CN115392752A - Subway short-time passenger flow prediction method and system, electronic equipment and storage medium - Google Patents
Subway short-time passenger flow prediction method and system, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115392752A CN115392752A CN202211065683.XA CN202211065683A CN115392752A CN 115392752 A CN115392752 A CN 115392752A CN 202211065683 A CN202211065683 A CN 202211065683A CN 115392752 A CN115392752 A CN 115392752A
- Authority
- CN
- China
- Prior art keywords
- social media
- subway
- passenger flow
- gcn
- gru
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000003860 storage Methods 0.000 title claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 7
- 230000000717 retained effect Effects 0.000 claims description 4
- 230000014759 maintenance of location Effects 0.000 claims description 3
- 230000002093 peripheral effect Effects 0.000 abstract description 4
- 230000007547 defect Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 12
- 230000008569 process Effects 0.000 description 10
- 230000002776 aggregation Effects 0.000 description 5
- 238000004220 aggregation Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 238000009826 distribution Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000005065 mining Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
-
- 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
- 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
- G06Q50/265—Personal security, identity or safety
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Educational Administration (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Quality & Reliability (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Security & Cryptography (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention discloses a method and a system for predicting short-time passenger flow of a subway, electronic equipment and a storage medium. The method comprises the following steps: constructing a subway network map based on the inbound traffic of other stations at different time periods; obtaining social media data in the geographic tag range of a target site at different time periods; constructing a GCN-GRU model based on attention; the attention-based GCN-GRU model comprises a plurality of GCN network models, a plurality of full connection layers and a plurality of GRU network models; and inputting the subway network map and the social media data into the GCN-GRU model based on attention, and predicting the outbound flow of the target station in the current time period. The invention fully considers the space structure of the urban rail transit network, predicts the outbound passenger flow of the peripheral sites during the special event by using the passenger flow coming from other sites and the sudden increase of the social media posting volume around the event, and aims to make up the defects of the prior art.
Description
Technical Field
The invention relates to the technical field of passenger flow prediction, in particular to a method and a system for predicting short-time passenger flow of a subway, electronic equipment and a storage medium.
Background
The existing short-term passenger flow prediction generally takes 15-60 min as time granularity, and means that the passenger flow condition of a predicted object after 15min is calculated according to data such as historical passenger flow, real-time passenger flow and the like by using a passenger flow prediction model, if the numerical value exceeds an industrial specification or a safety range given by an operation company, a relevant operation department and staff should immediately carry out actions according to a corresponding safety plan, for example, the passenger flow density of a station and a platform is ensured within a safety range by means of broadcasting, entrance flow limiting and other measures, the personal safety events of passengers such as trampling and the like are prevented, and the safety of passengers and the normal operation of a train are ensured.
There are two main limitations in the field of urban rail transit short-time passenger flow prediction in the event of an incident. First, existing models and data collection and processing approaches perform well in conventional short-term traffic prediction, but they ignore the impact of social media, are unable to identify sudden outbound traffic during special events, and are not well-studied in applying social media data sources in view of the impact of events. Secondly, the research of social media focuses on mining the relationship between text content and passenger flow, but the mining degree of the space-time data is insufficient, and the inbound traffic of other stations is considered to be the cause of sudden increase of the outbound passenger flow of rail transit at the stations around the event, and the social media is considered to be fully considered.
Disclosure of Invention
In order to solve the problems, the invention provides a method and a system for predicting the short-time passenger flow of a subway, electronic equipment and a storage medium.
In order to achieve the purpose, the invention provides the following scheme:
a subway short-time passenger flow prediction method comprises the following steps:
constructing a subway network map based on the inbound traffic of other stations at different time periods;
obtaining social media data in the geographic tag range of a target site at different time periods;
constructing a GCN-GRU model based on attention; the attention-based GCN-GRU model comprises a plurality of GCN network models, a plurality of full connection layers and a plurality of GRU network models;
inputting the subway network map and the social media data into the attention-based GCN-GRU model, and predicting the outbound flow of the target station in the current time period.
