment of the present invention;
fig. 3 is a schematic structural diagram of a graph convolution neural network in the urban travel demand prediction method according to the embodiment of the present invention;
fig. 4 is a schematic view of an overall structure of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic overall flow chart of a method for predicting urban travel demand according to an embodiment of the present invention, where the method includes: s101, for any area in a city, acquiring the travel demand characteristics and the crowd flow characteristics of the area in the previous time period of the current time period;
the city is further required to be divided into a plurality of areas before travel demand prediction is performed on the city, and the embodiment is not limited to the method for dividing the city area. The travel demand characteristics are characteristics reflecting the travel demands of the users, such as the number of the users needing to travel. The crowd flowing characteristics are characteristics representing crowd flowing, such as the population outflow volume, the population inflow volume and the like. And respectively acquiring the travel demand characteristics and the crowd flow characteristics of each region in the previous time period.
S102, inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period; the graph convolution neural network model is obtained after training based on training sample data and a predetermined travel demand prediction result.
The graph convolution neural network model is established in advance based on the spatial dependence relationship among all regions of a city, and the spatial dependence relationship among the regions is determined by the crowd flow quantity among the regions. And taking the trained graph convolution neural network model as a travel demand prediction model. And inputting the travel demand characteristics and the crowd flow characteristics of the previous time period of each region into a pre-trained graph convolution neural network model, and acquiring the output travel demand prediction result of each region of the city.
According to the embodiment, the travel demand is predicted by utilizing the graph convolution neural network model established based on the regional space dependency relationship according to the time correlation of the travel demands of each region and the correlation of regional crowd flow and the travel demands, the influence of the flow characteristics of the regional crowd and the historical data with the occurrence demand characteristics on the regional travel demands is fully considered, and therefore the urban travel demands are predicted accurately.
On the basis of the foregoing embodiment, in this embodiment, the step of acquiring the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period further includes: acquiring the training sample data, and constructing a directed graph according to the training sample data; constructing a graph convolution neural network model according to the directed graph; inputting the training sample data into the graph convolution neural network model, and training the graph convolution neural network model.
Specifically, before the travel demand of each region of the city is predicted by using a trained graph convolution neural network model, training sample data is obtained, and a directed graph is constructed according to the training sample data. Directed graph D is an ordered triple (V (D), A (D), φ (D)), where φ (D) is a correlation function that makes each element in A (D), called a directed edge or arc, correspond to an ordered pair of elements in V (D), called a vertex or point. In the embodiment of the invention, the spatial dependency relationship of the areas is expressed as a directed graph, elements in the directed graph are areas, and edges connecting different elements express the population outflow volume of the areas and the people flow moving direction of the areas.
And constructing a graph convolution neural network model based on the directed graph. The training sample data comprises input quantity of the graph convolution neural network model and a travel demand prediction result corresponding to the input quantity. The input quantity is node travel demand characteristic samples and crowd flow characteristic samples of multiple time intervals in each region of the city. And inputting training sample data into the graph convolution neural network model for training, and taking the trained graph convolution neural network model as a city trip demand prediction model. During training, sample data is divided into two parts, namely training sample data and test sample data, and training is performed by adopting a cross validation method.
On the basis of the above embodiment, the training sample data in this embodiment includes a crowd flow feature sample and a travel demand feature sample; correspondingly, the step of acquiring the training sample data specifically includes: corresponding to any one of the areas, acquiring mobile network data of the area; the mobile network data of the area comprises a user client set and a flow record, wherein each base station in the area is accessed in a preset time period immediately before the current time period; dividing the preset time period into a plurality of sub time periods, acquiring the position of a user client in each sub time period according to the mobile network data generated in each sub time period, and acquiring the crowd flow characteristic sample of each area in each sub time period according to the position of the user client in each sub time period; and performing network packet capturing on internet trip software in each sub-time period, performing protocol analysis on data packet contents acquired by the network packet capturing, and acquiring trip demand characteristic samples of each area in each sub-time period according to a protocol analysis result and the user client position corresponding to the protocol analysis result.
