CN113380043B - Bus arrival time prediction method based on deep neural network calculation - Google Patents

Bus arrival time prediction method based on deep neural network calculation Download PDF

Info

Publication number
CN113380043B
CN113380043B CN202110922683.6A CN202110922683A CN113380043B CN 113380043 B CN113380043 B CN 113380043B CN 202110922683 A CN202110922683 A CN 202110922683A CN 113380043 B CN113380043 B CN 113380043B
Authority
CN
China
Prior art keywords
traffic
extracting
bus
road
arrival 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.)
Active
Application number
CN202110922683.6A
Other languages
Chinese (zh)
Other versions
CN113380043A (en
Inventor
田锋
邓普阳
张枭勇
刘宇鸣
张炳振
陈振武
王宇
周勇
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.)
Shenzhen Urban Transport Planning Center Co Ltd
Original Assignee
Shenzhen Urban Transport Planning Center Co Ltd
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 Shenzhen Urban Transport Planning Center Co Ltd filed Critical Shenzhen Urban Transport Planning Center Co Ltd
Priority to CN202110922683.6A priority Critical patent/CN113380043B/en
Publication of CN113380043A publication Critical patent/CN113380043A/en
Application granted granted Critical
Publication of CN113380043B publication Critical patent/CN113380043B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G08G1/0129—Traffic data processing for creating historical data or processing based on historical data

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Development Economics (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention provides a bus arrival time prediction method based on deep neural network calculation, and belongs to the technical field of public transport information processing. The method specifically comprises the steps of firstly preprocessing traffic data; secondly, extracting segmented traffic information to obtain road traffic characteristics; secondly, expanding the sample size of the road traffic characteristics; secondly, extracting historical characteristics and selecting information of multiple road sections by using an attention mechanism model for historical driving time data of various road sections; secondly, inputting the selected feature vectors into a full-connection layer, and training an attention mechanism model by using a mean square error as a loss function; and finally, obtaining the time for predicting the bus arrival. The method solves the technical problem that the prediction of the bus arrival time in the prior art is not accurate, and achieves the technical effects that the prediction of the bus arrival time between the bus stops is more real-time and accurate.