Optionally, inputting the subway network map and the social media data into the attention-based GCN-GRU model, and predicting outbound traffic of a target site in a current time period, specifically including:
inputting the subway network map into the GCN network model, and extracting the inbound passenger flow characteristics of other stations at different time periods;
inputting the social media data into the full-connection layer, and extracting social media characteristics at different time intervals;
splicing the inbound passenger flow characteristics of other sites in different time periods and the social media characteristics in different time periods to obtain spliced characteristics;
inputting the spliced features into the GRU network model, and predicting the outbound flow of the target site in the current time period.
Optionally, the subway network map G = (V, E, S, a) V ) Wherein V is a node set, E is an edge set, each node V belongs to V and represents a rail transit station, each edge (u, V) belongs to E and represents a line passing through from the node u to the node V, and S represents a rail transit station adjacency matrix in two directions; a. The V Representing inbound traffic for each site.
Optionally, the obtaining social media data within the geographic tag range of the target site at different time periods specifically includes:
collecting, by a geo-location filter, social media posts having geo-coordinates;
screening out social media posts within the geographic tag range of a target site from the social media posts;
the timestamp and the geographic location of the screened social media posts are retained as social media data.
The invention also provides a system for predicting the short-time passenger flow of the subway, which comprises the following steps:
the subway network map building module is used for building a subway network map based on the inbound traffic of other stations at different time periods;
the social media data acquisition module is used for acquiring social media data in the geographic tag range of the target site at different time periods;
the model building module is used for building a GCN-GRU model based on attention; the attention-based GCN-GRU model comprises a plurality of GCN network models, a plurality of full connectivity layers and a plurality of GRU network models;
and the target station outbound flow prediction module is used for inputting the subway network map and the social media data into the GCN-GRU model based on attention and predicting the outbound flow of the target station in the current time period.
Optionally, the target station outbound traffic prediction module includes:
the inbound passenger flow characteristic extraction unit is used for inputting the subway network map into the GCN network model and extracting inbound passenger flow characteristics of other sites at different time periods;
the social media feature extraction unit is used for inputting the social media data into the full-connection layer and extracting social media features of different time periods;
the splicing unit is used for splicing the inbound passenger flow characteristics of other sites in different time periods and the social media characteristics in different time periods to obtain spliced characteristics;
and the target station outbound flow prediction unit is used for inputting the spliced characteristics into the GRU network model and predicting the outbound flow of the target station in the current time period.
Optionally, the subway network map G = (V, E, S, a) V ) Wherein V is a node set, E is an edge set, each node V belongs to V and represents a rail transit station, each edge (u, V) belongs to E and represents a line passing through from the node u to the node V, and S represents a rail transit station adjacency matrix in two directions; a. The V Representing the inbound traffic for each site.
Optionally, the social media data obtaining module includes:
a collection unit for collecting social media posts with geo-coordinates through a geo-location filter;
the screening unit is used for screening out social media posts within the range of the geographic tags of the target sites from the social media posts;
and the retention unit is used for retaining the time stamp and the geographic position of the screened social media post as social media data.
The invention also provides electronic equipment, which comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor runs the computer program to enable the electronic equipment to execute the subway short-time passenger flow prediction method.