Specifically, in order to utilize large-scale mobile network data and extract relevant data such as urban trip demands and the like so as to perform effective modeling, protocol analysis of various internet trip software is realized by describing and analyzing the urban trip demands and protocol reverse engineering based on network tracks, so that the description analysis of trip demand data is realized in the mobile network big data. For any area, mobile network data is obtained from data packets sent by user clients in the area to the base station. The mobile network data in the area includes a user client set accessed by each base station in a preset time period immediately before the current time period t and a traffic record, such as a user client ID, a base station ID, a URL (uniform Resource Locator) and a URI (uniform Resource Identifier) of the accessed content. A data packet sent to a base station by a user client connected with a cellular network is recorded based on a cellular network monitoring system, the data packet is stored in a database in a mode of multi-element group (user ID, equipment type, uplink flow, downlink flow, APP type, URL and URI), and incomplete data records in the database are roamed to local records to be cleaned. And estimating the position of the user client through the ID of the base station accessed by the user client, and acquiring the crowd flow sample characteristics of each area in each time period according to the position of the user client in each sub-time period. The method comprises the steps of carrying out network packet capturing on various internet trip software, carrying out protocol analysis on data packet contents obtained by the network packet capturing, matching URL and URI in the data packets with packet capturing analysis results, and obtaining trip demand characteristic samples of each region in each sub-time period. When training is carried out, the travel demand prediction result of the t time period is predicted by using training sample data of k time periods before the t time period. Fig. 2 shows a strong correlation between the regional travel demand and the regional population outflow, which indicates that the regional population flow characteristics can better reflect the regional potential travel demand, and meanwhile, the regional population outflow and the historical data of the travel demand characteristics are input into the convolutional neural network, and the two parts of characteristics are fused for learning, so that the trained convolutional neural network is more accurate.
On the basis of the foregoing embodiment, in this embodiment, the step of obtaining the crowd flow characteristic sample of each area in each sub-time period according to the user client position in each sub-time period specifically includes: for any one of the sub-time periods and any one of the areas, counting the number of users whose user client positions are in the area in the sub-time period before the sub-time period and whose user client positions are not in the area in the sub-time period, and taking the user data as the crowd flow characteristic sample of the area in the sub-time period.
Specifically, the travel demand characteristic of any region c in any time period t is specifically
Wherein,
the trip demand characteristics of the region c in the time period t, namely the total number of the user clients initiating the taxi taking request in the region c in the time period t, AT(c) The user client initiating the taxi taking request in the area c for the period t, u represents the user client initiating the taxi taking request,
and initiating the number of taxi taking requests for the user client u. The sample of the flow characteristics of the population in any region c at any time t is specifically
Wherein
Samples of flow characteristics of the population at any time t representing any region c, BT(c) Is in the t-1 time period region c, and the t time period is in the user client set of other regions. Correspondingly, the travel demand characteristic sample in the input quantity in the training sample is specifically
The crowd flow characteristic sample in the input quantity is specifically
Input quantityThe corresponding travel demand prediction result is specifically
Where d is the dependent edge of the region c vector.
On the basis of the foregoing embodiment, the step of constructing a directed graph according to the training sample data in this embodiment specifically includes: taking each region as a vertex of the directed graph; for any two of the regions, acquiring the population outflow from one of the two regions to the other of the two regions according to the training sample data; taking the direction from the one region to the other region as the direction of the edge between the vertex of the one region and the vertex of the other region; and taking the population outflow from the one area to the other area as the weight of the edge between the vertex of the one area and the vertex of the other area.
