CN114372830A - Network taxi booking demand prediction method based on space-time multi-graph neural network - Google Patents

Network taxi booking demand prediction method based on space-time multi-graph neural network Download PDF

Info

Publication number
CN114372830A
CN114372830A CN202210038032.5A CN202210038032A CN114372830A CN 114372830 A CN114372830 A CN 114372830A CN 202210038032 A CN202210038032 A CN 202210038032A CN 114372830 A CN114372830 A CN 114372830A
Authority
CN
China
Prior art keywords
graph
network
road section
correlation
time
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.)
Pending
Application number
CN202210038032.5A
Other languages
Chinese (zh)
Inventor
陈静
袁长伟
毛新华
王虎军
官文英
丁圣轩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changan University
Original Assignee
Changan University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Changan University filed Critical Changan University
Priority to CN202210038032.5A priority Critical patent/CN114372830A/en
Publication of CN114372830A publication Critical patent/CN114372830A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0202—Market predictions or forecasting for commercial activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G06Q50/40—

Abstract

The invention discloses a network taxi appointment demand prediction method based on a space-time multi-graph neural network, which comprises the following steps of: acquiring network car booking demand data, constructing an adjacency relation graph, a road section function correlation graph and a road section public traffic correlation graph according to urban road section space relation, road section functions and public traffic correlation, inputting the constructed graph data into an end-to-end space-time graph convolutional neural network, firstly capturing multi-graph correlation by using the graph convolutional network, then performing multi-graph fusion, secondly inputting the graph fused at each moment into a gated cyclic neural network to capture time correlation among graph data, constructing the space-time graph convolutional neural network, and inputting order time and the matched road section into the space-time graph convolutional neural network to obtain a predicted value of the network car booking demand on the road section. The method starts from the perspective of time-space distribution of network car booking requirements, multiple graphs are constructed, and the network car booking requirements of multiple time slices of all road sections in the future are predicted by utilizing an end-to-end time-space graph convolutional neural network.