The invention also provides a computer readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the method for predicting the short-time passenger flow of the subway is realized.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the subway short-time passenger flow prediction method provided by the invention fully considers the space structure of the urban rail transit network, predicts the outbound passenger flow of peripheral stations during a special event by utilizing the inbound passenger flow from other stations and the sudden increase of the posting amount of social media around the event, and aims to make up for the defects of the prior art.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a method for predicting short-term passenger flow of a subway provided by the invention;
fig. 2 is a schematic diagram of an updating process of a GRU network model provided in the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. 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 invention provides an attention-based GCN-GRU (Graph Convolutional Networks-Gate RecurrentUnit) model, which is a three-stage model, fully considers the space structure of an urban rail transit network, and predicts the outbound passenger flow of peripheral stations during a special event by using the inbound passenger flow from other stations and the sudden increase of social media posting volume around the event. The innovation points of the invention are as follows: 1) The GCN is used for extracting features in the inbound passenger flow of other sites to contribute to the prediction of the outbound passenger flow of the event site, and the GCN can extract the most effective features more efficiently by using the adjacency relation of each site. 2) The passenger flow characteristics of other stations entering the station and the social media characteristics around the event in each time period are input into the GRU for characteristic extraction and time sequence information transmission so as to predict the passenger flow of the event station leaving the station.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example one
As shown in fig. 1, the method for predicting the short-time passenger flow of the subway provided by the invention comprises the following steps:
step 101: and constructing a subway network map based on the inbound traffic of other stations in different periods.
Step 102: and acquiring social media data within the geographic tag range of the target site at different time periods. The method specifically comprises the following steps: collecting, by a geo-location filter, social media posts having geo-coordinates; screening out social media posts within the geographic tag range of the target site from the social media posts; the timestamp and the geographic location of the screened social media posts are retained as social media data.
Step 103: constructing a GCN-GRU model based on attention; the attention-based GCN-GRU model includes a plurality of GCN network models, a plurality of full connectivity layers, and a plurality of GRU network models. The method specifically comprises the following steps: inputting the subway network map into a GCN model, and extracting inbound passenger flow characteristics of other stations at different time periods; inputting social media data into a full connection layer, and extracting social media characteristics at different time intervals; splicing the inbound passenger flow characteristics of other sites in different time periods and the social media characteristics in different time periods to obtain spliced characteristics; inputting the spliced characteristics into a GRU network model, and predicting the outbound flow of the target site in the current time period.
Step 104: and inputting the subway network map and social media data into the GCN-GRU model based on attention, and predicting the outbound flow of the target station in the current time period.
The specific implementation process of step 101 is as follows:
space-time features, such as historical traffic and OD matrices, have been widely used for traffic flow prediction. However, mining temporal and spatial features alone does not work well due to special prediction challenges during the event. Mining spatiotemporal features of an event is widely used in order to improve performance of passenger flow prediction during the event. The predecessors propose y (t) = f (y (t-1),. Y (t-n))，u i (t-1)，...u i (t-m)) + ε (t), where y (t) represents the outbound traffic for the destination station at time period t. u. of i And (t) represents the i station inbound passenger flow when the time period is t, epsilon (t) is the estimation error when the time period is t, and f is a linear or nonlinear mapping function. However, when i increases, the input to the model is too large, making it difficult to train the model. The spatiotemporal features are defined as a matrix that can handle the increasing i-values and complex traffic networks in large cities. The horizontal axis and the vertical axis respectively represent space-time dimensions (the horizontal axis and the vertical axis are interchangeable), a space-time characteristic diagram is formed, and the road traffic speed and flow prediction method is good in performance. However, there is currently no spatiotemporal feature map suitable for a particular event.
During a special event, the outbound traffic of the destination station will increase significantly, and the outbound traffic during the event may be much higher than normal. This increase is caused by an increase in inbound traffic at other sites. Outbound passenger flow y for a predicted time period tmin t The inbound feature matrix of other stations of the target station in the time period t is defined as:
wherein u i a (t) represents the inbound traffic of i stations in a time period t, p represents the total number of stations in the subway system, and q represents the maximum hysteresis order for prediction.
The inbound traffic values of other sites can only be captured, and the difference between when an event occurs and when no event occurs cannot be reflected. Therefore, the average inbound traffic feature matrix of the target station in the time period t is defined as:
wherein u i b (t) represents the average inbound traffic for site i at time period t, with no events occurring (average of no events occurring over the past month). Finally, the inbound enhancement matrix is defined as:
wherein the symbols
Representing the subtraction of corresponding elements in the two matrices.