Specifically, the present embodiment creates a directed graph according to the spatial dependency relationship among the regions of the city. The directed graph includes a set of regions, a set of edges, and a weight vector for the edges. The directed graph is represented by G ═ C, E, W) G ═ C, E, W; wherein C is a set of regions, E is a set of edges, and W is a weight vector of the edges; the set of edges E includes at least one dependent edge, the dependent edge includes population outflow and an outflow moving direction of the area, and the weight of the dependent edge is the population outflow corresponding to the dependent edge. When in specific drawing, if the user moves from the area a to the area b, an edge is created between the two areas, and the direction can be noted on the edge, namely the direction of the outgoing flow movement of the area. By creating a directed graph between regions, the spatial dependency can be effectively reflected. In addition, the adjacency may be either directly adjacent or indirectly adjacent separated by several regions. If user u is located in region c during the period t-11In the region c during the period t2Then, it represents c1And c2Storing a dependent edge e1,2，e1,2Weight w of1,2Number vector defined as the class of users
Considering the existence of data noise, and the fact that more edges in the directed graph will greatly increase the complexity of the calculation, performance will be degraded. Therefore, the obtained set of edges may be filtered by using a method of setting a threshold, that is, a threshold is set, and only the spatial dependency relationship that the number of 0 in the weight vector is smaller than the threshold is retained.
The graph convolution neural network model is constructed based on a modeled urban area spatial dependence directed graph G, Fourier transformation is carried out on the directed graph G, and a Laplace matrix L of the directed graph G is defined as:
wherein D represents a diagonal matrix and Dii＝∑jWij，InIs an identity matrix, and L is a real symmetric positive semi-definite matrix and contains a complete set of orthogonal feature vectors
I.e., the fourier pattern of the directed graph, and the set of non-negative eigenvalues
I.e. the frequency of the directed graph. Eigenvalues of the directed graph G
May be expressed as gθ(Λ) ═ diag (θ), where θ ∈ RnIs a fourier coefficient vector learned from a neural network. The convolution operation for the graph feature x is defined as
y＝gθ(L)x＝gθ(UΛUT)x＝Ugθ(Λ)UTx
Wherein U ═ U1,…,un]∈Rn×nIs a matrix of eigenvectors, Λ ═ diag ([ λ [ ])0,…,λn-1])∈Rn×nIs the eigenvalue diagonal matrix of the laplacian matrix L of the normalized graph. For larger graphs, the eigenvalue decomposition computation of L is large, and to solve this problem, a polynomial filter g is usedθ(Λ):
Wherein the parameter theta is equal to RKIs a polynomial coefficient matrix and K is a hyperparameter. Due to dG(i,j)>K, indicates (L)K)i,jIs 0, wherein dGIs the shortest path distance. During training, a graph convolution neural network with Laplace K-order polynomial as a spectrum filter is used to realize modeling learning of spatial correlation.
Furthermore, residual connectivity is introduced in the graph convolution neural network. Residual learning, namely, in order to model the spatial dependence of a large-scale city and increase the depth of the network to realize a better prediction effect, an identical mapping layer with y being equal to x is superposed on the basis of a shallow network, so that the network is not degraded along with the increase of the depth; a residual unit containing identity mapping is defined as:
X(l+1)＝X(l)+F(X(l))
wherein X(l)And X(l+1)Which respectively represent the input and output of the l-th layer residual unit, F is a residual method.
In the embodiment, the travel demand prediction model is established based on the graph convolution neural network and residual error learning, and the influence of time-space correlation, people flow characteristics and the like on travel is fully considered, so that the travel demand is accurately predicted.
On the basis of the foregoing embodiments, in this embodiment, the step of acquiring the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period further includes: acquiring travel demand characteristics and crowd flow characteristics of the region in a daily periodic time period before the current period, and inputting the travel demand characteristics and the crowd flow characteristics in the daily periodic time period into a graph convolution neural network model; correspondingly, the step of inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model and outputting the travel demand prediction result of the area in the current time period specifically comprises the following steps: inputting the travel demand characteristics and the crowd flow characteristics into the graph convolution neural network model to obtain first characteristics output by the last convolution layer of the graph convolution neural network model; inputting the travel demand characteristics and the crowd flow characteristics in the daily periodic time period into the graph convolutional neural network model to obtain second characteristics output by the last convolutional layer; and fusing the first characteristic and the second characteristic, and acquiring the travel demand prediction result according to a fusion result.