Description

Network taxi booking demand prediction method based on space-time multi-graph neural network
Technical Field
The invention relates to the technical field of traffic prediction, in particular to a network taxi appointment demand prediction method based on a space-time multi-graph neural network.
Background
In the era of diversified traffic development, taking the net appointment vehicle for traveling becomes one of important traveling modes due to the characteristics of flexibility and convenience of the net appointment vehicle. However, many cities have the problem of 'difficulty in taxi calling', and the main reason is that a supply and demand contradiction exists between the network taxi booking service provided by a driver and taxi taking requirements generated by passengers, so that the operation cost of the network taxi booking is increased, the utilization rate of the network taxi booking and the income of the driver are reduced, and the waiting time of the passengers is also increased. The accurate prediction network car booking requirement can help the network car booking platform to distribute orders and dispatch vehicles more reasonably, and meanwhile, the accurate prediction network car booking platform has a very positive effect on traffic management, design, travel guidance and the like.
When the prior art predicts the demand of network taxi booking, the city is divided into grids with fixed size or grid areas with a certain rule, and the problems that the grid is not consistent with the actual road network condition, the same road section can be divided into a plurality of grids, different road sections in the same grid are far from the actual reachable path and the like exist; in addition, when the space-time diagram network modeling is carried out on non-European space to predict the network appointment demand, the multiple spatial correlation existing between the road sections is not fully considered.
Disclosure of Invention
In order to avoid the defects of the prior art, the invention provides a network taxi appointment demand prediction method based on a space-time multi-graph neural network, which is used for capturing non-European-space multiple correlation based on urban actual road network data by utilizing a graph convolution network to capture adjacency relation, road function correlation and public traffic correlation among roads, and then capturing time correlation of the graph network by utilizing a gated cyclic neural network to solve the problem of road section non-European-space multiple correlation modeling.
The embodiment of the invention provides a network appointment demand prediction method based on a space-time multi-graph neural network, which comprises the following steps:
extracting urban road network data and performing map matching;
constructing a road section adjacency relation graph according to whether the road section spaces are adjacent or not;
constructing a road section function correlation graph according to the road section function;
obtaining a public transport correlation diagram according to the correlation of the subway and bus stop at the road section;
respectively carrying out graph convolution operation on the road section adjacency relation graph, the road section function correlation graph and the public traffic correlation graph by utilizing a graph convolution neural network, and capturing multiple correlations including adjacency relation, function correlation and public traffic correlation among roads;
inputting the fused road section adjacency graph, road section function correlation graph and public traffic correlation graph into a recurrent neural network, and capturing time correlation at different moments;
constructing a space-time graph convolutional neural network based on the multiple spatial correlation of the graph convolutional neural network and the time correlation of the cyclic neural network;
and inputting the order time and the matched road section into a space-time graph convolutional neural network to obtain a predicted value of the network appointment vehicle demand in the future period of time on the road section.
Further, the extracting the urban road network data and performing map matching comprises:
cleaning the data, and deleting order data of missing positions and time;
extracting urban road network data through an open source map service OpenStreetMap to obtain road section information;
matching the order to the nearest road section by using a map matching algorithm according to the longitude and latitude information in the order data;
and designing the length of the time slice, and counting the quantity of the network appointment demands of all road sections of all the time slices according to the order time and the matched road sections to be used as the input of the network.
Further, constructing a road section adjacency graph, comprising:
and constructing an adjacency relation graph according to the extracted link adjacency relation in the road network, representing vertexes by using the links, representing edges by using the relation between the links, and representing whether the two links are adjacent or not by using the edge weight between the two vertexes.
Further, constructing a road section function correlation diagram, comprising:
the road section function correlation is obtained by calculating the similarity of POI distribution on the road section, a graph structure is represented by a vertex, an edge and an edge weight matrix, and the edge weight represents the function correlation of the two road sections.
Further, constructing a road section public transportation correlation diagram comprises the following steps:
the public traffic correlation is obtained by calculating the distribution similarity of subways and bus stops on road sections, a graph structure is represented by using vertex, edge and edge weight matrixes, and the edge weight matrix represents the public traffic correlation of the two road sections.
Further, still include: training a space-time graph convolutional network, which specifically comprises the following steps:
inputting the statistical network taxi booking demand matrix into a network, setting a time slice step length, and dividing the demand matrix according to the step length;
carrying out graph convolution on the established road section adjacency graph, the road section function correlation graph and the road section public traffic correlation graph to capture multiple correlations among the road sections, respectively carrying out graph convolution operation on the three graphs at each moment, and then carrying out multi-graph fusion;