Each element in (a) can account for an enhanced value of inbound traffic for each site compared to regular time. If the element value is large, indicating that the station's inbound traffic is significantly increased, the resulting outbound traffic is likely to contribute to the destination's outbound traffic.
The following problems are encountered at this time: the features in the form of the matrix are often extracted by a deep convolutional neural network, however, the convolutional neural network ignores the adjacency relation among all the sites, and the convolutional neural network is a graph network in fact, and the graph convolutional neural network can solve the problem just, so that the convolutional neural network is a deep learning model which is most suitable for processing the problems. The research models the urban rail transit passenger characteristic network into a directed graph, and defines a subway network graph G = (V, E, S), wherein V is a node set, E is an edge set, each node V ∈ V represents an urban rail transit station, each edge (u, V) ∈ E represents a line capable of traversing from u to V, and V and E are obtained according to the actual structure of the urban rail transit network in Beijing.
Is a set of two-directional rail transit station adjacency matrices of graph G. The invention adds the station-entering flow into the finally constructed subway network diagram which is G = (V, E, S, A) V )，A V Attribute representing node, i.e. inbound direction per site time period tQuantity of
Namely that
The specific implementation process of step 102 is as follows:
social media with geotags have been shown to be closely related to the occurrence of outflow events. Social media provides an economical and effective method for obtaining traffic related data, and fills the gap between short-term passenger flow and social media distribution amount.
When an event occurs, the social media posting volume has certain correlation with the subway passenger flow volume. Therefore, the time series of the social media information amount can reflect the changing trend of the passenger flow. To predict yt (targeted site outbound traffic), the social media feature vector is defined as:
s(t-1)＝{H t-1 ，H t-2 ，...H t-m } (4)
wherein H t Represents the total amount of social media posts that the event is within the geotag of the targeted site during time period t, and m represents the maximum hysteresis order for the social media posts to predict. Social media data is obtained from social media (e.g., twitter and twigs) based on geographic coordinates. In particular, social media posts are collected by a geo-location filter having geo-coordinates with the same time window as the event. The spatial bounding box is set to cover only the subway station and the stadium. These posts are all considered related posts because their geo-tags are close to subway stations and stadiums. Finally, the timestamp, geographic location of the valid posts are retained and the social media feature vector is calculated using these data.
The specific implementation process of steps 103-104 is as follows:
the GCN network model is a neural network that operates on graphs and is composed of one or more graph convolutional layers. Graph convolution network GCN is represented by graph G = (V, E, S, A) V ) As inputs, the GCN is calculated as follows: given these inputsAnd when the k-th layer graph convolution neural network calculates and outputs, the following conditions are satisfied:
where σ is the activation function, AGGREGATE:
is a function of the aggregation of the domains,
is a learned weight matrix. Attribute A with the initial node V In a similar manner to that described above,
each row in (a) is a vector representation of a node. In some cases, A V Using matrix multiplication and weight matrix W prior to aggregation k A linear transformation is performed while in other cases the weight multiplication is performed after the aggregation, as shown in equation (6).
The aggregation function of equation (6) derives a new representation of node v by aggregating the neighborhood representations of node v. Although the aggregation function is a distinction between graph-convolved neural network architectures, many architectures can be represented as a weighted sum:
wherein
a (v，n) Is the aggregate weight of neighbors n of node v. The GCN network model is defined as a g (-) function. In the framework, q GCN network models are shared, and q groups of map features (inbound traffic features) are output, namely:
the social media feature vector set P (t-1) is characterized by a plurality of fully connected layers, each of which processes s (t-1), s (t-2).., s (t-q), respectively. There are q full-connected layers in the framework, and q sets of feature vectors (social media features) are output in total, that is: dense (s (t-1)), dense (s (t-2)).. Dense (s (t-q)), where Dense (·) denotes the calculation function of the fully-connected layer.
The inbound passenger flow features extracted by the GCN and the social media features extracted by the full connection layer are input into the subsequent GRU network model after each series concatenation, and the features input into the subsequent GRU network model can be expressed as:
wherein f is t-1 ，f t-2 ，...，f t-q The data are respectively processed by different GRU network models, each GRU network model outputs t-1, t-2, \ 8230, outbound flow in t-n time period, and n is a hysteresis order.