The daily cycle time period refers to the same time period of each day in preset days before the current time period. The prediction of the travel demands of different regions of the city can be influenced by time correlation and space correlation at the same time, the travel demands of the city often show obvious periodicity, the daily periodic travel demand characteristics and the crowd flow characteristics are extracted for modeling and re-fusion, and the prediction method can be used for predicting the travel demands of different time dimensions, and is shown in fig. 3.
On the basis of the foregoing embodiments, in this embodiment, the step of acquiring the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period further includes: acquiring cross-domain information characteristics of the region in a time period before the current time period; wherein the cross-domain information features include one or more of weather, holidays, and festivals; correspondingly, the step of inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model and outputting the travel demand prediction result of the area in the current time period specifically comprises the following steps: and inputting the travel demand characteristics, the crowd flow characteristics and the cross-domain information characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period.
Specifically, cross-domain data are fused for more accurately predicting the travel demand of each region in a future period of time. Multisource information data such as holidays, holidays and weather conditions have certain influence on the travel demands of different types of areas, the taxi taking demand suddenly increases in general holidays, the taxi taking demand of an office area can be obviously reduced on weekends, and the taxi taking demand is obviously increased in rainy and snowy weather. In order to integrate the cross-domain data information in deep learning, and simultaneously considering the flexibility of a model, a two-layer full-connection network is used for processing external data. The first layer can be considered an embedding layer, and the second layer converts the dimension of external data output to be the same as the dimension of the network output modeled by the spatio-temporal sequence. And then performing output fusion, putting the fusion into a new full-connection network to learn again, and outputting a final result, as shown in fig. 3.
According to the embodiment, the trip requirements of users in different regions of the city at different time periods are accurately predicted by using the large-scale mobile network data, the factors such as space, time, weather and events and the relevance among the factors according to the spatial difference and the dependency of the trip requirements among the regions in the large-scale city.
On the basis of the foregoing embodiments, in this embodiment, before the step of acquiring the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period, the method further includes: dividing the city into a plurality of parts according to the road network of the city; and clustering all parts of the city according to the population outflow of each part of the city, and taking the clustering result as the region division result of the city.
The road network is a road system formed by interconnecting and interlacing various roads in a certain area. Due to the spatial dynamics of urban population traveling, in order to carry out reasonable area division, the influence of road networks and urban pedestrian flow characteristics is fully considered, and urban irregular areas are divided. The present embodiment first performs fine-grained division on cities according to a road network. The divided areas are too small only by utilizing the road network, and in order to increase the size of the divided areas, each area contains a base station, so that the division of the areas is meaningful, and the fine-grained parts of the road network division are clustered by further utilizing the crowd flow characteristics. The crowd flowing characteristics of the region can well reflect the functionality of the region and the cityThe attractiveness of the population. The population flow characteristic of a region is the population outflow of the region, which is the number of people moving from a region to another region. For regional population outflow calculation, P is usedt(ri) Representing the time period t, t + Δ t]Is located in the region riThe total number of people in the area riThe number of candidate people in this time period is Ct(ri)＝Pt(ri)\Pt+1(ri) I.e. if a user client is not in the region for the next time period, it is considered that this time period is leaving the region ri. However, considering that a user client leaves an area, it may leave the area after staying in the area for a long time, or it may just pass on the way and generate data information. In order to reduce the influence of the user on the analysis caused only in the way, the user client is set to count the user client into the outflow population of the area only when the user client stays in the area for more than a preset time, for example, leaves the area after 1 hour, so that the real outflow population is supposed to be selected from the candidate outflow populations.