capturing the time correlation among the fused images by using a gated cyclic neural network, inputting the fused images into the cyclic neural network at each moment, and capturing the correlation existing at different moments;
outputting the required quantity of the network appointment vehicles of all road sections in a plurality of time slices in the future, reversely adjusting parameters according to the loss between the prediction result and the true value, completing network training through iterative training, and storing the training network parameters.
Further, still include:
inputting test data into the trained network, calculating the quantity of network car booking demands on the road section in future specific time according to the trained network and parameters thereof, and outputting a predicted value;
and calculating the direct error of the true value and the predicted value according to the evaluation index, and evaluating the network performance.
The embodiment of the invention provides a network appointment demand prediction method based on a space-time multi-graph neural network, which has the following beneficial effects compared with the prior art:
1. the invention provides a method for predicting the demand of vehicle booking on a road section, which solves the problems that the road section is inconsistent with the actual road network condition, the same road section can be divided into a plurality of grids, the actual reachable paths of different road sections in the same grid are far and even unreachable and the like in the demand prediction of the current method by using a region division method, is favorable for better dispatching orders and path planning for vehicle booking on the network, and effectively relieves the contradiction between supply and demand between the vehicle booking and passengers. The method has positive effects on better traffic management and planning of traffic management departments.
2. The invention provides a network car booking requirement prediction method based on a space-time multi-graph convolutional neural network, which fully considers various factors influencing the network car booking requirement, captures the multi-time and space correlation influencing the network car booking requirement, and effectively improves the accuracy of network prediction.
Drawings
FIG. 1 is a step diagram of a network taxi appointment demand prediction method based on a spatio-temporal multi-graph neural network according to an embodiment of the present invention;
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying 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.
Referring to fig. 1, an embodiment of the present invention provides a method for predicting a network taxi appointment demand based on a spatiotemporal multi-graph neural network, where the method includes:
step 1, data preprocessing is carried out, and data of missing positions and time information are deleted.
Step 2, urban road network extraction and map matching, comprising the following steps:
and step 21, acquiring city IDs and city IDs according to city names based on interfaces provided by the OpenStreetMap map service, extracting city road network data according to the interfaces by using a Python program, and acquiring and storing road section information.
And step 22, matching the order to the nearest road section in the urban road network by using a map matching algorithm according to the longitude and latitude information in each order data.
And 23, dividing the time into slices with the same time length, and counting the quantity of the network car booking demands of all the time slices of all the road sections according to the order time and the quantity of the road sections matched with the orders in the step 22 to obtain a network car booking demand matrix.
And 3, taking the statistical network car booking requirement matrix as input, wherein the size of the matrix is NxW, N represents the number of all the sections, and W represents the total time slice number of the historical network car booking requirement. m isij∈M,0<i≤W,0<And j is less than or equal to N, which represents the net appointment demand quantity of the jth road section at the ith moment. Setting step length d for dividing N, dividing input demand matrix into
Figure BDA0003469183890000051
A submatrix, which takes the time slices of d steps as a time and has the same time
Figure BDA0003469183890000052
And (4) the time.
And 4, constructing a road section spatial correlation multi-graph which comprises a road section adjacency relation graph, a road section function correlation graph and a road section public transport correlation graph. The method comprises the following steps:
step 41, constructing an adjacency graph G according to whether road sections are adjacent to each other in road network geographic spacerIs shown as Gr＝(V,E,Wr) V denotes a set of vertices in the graph, one vertex denotes a link, E denotes a set of edges, Wr(i, j) indicates whether or not the link i is adjacent to the link j, 1 indicates adjacent, and 0 indicates non-adjacent.
Step 42, road segment functional correlation graph Gf: g for road segment function correlation graphf＝(V,E,Wf) Is represented by the formula, wherein WfAnd (i, j) representing POI distribution on the road sections into vectors for the functional relevance of the two road sections, and calculating the POI vector similarity of the two road sections to obtain the functional relevance of the road sections.
Step 43, road section public transport correlation diagram Gp: g for road section public traffic correlation graphp＝ (V,E,Wp) And (4) showing. WpAnd (i, j) representing the public traffic correlation of the two road sections, representing the distribution of the subway and the bus stop on the road section into vectors, and calculating the similarity of the stop vectors of the two road sections to obtain the public traffic correlation of the road section.
And 5, capturing multiple correlations existing among the road sections by using the graph convolution neural network, and further fusing the graphs after graph convolution. The method comprises the following steps:
step 51, capturing road section adjacency correlation, road section function correlation and road section public traffic correlation by utilizing a two-layer graph convolution neural network, wherein the graph convolution process is expressed as
Figure RE-GDA0003535117520000053
Wherein
Figure RE-GDA0003535117520000054
A is an adjacency matrix, I is an identity matrix,
Figure RE-GDA0003535117520000055
then is
Figure RE-GDA0003535117520000056