The GRU is a special form of Recurrent Neural Network (RNN), which is an extension of the traditional forward propagation neural network and can be used to solve sequence problems. Specifically, one sequence x = (x) is given 1 ，x 2 ，...x T ) Cyclic hidden state h of RNN t Updated by the following equation:
wherein
Is a non-linear function. Then, the output sequence y = (y) may be calculated 1 ，y 2 ，...y T )。
Generally, the update of the cyclic hidden state in equation (8) is calculated as follows:
h t ＝g(Wx t +Uh t-1 ) (9)
where g represents a smooth and bounded function.
The GRU network model may output a probability distribution as the next element of the sequence, given the current state h t Then, the GRU network model applies a special output symbol to indicate the end of the sequence, and extracts the distribution on the sequence. The model decomposes the sequence probability into:
p(x 1 ，x 2 ，...x T )＝p(x 1 )p(x 2 |x 1 )p(x 3 |x 1 ，x 2 )...p(x T |x 1 ，...，x T-1 ) (10)
each conditional probability distribution is calculated by the following equation:
p(x T |x 1 ，...，x T-1 )＝g(h t ) (11)
the GRU network model may also be used as a rotation unit to adaptively extract the correlation of each element in the sequence. Like the LSTM network model, the GRU network model has several gate units that can regulate the flow of information within a unit without a separate storage unit. Fig. 2 illustrates an update process of the GRU network model. Activation of GRU network model at time t
Is a previous activation
And candidate activation
Linear interpolation between:
wherein the refresh door
The determination unit updates the degree of activation. Updating the gate may be handled by:
candidate activation
The calculation process of (2) is similar to that of a conventional RNN:
wherein r is t A reset gate for a specified group indicates that the corresponding element is multiplied. When the temperature is higher than the set temperature
Near 0, the reset gate effectively causes the cell to behave as if it were reading the first symbol of the input sequence, thereby causing it to forget the state previously calculated. Reset door
The calculation process of (a) is similar to that of the refresh gate:
note that the mechanism is one between GRU and f t The process between, in which a full join is used, can extract the most useful features from the input features. The attention weight is learned through the fully connected layer and SoftMax functions, and the input is multiplied by the attention weight. That is, the attention score is calculated as SoftMax (dense (f) t ))，f t ×SoftMax(dense(f t ) Is input into the GRU. The entire attention-based GRU process is defined as the function AGRU (·):
the overall model is to solve an optimization problem. The decision variables are parameters of the whole framework, and the objective function is the Mean Square Error (MSE) of the prediction result, as follows:
wherein y is t Is the true value of the outbound traffic of the target station,
is a prediction of the outbound traffic of the target station. Theta represents the parameters of the entire framework and the invention learns by back propagation using an Adam optimizer.
The invention fully considers the space structure of the urban rail transit network, predicts the outbound passenger flow of the peripheral sites during the special event by using the passenger flow coming from other sites and the sudden increase of the social media posting volume around the event, and aims to make up the defects of the prior art.
Example two
In order to implement a corresponding method of the above embodiment to achieve corresponding functions and technical effects, the following provides a system for predicting short-term passenger flow in a subway, including:
the subway network map building module is used for building a subway network map based on the inbound traffic of other stations at different time periods;
the social media data acquisition module is used for acquiring social media data in the geographic tag range of the target site at different time periods;
the model building module is used for building a GCN-GRU model based on attention; the attention-based GCN-GRU model comprises a plurality of GCN network models, a plurality of full connection layers and a plurality of GRU network models;
and the target station outbound flow prediction module is used for inputting the subway network map and the social media data into the GCN-GRU model based on attention and predicting the outbound flow of the target station in the current time period.