On the basis of the above embodiment, in this embodiment, the clustering is spectral clustering; correspondingly, according to the population outflow of each part of the city, clustering all parts of the city, and taking the clustering result as the region division result of the city specifically comprises the following steps: corresponding to any two parts of the city, constructing a people stream feature sequence of one part according to the population outflow of one part of the two parts in a plurality of historical time periods, and constructing a people stream feature sequence of the other part according to the population outflow of the other part of the two parts in the plurality of historical time periods; calculating a similarity between the two portions of their respective human stream feature sequences based on minkowski distance; calculating an adjacency matrix of the spectral clustering according to the corresponding similarity of any two parts of the city; wherein, the weight values of two parts which are not adjacent in the adjacent matrix are 0, and the similarity corresponding to the two parts is in direct proportion to the weight values of the two parts; and acquiring a Laplacian matrix of the spectral cluster according to the adjacency matrix, standardizing the Laplacian matrix of the spectral cluster, clustering the standardized Laplacian matrix, and acquiring a region division result of the city.
In particular, for clustering of fine-grained portions, the similarity between the human-stream feature sequences of the portions is computed using Minkowski distances after normalizing the population outflow of the portions. The characteristic sequence of the stream of people like any part is an n-dimensional variable a (x)11,x12,…,x1n) And b (x)11,x12,…,x1n) The minkowski distance between is defined as:
wherein n is the number of the historical time periods, and p is a preset value, for example, p is 2. The embodiment clusters the fine-grained part by using a spectral clustering mode; firstly, a similarity matrix is used for calculating an adjacent matrix in a spectrum cluster, the two parts with small similarity have lower weight values and the parts with large similarity have high weight values, and meanwhile, the weight values exist between the adjacent parts. Since the adjacency matrix W requires an asymmetric matrix, for any riAnd rjTwo-part adjacency matrix Wij：
Wherein alpha isijIs the human-stream characteristic minkowski distance between two regions, which is the standard deviation of all sequences of minkowski distances. A degree matrix D and a Laplace matrix L are obtained through calculation of an adjacency matrix W, the Laplace matrix is normalized, and eigenvectors corresponding to eigenvalues of the normalized Laplace matrix are obtained. And clustering the Laplace matrix standardized by rows by using a traditional clustering mode to form a final urban area division result.
The embodiment provides an electronic device, and fig. 4 is a schematic view of an overall structure of the electronic device according to the embodiment of the present invention, where the electronic device includes: at least one processor 401, at least one memory 402, and a bus 403; wherein,
the processor 401 and the memory 402 communicate with each other via a bus 403;
the memory 402 stores program instructions executable by the processor 401, and the processor calls the program instructions to perform the methods provided by the above method embodiments, for example, the methods include: for any region in a city, acquiring the travel demand characteristics and the crowd flow characteristics of the region in the previous time period of the current time period; inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period; the graph convolution neural network model is obtained after training based on training sample data and a predetermined travel demand prediction result.
The present embodiments provide a non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the methods provided by the above method embodiments, for example, including: for any region in a city, acquiring the travel demand characteristics and the crowd flow characteristics of the region in the previous time period of the current time period; inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period; the graph convolution neural network model is obtained after training based on training sample data and a predetermined travel demand prediction result.
Those of ordinary skill in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
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 (8)
1. A method for predicting urban travel demand is characterized by comprising the following steps:
for any region in a city, acquiring the travel demand characteristics and the crowd flow characteristics of the region in the previous time period of the current time period; the travel demand characteristics comprise the number of users to be traveled;
inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period;
the graph convolution neural network model is obtained by training based on training sample data and a predetermined travel demand prediction result;
wherein, the step of obtaining the travel demand characteristics and the crowd flow characteristics of the area in the previous time quantum of the current time quantum further comprises the following steps:
acquiring the training sample data, and constructing a directed graph according to the training sample data;
constructing a graph convolution neural network model according to the directed graph;
inputting the training sample data into the graph convolution neural network model, and training the graph convolution neural network model;
the step of constructing the directed graph according to the training sample data specifically comprises the following steps:
taking each region as a vertex of the directed graph;
for any two of the regions, acquiring the population outflow from one of the two regions to the other of the two regions according to the training sample data;
taking the direction from the one region to the other region as the direction of the edge between the vertex of the one region and the vertex of the other region;
and taking the population outflow from the one area to the other area as the weight of the edge between the vertex of the one area and the vertex of the other area.