Degree matrix of (W)(l)Is disciplinableA refined weight matrix. σ (-) is the activation function. H(l)And H(l+1)The input and the output of the l-th layer neural network are respectively.
And step 52, carrying out graph convolution on the road section adjacent correlation graph, the road section function correlation graph and the road section public transport correlation graph, and fusing the graphs.
And 6, capturing the time correlation between the fused images by using a gated Recurrent neural network GRU (gate Recurrent Unit). And at each moment, inputting the fused graph into a gated recurrent neural network to capture the correlation existing at different moments.
And 7, repeating the step 5 and the step 6, performing iterative training for K times, finishing the training of the space-time diagram convolutional neural network, and storing parameters.
And 8, inputting test sample data according to the trained network and the stored parameters, calculating a predicted value through the parameters, and evaluating the performance of the space-time graph convolutional neural network according to the error between the true value and the predicted value.
Example 1:
step 1, inputting network car booking order data, and deleting data of missing position and time information in the order.
Step 2, urban road network extraction and map matching, and specifically executing the following steps:
and step 21, acquiring a city ID according to the city name based on the OpenStreetMap map service, extracting city road network data according to the city ID by using a Python program according to an interface, and acquiring and storing road section information.
And step 22, matching the starting point of the order to the nearest road section in the urban road network by using a map matching algorithm based on hidden Markov according to the starting point longitude lon and the latitude lat in each piece of order data.
Step 23, setting the unit time slice as e minutes, wherein T represents the total minutes covered by all orders, and the T represents the total minutes covered by all orders
Figure BDA0003469183890000061
Time slicing, and according to the road network data, if a total number of N road sections exist, the network taxi appointment demand matrix is expressed as M belongs to RW×N。
Step 3, inputting the network taxi appointment demand matrix M into the empty graph convolution network, wherein the matrix size is W multiplied by N, Mij∈M,0<i≤W,0<N is more than or equal to j, the net appointment demand quantity of the jth road section at the ith moment is represented, step length d is set for dividing N, and the input demand matrix is divided into
Figure BDA0003469183890000062
A submatrix, which takes the time slices of d steps as a time and has a total
Figure BDA0003469183890000063
And (4) the time.
And 4, constructing a road section spatial correlation multi-graph which comprises a road section adjacency relation graph, a road section function correlation graph and a road section public transport correlation graph. The method comprises the following steps:
step 41, building a road section adjacency graph Gr. Constructing an adjacency graph G according to whether road sections are adjacent in road network ground spacerIs shown as Gr＝(V,E,Wr) V denotes a set of vertices in the graph, one vertex denotes a road segment, E denotes a set of edges, Wr(i, j) represents whether or not the link i is adjacent to the link j, 1 represents adjacent, 0 represents not adjacent, WrIs shown as
Figure BDA0003469183890000071
Step 42, building a road section function correlation graph GfIs shown as Gf＝(V,E,Wf) Wherein W isf(i, j) is the functional relevance of two road segments. The method comprises the steps of crawling interest Point (POI) distribution on a road section, wherein the POI distribution comprises interest points of catering, work (such as companies, banks and the like), business (such as shopping malls, shops and the like), houses, science and education, medical treatment and the like to form a feature vector. By calculating POI vectors pv of two road segmentsiAnd pvjThe cosine similarity yields a road segment functional correlation, expressed as:
Figure BDA0003469183890000072
Wfis expressed as the following formula, wherein WfThe value of (i, j) is normalized Sim (pv)i,pvj) The value is obtained.
Figure BDA0003469183890000073
Step 43, constructing a road section public transport correlation diagram Gp: g for road section public traffic correlation graphp＝ (V,E,Wp) And (4) showing. Wp(i, j) represents the public traffic correlation of two road sections, the distribution of subway and bus stop on the road sections is represented as a vector, and the stop vector tv of the two road sections is calculatediAnd tvjCosine similarity Sim (tv)i,tvj) And obtaining the public traffic correlation of the road sections. WpIs expressed as the following formula, wherein WpThe value of (i, j) is the normalized Sim (tv)i,tvj) The value is obtained.
Figure BDA0003469183890000074
And 5, capturing the road section adjacency correlation, the road section function correlation and the road section public traffic correlation through a graph convolution network, and further fusing the graph after graph convolution. The method comprises the following steps:
step 51, convolution with a two-layer graph, denoted as
Figure BDA0003469183890000075
Wherein
Figure BDA0003469183890000076
Figure BDA0003469183890000077
A is an adjacency matrix, I is an identity matrix,
Figure BDA0003469183890000078
then is
Figure BDA0003469183890000079
Degree matrix of (W)(l)Is a trainable weight matrix. σ (-) is the activation function. H(l)And H(l+1)The input and the output of the l-th layer neural network are respectively.
Step 52, fusing the convolved images,
Figure BDA0003469183890000081
is the fused output matrix, expressed as:
Figure BDA0003469183890000082
and 6, capturing the time correlation between the fused images by utilizing the GRU network. At each time t, a fused matrix is generated
Figure BDA0003469183890000083
Will be provided with
Figure BDA0003469183890000084
Inputting the input into a GRU network for training.
And 7, repeating the step 5 and the step 6, performing iterative training for K times, finishing the training of the space-time diagram convolutional neural network, and storing parameters.
And 8, inputting test sample data according to the trained network and the stored parameters, calculating a predicted value through the parameters, and evaluating the performance of the space-time graph convolutional neural network.
Although the present invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.