The target station outbound flow prediction module comprises:
the inbound passenger flow characteristic extraction unit is used for inputting the subway network map into the GCN network model and extracting inbound passenger flow characteristics of other sites at different time periods;
the social media feature extraction unit is used for inputting social media data into the full connection layer and extracting social media features of different time periods;
the splicing unit is used for splicing the inbound passenger flow characteristics of other sites in different time periods and the social media characteristics in different time periods to obtain spliced characteristics;
and the target station outbound flow prediction unit is used for inputting the spliced characteristics into the GRU network model and predicting the outbound flow of the target station in the current time period.
Wherein, social media data acquisition module includes:
a collection unit for collecting social media posts with geo-coordinates through a geo-location filter;
the screening unit is used for screening out social media posts within the range of the geographic tags of the target sites from the social media posts;
and the retention unit is used for retaining the time stamp and the geographical position of the screened social media post as social media data.
EXAMPLE III
The third embodiment of the invention provides a memory and a processor, wherein the memory is used for storing a computer program, and the processor runs the computer program to enable electronic equipment to execute the method for predicting the short-time passenger flow of the subway in the first embodiment.
The electronic device may be a server.
Example four
The fourth embodiment of the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the method for predicting short-time passenger flow in a subway according to the first embodiment is implemented.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the description of the method part.
The principle and the embodiment of the present invention are explained by applying specific examples, and the above description of the embodiments is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In summary, this summary should not be construed to limit the present invention.
Claims (10)
1. A subway short-time passenger flow prediction method is characterized by comprising the following steps:
constructing a subway network map based on the inbound traffic of other stations in different periods;
obtaining social media data in the geographic tag range of a target site at different time periods;
constructing a GCN-GRU model based on attention; the attention-based GCN-GRU model comprises a plurality of GCN network models, a plurality of full connectivity layers and a plurality of GRU network models;
inputting the subway network map and the social media data into the attention-based GCN-GRU model, and predicting the outbound flow of the target station in the current time period.
2. A method as claimed in claim 1, wherein the step of inputting the subway network map and the social media data into the attention-based GCN-GRU model to predict outbound traffic of a target site in a current time period comprises:
inputting the subway network map into the GCN network model, and extracting the inbound passenger flow characteristics of other stations at different time periods;
inputting the social media data into the full-connection layer, and extracting social media characteristics at different time intervals;
splicing the inbound passenger flow characteristics of other sites in different time periods and the social media characteristics in different time periods to obtain spliced characteristics;
inputting the spliced characteristics into the GRU network model, and predicting the outbound flow of the target station in the current time period.
3. A subway short-time passenger flow prediction method as claimed in claim 1, wherein said subway network map G = (V, E, S, a) V ) Wherein V is a node set, E is an edge set, each node V belongs to V and represents a rail transit station, each edge (u, V) belongs to E and represents a line passing through from the node u to the node V, and S represents a rail transit station adjacency matrix in two directions; a. The V Representing inbound traffic for each site.
4. A subway short-time passenger flow prediction method as claimed in claim 1, wherein obtaining social media data within the geographic tag range of target sites at different time periods specifically comprises:
collecting, by a geo-location filter, social media posts having geo-coordinates;
screening out social media posts within the geographic tag range of the target site from the social media posts;
the timestamp and the geographic location of the screened social media posts are retained as social media data.
5. A subway short-time passenger flow prediction system is characterized by comprising:
the subway network map building module is used for building a subway network map based on the inbound traffic of other stations in different periods;
the social media data acquisition module is used for acquiring social media data in the geographic tag range of the target site at different time periods;
the model building module is used for building a GCN-GRU model based on attention; the attention-based GCN-GRU model comprises a plurality of GCN network models, a plurality of full connection layers and a plurality of GRU network models;
and the target station outbound flow prediction module is used for inputting the subway network map and the social media data into the GCN-GRU model based on attention and predicting the outbound flow of the target station in the current time period.