2. The urban travel demand forecasting method according to claim 1, wherein the training sample data comprises crowd flow feature samples and travel demand feature samples;
correspondingly, the step of acquiring the training sample data specifically includes:
corresponding to any one of the areas, acquiring mobile network data of the area; the mobile network data of the area comprises a user client set and a flow record, wherein each base station in the area is accessed in a preset time period immediately before the current time period;
dividing the preset time period into a plurality of sub time periods, acquiring the position of a user client in each sub time period according to the mobile network data generated in each sub time period, and acquiring the crowd flow characteristic sample of each area in each sub time period according to the position of the user client in each sub time period;
and performing network packet capturing on internet trip software in each sub-time period, performing protocol analysis on data packet contents acquired by the network packet capturing, and acquiring trip demand characteristic samples of each area in each sub-time period according to a protocol analysis result and the user client position corresponding to the protocol analysis result.
3. The urban travel demand prediction method according to claim 2, wherein the step of obtaining the crowd flow characteristic samples of each of the regions in each of the sub-time periods according to the user client position in each of the sub-time periods specifically comprises:
for any one of the sub-time periods and any one of the areas, counting the number of users whose user client positions are in the area in the sub-time period before the sub-time period and whose user client positions are not in the area in the sub-time period, and taking the number of users as the crowd flow characteristic sample of the area in the sub-time period.
4. The urban travel demand prediction method according to any one of claims 1 to 3, wherein the step of obtaining the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period further comprises:
acquiring travel demand characteristics and crowd flow characteristics of the region in a daily periodic time period before the current time period, and inputting the travel demand characteristics and the crowd flow characteristics in the daily periodic time period into a graph convolution neural network model;
correspondingly, the step of inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model and outputting the travel demand prediction result of the area in the current time period specifically comprises the following steps:
inputting the travel demand characteristics and the crowd flow characteristics into the graph convolution neural network model to obtain first characteristics output by the last convolution layer of the graph convolution neural network model;
inputting the travel demand characteristics and the crowd flow characteristics in the daily periodic time period into the graph convolutional neural network model to obtain second characteristics output by the last convolutional layer;
and fusing the first characteristic and the second characteristic, and acquiring the travel demand prediction result according to a fusion result.
5. The urban travel demand prediction method according to any one of claims 1 to 3, wherein the step of obtaining the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period further comprises:
acquiring cross-domain information characteristics of the region in a time period before the current time period; wherein the cross-domain information features include one or more of weather, holidays, and festivals;
correspondingly, the step of inputting the travel demand characteristics and the crowd flow characteristics into a graph convolution neural network model and outputting the travel demand prediction result of the area in the current time period specifically comprises the following steps:
and inputting the travel demand characteristics, the crowd flow characteristics and the cross-domain information characteristics into a graph convolution neural network model, and outputting a travel demand prediction result of the region in the current time period.
6. The urban travel demand prediction method according to any one of claims 1 to 3, wherein the step of obtaining the travel demand characteristic and the crowd flow characteristic of the area in the time period before the current time period further comprises:
dividing the city into a plurality of parts according to the road network of the city;
and clustering all parts of the city according to the population outflow of each part of the city, and taking the clustering result as the region division result of the city.