Claims (7)

1. A network taxi appointment demand prediction method based on a spatiotemporal multi-graph neural network is characterized by comprising the following steps:
extracting urban road network data and performing map matching;
constructing a road section adjacency relation graph according to whether the road section spaces are adjacent or not;
constructing a road section function correlation graph according to the road section function;
obtaining a public transport correlation diagram according to the correlation of the subway and bus stop at the road section;
respectively carrying out graph convolution operation on the road section adjacency relation graph, the road section function correlation graph and the public traffic correlation graph by utilizing a graph convolution neural network, and capturing multiple correlations including adjacency relation, function correlation and public traffic correlation among roads;
inputting the fused road section adjacency graph, road section function correlation graph and public traffic correlation graph into a recurrent neural network, and capturing time correlation at different moments;
constructing a space-time graph convolutional neural network based on the multiple spatial correlation of the graph convolutional neural network and the time correlation of the cyclic neural network;
and inputting the order time and the matched road section into a space-time graph convolutional neural network to obtain a predicted value of the network car-booking demand of the road section in a future period of time.
2. The method for predicting vehicle appointment demand based on the spatio-temporal multi-graph neural network as claimed in claim 1, wherein the extracting city road network data and performing map matching comprises:
cleaning the data, and deleting order data of missing positions and time;
extracting urban road network data through an open source map service OpenStreetMap to obtain road section information;
matching the order to the nearest road section by using a map matching algorithm according to the longitude and latitude information in the order data;
and designing the length of the time slice, and counting the required quantity of the network appointment of each road section of each time slice according to the order time and the matched road section to be used as the input of the network.
3. The method for forecasting network appointment demand based on the spatio-temporal multi-graph neural network as claimed in claim 1, wherein the step of constructing the road section adjacency graph comprises the following steps:
and constructing an adjacency relation graph according to the extracted link adjacency relation in the road network, representing vertexes by using the links, representing edges by using the relation between the links, and representing whether the two links are adjacent or not by using the edge weight between the two vertexes.
4. The method for forecasting network appointment demand based on the spatiotemporal multi-graph neural network as claimed in claim 1, wherein the constructing of the road section function correlation graph comprises:
and obtaining the functional relevance of the road sections according to the similarity of POI distribution on the road sections, representing the graph structure by using a vertex, an edge and an edge weight matrix, and representing the functional relevance of the two road sections by using the edge weight.
5. The method for predicting the vehicle appointment demand based on the spatiotemporal multi-graph neural network as claimed in claim 1, wherein the step of constructing the road section public transportation correlation graph comprises the following steps:
public transportation relativity is obtained by calculating distribution similarity of subways and bus stops on road sections, and a graph structure is represented by using a vertex, an edge and an edge weight matrix, wherein the edge weight represents the public transportation relativity of the two road sections.
6. The method for forecasting net appointment of vehicle demand based on spatiotemporal multi-graph neural network as claimed in claim 1, training spatiotemporal graph convolutional network, comprising:
inputting the statistical network car booking demand matrix into a network, setting a time slice step length, and dividing the demand matrix according to the step length;
performing multi-graph convolution on the established road section adjacency graph, the road section function correlation graph and the road section public traffic correlation graph to acquire multiple correlations among road sections, performing graph convolution operation on the three graphs at each moment, and then performing multi-graph fusion;
capturing the time correlation among the fused images by using a gated cyclic neural network, inputting the fused images into the cyclic neural network at each moment, and capturing the correlation existing at different moments;
outputting the required quantity of the network appointment vehicles of all road sections in a plurality of time slices in the future, reversely adjusting parameters according to the loss between the prediction result and the true value, completing network training through iterative training, and storing the training network parameters.
7. The method of claim 6, further comprising:
inputting the test data into the trained network, calculating the quantity of network car booking demands in future specific time on the road section according to the trained network and the parameters thereof, and outputting a predicted value;
and calculating the direct error of the true value and the predicted value according to the evaluation index, and evaluating the network performance.
CN202210038032.5A 2022-01-13 2022-01-13 Network taxi booking demand prediction method based on space-time multi-graph neural network Pending CN114372830A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210038032.5A CN114372830A (en) 2022-01-13 2022-01-13 Network taxi booking demand prediction method based on space-time multi-graph neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210038032.5A CN114372830A (en) 2022-01-13 2022-01-13 Network taxi booking demand prediction method based on space-time multi-graph neural network