6. The subway short-time passenger flow prediction system as claimed in claim 5, wherein said target station outbound traffic prediction module comprises:
the inbound passenger flow characteristic extraction unit is used for inputting the subway network map into the GCN network model and extracting inbound passenger flow characteristics of other sites at different time periods;
the social media feature extraction unit is used for inputting the social media data into the full-connection layer and extracting social media features of different time periods;
the splicing unit is used for splicing the inbound passenger flow characteristics of other sites in different time periods and the social media characteristics in different time periods to obtain spliced characteristics;
and the target station outbound flow prediction unit is used for inputting the spliced characteristics into the GRU network model and predicting the outbound flow of the target station in the current time period.
7. A subway short-time passenger flow prediction system as claimed in claim 5, wherein said subway network map G = (V, E, S, A) V ) Wherein V is a node set, E is an edge set, each node V belongs to V and represents a rail transit station, each edge (u, V) belongs to E and represents a line passing through from the node u to the node V, and S represents a rail transit station adjacency matrix in two directions; a. The V Representing inbound traffic for each site.
8. The subway short-time passenger flow prediction system as claimed in claim 5, wherein said social media data acquisition module comprises:
a collection unit to collect social media posts with geo-coordinates through a geo-location filter;
the screening unit is used for screening out social media posts in the range of the geographic tags of the target sites from the social media posts;
and the retention unit is used for retaining the time stamp and the geographic position of the screened social media post as social media data.
9. An electronic device, comprising a memory for storing a computer program and a processor for executing the computer program to cause the electronic device to perform the method of predicting short-term passenger flow in a subway according to any one of claims 1 to 4.
10. A computer-readable storage medium, characterized in that it stores a computer program which, when being executed by a processor, implements the method of predicting short-time passenger flow in a subway according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211065683.XA CN115392752B (en) | 2022-09-01 | 2022-09-01 | Subway short-time passenger flow prediction method, system, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211065683.XA CN115392752B (en) | 2022-09-01 | 2022-09-01 | Subway short-time passenger flow prediction method, system, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115392752A true CN115392752A (en) | 2022-11-25 |
| CN115392752B CN115392752B (en) | 2023-11-28 |
Family
ID=84124085
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211065683.XA Active CN115392752B (en) | 2022-09-01 | 2022-09-01 | Subway short-time passenger flow prediction method, system, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115392752B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116128122A (en) * | 2023-01-03 | 2023-05-16 | 北京交通大学 | Urban rail transit short-time passenger flow prediction method considering burst factors |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110866649A (en) * | 2019-11-19 | 2020-03-06 | 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 |
| CN111860787A (en) * | 2020-07-27 | 2020-10-30 | 南京工程学院 | Short-term prediction method and device for coupling directed graph structure flow data containing missing data |
| CN112435462A (en) * | 2020-10-16 | 2021-03-02 | 同盾控股有限公司 | Method, system, electronic device and storage medium for short-time traffic flow prediction |
| CN112801355A (en) * | 2021-01-20 | 2021-05-14 | 南京航空航天大学 | Data prediction method based on multi-graph fusion space-time attention of long-short-term space-time data |
| CN113077090A (en) * | 2021-04-09 | 2021-07-06 | 上海大学 | Passenger flow prediction method, system and computer readable storage medium |
| CN113487856A (en) * | 2021-06-04 | 2021-10-08 | 兰州理工大学 | Traffic flow combination prediction model based on graph convolution network and attention mechanism |
| US11238729B1 (en) * | 2020-09-11 | 2022-02-01 | Toyota Motor Engineering & Manufacturing North America, Inc. | Systems and methods for traffic flow prediction |
-
2022