7. The urban travel demand prediction method according to claim 6,
the clustering is spectral clustering;
correspondingly, according to the population outflow of each part of the city, clustering all parts of the city, and taking the clustering result as the region division result of the city specifically comprises the following steps:
corresponding to any two parts of the city, constructing a people stream feature sequence of one part according to the population outflow of one part of the two parts in a plurality of historical time periods, and constructing a people stream feature sequence of the other part according to the population outflow of the other part of the two parts in the plurality of historical time periods;
calculating a similarity between the two portions of their respective human stream feature sequences based on minkowski distance;
calculating an adjacency matrix of the spectral clustering according to the corresponding similarity of any two parts of the city; wherein, the weight values of two parts which are not adjacent in the adjacent matrix are 0, and the similarity corresponding to the two parts is in direct proportion to the weight values of the two parts;
and acquiring a Laplacian matrix of the spectral cluster according to the adjacency matrix, standardizing the Laplacian matrix of the spectral cluster, clustering the standardized Laplacian matrix, and acquiring a region division result of the city.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the method for predicting urban travel demand according to any one of claims 1 to 7 are implemented when the processor executes the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910570152.8A CN110322064B (en) | 2019-06-27 | 2019-06-27 | Urban trip demand prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910570152.8A CN110322064B (en) | 2019-06-27 | 2019-06-27 | Urban trip demand prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110322064A CN110322064A (en) | 2019-10-11 |
| CN110322064B true CN110322064B (en) | 2022-04-15 |
Family
ID=68120487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910570152.8A Active CN110322064B (en) | 2019-06-27 | 2019-06-27 | Urban trip demand prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110322064B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110751576B (en) * | 2019-10-21 | 2023-07-25 | 中国民航信息网络股份有限公司 | Passenger journey determining method, device and server |
| CN110866649B (en) * | 2019-11-19 | 2022-09-20 | 中国科学院深圳先进技术研究院 | Method and system for predicting short-term subway passenger flow and electronic equipment |
| CN111062589B (en) * | 2019-12-02 | 2022-08-16 | 武汉理工大学 | Urban taxi scheduling method based on destination prediction |
| CN111160747B (en) * | 2019-12-23 | 2022-07-22 | 北京百度网讯科技有限公司 | Scheduling method and device for unmanned robot taxi and electronic equipment |
| CN111612206B (en) * | 2020-03-30 | 2022-09-02 | 清华大学 | Neighborhood people stream prediction method and system based on space-time diagram convolution neural network |
| CN111815075B (en) * | 2020-08-27 | 2021-01-15 | 北京航空航天大学 | Prediction method for transportation travel demand under major public health incident |
| CN112183824B (en) * | 2020-09-11 | 2022-07-19 | 武汉大学 | Online and offline correlated urban passenger flow prediction method |
| CN112116155B (en) * | 2020-09-18 | 2023-07-25 | 平安科技（深圳）有限公司 | Population flow prediction method and device based on intelligent decision and computer equipment |
| CN112291807B (en) * | 2020-10-15 | 2022-09-23 | 山东科技大学 | Wireless cellular network traffic prediction method based on deep migration learning and cross-domain data fusion |
| CN112633608B (en) * | 2021-01-06 | 2022-04-12 | 南方科技大学 | People flow transfer prediction method, device, equipment and storage medium |
| CN113837383B (en) * | 2021-10-18 | 2023-06-23 | 中国联合网络通信集团有限公司 | Model training method and device, electronic equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109190795A (en) * | 2018-08-01 | 2019-01-11 | 中山大学 | A kind of interregional Travel Demand Forecasting method and device |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10817775B2 (en) * | 2017-01-12 | 2020-10-27 | International Business Machines Corporation | Neural network computing systems for predicting vehicle requests |
| CN108234198B (en) * | 2017-12-19 | 2020-07-07 | 清华大学 | Base station flow prediction method and equipment |
| CN108108844A (en) * | 2017-12-25 | 2018-06-01 | 儒安科技有限公司 | A kind of urban human method for predicting and system |