Publications (1)

Publication Number Publication Date
CN114372830A true CN114372830A (en) 2022-04-19

Family

ID=81144790

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210038032.5A Pending CN114372830A (en) 2022-01-13 2022-01-13 Network taxi booking demand prediction method based on space-time multi-graph neural network

Country Status (1)

Country Link
CN (1) CN114372830A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115618986A (en) * 2022-09-29 2023-01-17 北京骑胜科技有限公司 Method and device for coordinating resources
CN116187611A (en) * 2023-04-25 2023-05-30 南方科技大学 Multi-agent path planning method and terminal

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115618986A (en) * 2022-09-29 2023-01-17 北京骑胜科技有限公司 Method and device for coordinating resources
CN116187611A (en) * 2023-04-25 2023-05-30 南方科技大学 Multi-agent path planning method and terminal

Similar Documents

Publication Publication Date Title
Huang et al. A two-phase optimization model for the demand-responsive customized bus network design
Chu et al. Deep multi-scale convolutional LSTM network for travel demand and origin-destination predictions
US10639995B2 (en) Methods, circuits, devices, systems and associated computer executable code for driver decision support
Swamy et al. Hurricane evacuation planning using public transportation
Gao et al. Optimize taxi driving strategies based on reinforcement learning
US11094028B2 (en) System and method for determining passenger-seeking ride-sourcing vehicle navigation
CN112074845A (en) Deep reinforcement learning for optimizing car pooling strategies
Wong et al. A cell-based logit-opportunity taxi customer-search model
US20140035921A1 (en) Analysis and visualization of passenger movement in a transportation system
CN104025075A (en) Method and system for fleet navigation, dispatching and multi-vehicle, multi-destination routing
Zheng Urban computing
CN114372830A (en) Network taxi booking demand prediction method based on space-time multi-graph neural network
Yamamoto et al. Structured random walk parameter for heterogeneity in trip distance on modeling pedestrian route choice behavior at downtown area
CN104024801A (en) Method and system for navigation using bounded geograhic regions
Zhao et al. MePark: Using meters as sensors for citywide on-street parking availability prediction
CN114548811A (en) Airport accessibility detection method and device, electronic equipment and storage medium
CN114358808A (en) Public transport OD estimation and distribution method based on multi-source data fusion
CN115034429A (en) Urban taxi pick-up point recommendation method based on depth map neural network
Xu et al. A taxi dispatch system based on prediction of demand and destination
CN110262863A (en) A kind of methods of exhibiting and device of terminal main interface
Zhao et al. Developing a multiview spatiotemporal model based on deep graph neural networks to predict the travel demand by bus
Faial et al. A methodology for taxi demand prediction using stream learning
Ara et al. Ride Hailing Service Demand Forecast by Integrating Convolutional and Recurrent Neural Networks.
CN115994787A (en) Car pooling demand prediction matching method based on neural network
CN115759660A (en) Scheduling method, device, equipment and medium for unmanned vehicle

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