- 2022-09-01 CN CN202211065683.XA patent/CN115392752B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110866649A (en) * | 2019-11-19 | 2020-03-06 | 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 |
| CN111860787A (en) * | 2020-07-27 | 2020-10-30 | 南京工程学院 | Short-term prediction method and device for coupling directed graph structure flow data containing missing data |
| US11238729B1 (en) * | 2020-09-11 | 2022-02-01 | Toyota Motor Engineering & Manufacturing North America, Inc. | Systems and methods for traffic flow prediction |
| CN112435462A (en) * | 2020-10-16 | 2021-03-02 | 同盾控股有限公司 | Method, system, electronic device and storage medium for short-time traffic flow prediction |
| CN112801355A (en) * | 2021-01-20 | 2021-05-14 | 南京航空航天大学 | Data prediction method based on multi-graph fusion space-time attention of long-short-term space-time data |
| CN113077090A (en) * | 2021-04-09 | 2021-07-06 | 上海大学 | Passenger flow prediction method, system and computer readable storage medium |
| CN113487856A (en) * | 2021-06-04 | 2021-10-08 | 兰州理工大学 | Traffic flow combination prediction model based on graph convolution network and attention mechanism |
Non-Patent Citations (2)
| Title |
|---|
| 荣斌;武志昊;刘晓辉;赵苡积;林友芳;景一真;: "基于时空多图卷积网络的交通站点流量预测", 计算机工程, no. 05 * |
| 闫杨;孙丽?;朱兰婷;: "基于时空相关性的短时交通流量预测方法", 计算机工程, no. 01 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116128122A (en) * | 2023-01-03 | 2023-05-16 | 北京交通大学 | Urban rail transit short-time passenger flow prediction method considering burst factors |
| CN116128122B (en) * | 2023-01-03 | 2023-09-12 | 北京交通大学 | Urban rail transit short-time passenger flow prediction method considering burst factors |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115392752B (en) | 2023-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111223301B (en) | Traffic flow prediction method based on graph attention convolution network | |
| Chen et al. | A graph convolutional stacked bidirectional unidirectional-LSTM neural network for metro ridership prediction | |
| CN111242292B (en) | OD data prediction method and system based on deep space-time network | |
| CN113673769B (en) | Traffic flow prediction method of graph neural network based on multivariate time sequence interpolation | |
| Gong et al. | Potential passenger flow prediction: A novel study for urban transportation development | |
| CN111242395B (en) | Method and device for constructing prediction model for OD (origin-destination) data | |
| CN112801355A (en) | Data prediction method based on multi-graph fusion space-time attention of long-short-term space-time data | |
| CN114519932B (en) | Regional traffic condition integrated prediction method based on space-time relation extraction | |
| CN115204478A (en) | Public traffic flow prediction method combining urban interest points and space-time causal relationship | |
| CN113673749B (en) | Vehicle-pedestrian map track prediction method and device | |
| CN116895157A (en) | Traffic flow prediction method, system and medium based on trend time-space diagram convolution | |
| CN113496310A (en) | Atmospheric pollutant prediction method and system based on deep learning model | |
| CN115392752A (en) | Subway short-time passenger flow prediction method and system, electronic equipment and storage medium | |
| CN114021809A (en) | Unmanned driving is with picture level convolution neural network model of traffic flow prediction | |
| Yu et al. | A short-term traffic flow prediction method based on spatial–temporal correlation using edge computing | |
| Fang et al. | Traffic speed prediction based on LSTM-graph attention network (L-GAT) | |
| CN115762147A (en) | Traffic flow prediction method based on adaptive graph attention neural network | |
| Yang et al. | Short‐Term Forecasting of Dockless Bike‐Sharing Demand with the Built Environment and Weather | |
| Xiong et al. | Traffic flow prediction based on extended multi-component graph convolutional network | |
| CN115565370A (en) | Local space-time graph convolution traffic flow prediction method and system | |
| CN111327453B (en) | Communication bandwidth estimation method considering gridding dynamic and static components | |
| CN115456238A (en) | Urban trip demand prediction method based on dynamic multi-view coupling graph convolution | |
| CN114548572A (en) | Method, device, equipment and medium for predicting urban road network traffic state | |
| Song et al. | Sparse trip demand prediction for shared E-scooter using spatio-temporal graph neural networks | |
| CN111242394B (en) | Method and system for extracting spatial correlation characteristics |
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 |