Description

Bus arrival time prediction method based on deep neural network calculation
Technical Field
The application relates to a bus arrival time prediction method, in particular to a bus arrival time prediction method based on deep neural network calculation, and belongs to the technical field of public transport information processing.
Background
The prediction of bus arrival time is one of the important components of the intelligent traffic system. Accurate real-time bus arrival time prediction can help travelers to select and optimize travel routes by themselves. In addition, accurate arrival time prediction can help public transport management personnel to adjust and optimize the bus scheduling time under the influence of an emergency. Therefore, real-time and accurate prediction of bus arrival time is one of the important methods for improving urban traffic efficiency by intelligent traffic. The traditional bus arrival time prediction is used for calculating historical average running time among bus stops by acquiring bus running time among all line stops and geographic information of buses and predicting the arrival time. At present, scholars at home and abroad have a great deal of research on a method for predicting bus arrival time, and used prediction models comprise: differential integration Moving Average autoregressive model (ARIMA), Support Vector Machines (SVMs), and Long-Short Term Memory artificial neural networks (LSTM). However, the time series prediction model based on the method is difficult to support the large data requirement of the intelligent public transportation at the present stage. The current public transit trip demand is bigger and bigger, and the kind of trip mode is more and more to and the rapid development of GPS positioning technology and communication technology.
The common bus arrival time prediction technology in the prior art is that an LSTM neural network is used for predicting bus arrival time, the bus arrival time prediction generally belongs to the prediction problem of time sequence, and under the assumption of the development continuity of the bus arrival time, the trend prediction is carried out on the bus arrival time in a future period according to the bus arrival time in a historical period and traffic environment data. The method for predicting the arrival time by using the time series prediction mode has many defects, for example, it is difficult to capture the trend of long period under the condition of limited time slice window, when the time window is too long, the calculation increases the difficulty in training the model, and the systematic extraction of real-time traffic information is lacked.
Therefore, the method for predicting the future arrival time by using the time sequence analysis cannot meet the requirement that the existing bus station can display the accurate arrival time in real time, and is more used for predicting the bus running time in a future period of time in future journey planning.
Disclosure of Invention
The following presents a simplified summary of the invention in order to provide a basic understanding of some aspects of the invention. It should be understood that this summary is not an exhaustive overview of the invention. It is not intended to determine the key or critical elements of the present invention, nor is it intended to limit the scope of the present invention. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is discussed later.
In view of the above, in order to solve the technical problem that the prediction of the bus arrival time is not accurate in the prior art, the invention provides a method, a device and a storage medium for predicting the bus arrival time based on deep neural network calculation.
A bus arrival time prediction method based on deep neural network calculation comprises the following steps:
s1, preprocessing traffic data;
s2, extracting segmented traffic information to obtain road traffic characteristics;
s3, expanding the sample size of the road section traffic characteristics;
s4, extracting historical characteristics and selecting information of multiple road sections by using an attention mechanism model for historical driving time data of various road sections;
s5, inputting the selected feature vectors into a full-connection layer, and training an attention mechanism model by using a mean square error as a loss function;
and S6, obtaining the time for predicting the bus arrival.
Preferably, the specific method for preprocessing the data in step S1 is: dividing a bus route into road sections between front and rear stops, and extracting traffic information of each road section; the traffic information specifically includes discrete variables and continuous variables.
Preferably, the step S2 of extracting the segmented traffic information to obtain the road traffic characteristics includes the following steps:
s2.1, establishing a traffic characteristic extraction framework of the bus line;
s2.2, extracting segmented traffic information between two stations on a line;
s2.3, extracting and strengthening the relation between the discrete features;
s2.4, extracting and strengthening the relation between the continuous features;
and S2.5, fusing the discrete features and the continuous features to obtain the extracted traffic features of the road section.
Preferably, the specific method for expanding the sample size of the road segment traffic characteristics in step S3 is as follows: mean-shift clustering is carried out according to the traffic characteristics of each road section, k cluster families are divided according to the clustering result, and the training samples of the cluster families are expanded.
Preferably, in step S4, the specific method for selecting the multi-route information from the historical travel time data of various routes using the attention mechanism model is as follows:
the information selection method is to use an attention mechanism model to select the information of the extracted features, and the specific formula is as follows:
Figure 590588DEST_PATH_IMAGE001
；
wherein the content of the first and second substances,
Figure 835624DEST_PATH_IMAGE002
for historical traffic characteristic data for that type of road segment,
Figure 415029DEST_PATH_IMAGE003
is composed of
Figure 606976DEST_PATH_IMAGE002
A matrix of the states is formed of,
Figure 218086DEST_PATH_IMAGE004
is composed of
Figure 102865DEST_PATH_IMAGE005
The number of the neuron is equal to the number of the neuron,
Figure 432215DEST_PATH_IMAGE006
the function activates it to a value of 0-1 and becomes the attention weight matrix, and finally
Figure 427853DEST_PATH_IMAGE002
The corresponding elements are multiplied one by one to become the feature vector selected by attention,
Figure 627890DEST_PATH_IMAGE007
is the bias vector for that neuron.
Preferably, the specific method for inputting the selected feature vector into the fully-connected layer and training the attention mechanism model by using the mean square error as the loss function in step S5 is as follows:
Figure 417992DEST_PATH_IMAGE008
wherein the content of the first and second substances,
Figure 234638DEST_PATH_IMAGE009
for the real travel time value of the road section,
Figure 768388DEST_PATH_IMAGE010
a travel time value is predicted for the road segment.
Preferably, the specific method for extracting and enhancing the relationship between the discrete features in step S2.3 is as follows:
Figure 91440DEST_PATH_IMAGE011
wherein the content of the first and second substances,