| CN108615360B (en) * | 2018-05-08 | 2022-02-11 | 东南大学 | Traffic demand day-to-day evolution prediction method based on neural network |
| CN108985475B (en) * | 2018-06-13 | 2021-07-23 | 厦门大学 | Network taxi appointment and taxi calling demand prediction method based on deep neural network |
| CN109165779B (en) * | 2018-08-12 | 2022-04-08 | 北京清华同衡规划设计研究院有限公司 | Population quantity prediction method based on multi-source big data and long-short term memory neural network model |
| CN109376969A (en) * | 2018-12-14 | 2019-02-22 | 中南大学 | City fining population distribution dynamic prediction method and device based on deep learning |
-
2019
- 2019-06-27 CN CN201910570152.8A patent/CN110322064B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109190795A (en) * | 2018-08-01 | 2019-01-11 | 中山大学 | A kind of interregional Travel Demand Forecasting method and device |
Non-Patent Citations (3)
| Title |
|---|
| Gaussian Process Decentralized Data Fusion and Active Sensing for Spatiotemporal Traffic Modeling and Prediction in Mobility-on-Demand Systems;Jie Chen等;《IEEE Transactions on Automation Science and Engineering》;20150506;第12卷(第3期);第901-921页 * |
| Improving Urban Crowd Flow Prediction on Flexible Region Partition;Xu Wang等;《IEEE Transactions on Mobile Computing》;20190820;第19卷(第12期);第2804-2817页 * |
| 基于深度CNN-LSTM-ResNet组合模型的出租车需求预测;段宗涛等;《交通运输系统工程与信息》;20180815(第04期);第215-223页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110322064A (en) | 2019-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110322064B (en) | Urban trip demand prediction method | |
| CN111612206B (en) | Neighborhood people stream prediction method and system based on space-time diagram convolution neural network | |
| Tekouabou et al. | Improving parking availability prediction in smart cities with IoT and ensemble-based model | |
| Chu et al. | Deep multi-scale convolutional LSTM network for travel demand and origin-destination predictions | |
| Han et al. | Analysis of large-scale traffic dynamics in an urban transportation network using non-negative tensor factorization | |
| Faro et al. | Evaluation of the traffic parameters in a metropolitan area by fusing visual perceptions and CNN processing of webcam images | |
| CN115440032B (en) | Long-short-period public traffic flow prediction method | |
| CN112200351A (en) | Urban area passenger flow volume prediction method based on mobile phone signaling data | |
| CN113313947A (en) | Road condition evaluation method of short-term traffic prediction graph convolution network | |
| CN113792929B (en) | Traffic flow prediction method, electronic device and storage medium | |
| CN104657424A (en) | Clustering method for interest point tracks under multiple temporal and spatial characteristic fusion | |
| Gong et al. | Potential passenger flow prediction: A novel study for urban transportation development | |
| EP2590151A1 (en) | A framework for the systematic study of vehicular mobility and the analysis of city dynamics using public web cameras | |
| Chu et al. | Passenger demand prediction with cellular footprints | |
| CN110472999B (en) | Passenger flow mode analysis method and device based on subway and shared bicycle data | |
| CN114692984B (en) | Traffic prediction method based on multi-step coupling graph convolution network | |
| CN113763700B (en) | Information processing method, information processing device, computer equipment and storage medium | |
| CN113283581B (en) | Multi-fusion graph network collaborative multi-channel attention model and application method thereof | |
| CN111695046B (en) | User portrait inference method and device based on spatio-temporal mobile data representation learning | |
| CN113643532B (en) | Regional traffic prediction method and device | |
| Bachir et al. | Combining bayesian inference and clustering for transport mode detection from sparse and noisy geolocation data | |
| CN118262516A (en) | Vehicle traffic prediction method and system based on multiple attention neural networks | |
| Yang et al. | MGSTCN: a multi-graph spatio-temporal convolutional network for metro passenger flow prediction | |
| Chen et al. | A Data-driven Region Generation Framework for Spatiotemporal Transportation Service Management | |
| Faghih et al. | Predicting short-term Uber demand using spatio-temporal modeling: A New York City Case Study |
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 |