Figure 583601DEST_PATH_IMAGE003
is that
Figure 621965DEST_PATH_IMAGE012
The matrix of (a) is,
Figure 693826DEST_PATH_IMAGE013
the number of the discrete variables is the same as the number of the discrete variables,
Figure 134034DEST_PATH_IMAGE014
to input and use the one-hot coded discrete variables,
Figure 265938DEST_PATH_IMAGE007
is the bias vector for that neuron.
Preferably, the specific method for extracting and enhancing the relationship between the continuous features in step S2.4 is:
Figure 791598DEST_PATH_IMAGE015
wherein the content of the first and second substances,
Figure 667150DEST_PATH_IMAGE003
is that
Figure 961865DEST_PATH_IMAGE016
The matrix of (a) is,
Figure 530249DEST_PATH_IMAGE017
in order to continuously change the number of the variables,
Figure 546135DEST_PATH_IMAGE018
in order to be compressed to the number of dimensions,
Figure 225378DEST_PATH_IMAGE019
is a continuous variable of the input and is,
Figure 109020DEST_PATH_IMAGE007
is the bias vector for that neuron.
A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of a bus arrival time prediction method based on deep neural network calculation when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, implements a method for predicting bus arrival times based on deep neural network calculations.
The invention has the following beneficial effects: the invention extracts the road section information among the traffic stops in real time by establishing a traffic characteristic information extraction frame, increases the real-time prediction performance, enlarges the sample number of the traffic characteristic road sections through clustering, ensures more accurate prediction, trains prediction models respectively aiming at different types of road sections, solves the technical problem that the prediction of the bus arrival time in the prior art is not accurate, and realizes the technical effect that the prediction of the bus arrival time among the bus stops has more real-time performance and accuracy.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a schematic flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a clustering model according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an attention mechanism model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an RNN neuron historical time series feature extractor according to an embodiment of the present invention;
fig. 5 is an expanded schematic diagram of a RNN neuron historical time series feature extractor structure according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions and advantages of the embodiments of the present application more apparent, the following further detailed description of the exemplary embodiments of the present application with reference to the accompanying drawings makes it clear that the described embodiments are only a part of the embodiments of the present application, and are not exhaustive of all embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
Example one
Referring to fig. 1 to 5 to illustrate the embodiment, a method for predicting bus arrival time based on deep neural network calculation includes the following steps:
the bus route is a fixed route driven by the bus, and the bus route is the bus route in the invention
Figure 848306DEST_PATH_IMAGE020
Is composed of a series of GPS points
Figure 348557DEST_PATH_IMAGE021
The components of the composition are as follows,
Figure 565912DEST_PATH_IMAGE022
. GPS point
Figure 835219DEST_PATH_IMAGE023
Is the geographic position of the bus stop on the bus route,
Figure 745407DEST_PATH_IMAGE024
the problem to be solved by the invention is based on the bus line
Figure 467375DEST_PATH_IMAGE025
Last two sites
Figure 488420DEST_PATH_IMAGE023
，
Figure 451101DEST_PATH_IMAGE026
On real-time and
Figure 532190DEST_PATH_IMAGE025
last two sites
Figure 741454DEST_PATH_IMAGE023
，
Figure 300612DEST_PATH_IMAGE026
Historical traffic information is used for predicting the bus travel time between two stops of a bus line.
Firstly, preprocessing traffic data; the specific method comprises the following steps: dividing a bus route into a pair of road sections between front and rear stops, and extracting traffic information of each road section; the traffic information specifically comprises a discrete variable and a continuous variable; the discrete variables are as follows: day/week (1, 2, 3, 4, 5, 6, 7), hour/day (8, 9, 10 … … 20), continuous variables such as: the average station waiting time within 30 minutes and the average inter-station driving time within 30 minutes are shown in the table I in detail, and the table I is a traffic information table.
Figure 278932DEST_PATH_IMAGE027
Extracting segmented traffic information to obtain road traffic characteristics; the method comprises the following steps:
step two, establishing a traffic characteristic extraction frame of the bus line to extract the line
Figure 796501DEST_PATH_IMAGE025
Last two sites
Figure 493062DEST_PATH_IMAGE023
，
Figure 855910DEST_PATH_IMAGE026
Traffic information of the segments.
Secondly, extracting segmented traffic information between two stations on a line;
and step two and step three, extracting and strengthening the relation between the discrete features by using a Wide component by using a Wide & Deep model, wherein the relation is as follows, such as Monday, 7: 00-9: 00 "and" site 5-site 6 "may have strong correlation. The Wide component can be generalized to a generalized multi-layered linear model.
Extracting and strengthening the relationship between the discrete features, specifically:
Figure 954316DEST_PATH_IMAGE028
wherein, in the step (A),
Figure 111628DEST_PATH_IMAGE029
is that
Figure 298414DEST_PATH_IMAGE030
The matrix of (a) is,
Figure 464953DEST_PATH_IMAGE031
the number of the discrete variables is the same as the number of the discrete variables,
Figure 683445DEST_PATH_IMAGE032
to input and use the one-hot coded discrete variables,
Figure 277237DEST_PATH_IMAGE033
is the bias vector for that neuron.
And step two, simultaneously, extracting the relation between continuous variables, such as acceleration, average speed, traffic flow and the like of each section of road section by using a Deep component. Extracting and strengthening the relation between the continuous features, specifically:
Figure 682811DEST_PATH_IMAGE034
，
Figure 653041DEST_PATH_IMAGE029
is that
Figure 460460DEST_PATH_IMAGE035
The matrix of (a) is,
Figure 756312DEST_PATH_IMAGE036
in order to continuously change the number of the variables,
Figure 649182DEST_PATH_IMAGE037
in order to be compressed to the number of dimensions,
Figure 157523DEST_PATH_IMAGE038
is a continuous variable of the input and is,
Figure 822379DEST_PATH_IMAGE033
is the bias vector for that neuron.
And step two, fusing the discrete features and the continuous features to obtain the extracted traffic features of the road section. The specific method comprises the following steps:
Figure 23553DEST_PATH_IMAGE039
。
and step three, expanding the sample size of the road section traffic characteristics, performing mean-shift clustering according to the traffic characteristics of each road section, and dividing k cluster groups according to clustering results. The training samples of all the cluster groups are expanded, for example, the road sections in the special traffic rush hour can be clustered into similar cluster groups, and the sample size of the road sections with the traffic characteristics is expanded, so that the prediction is more accurate;
specifically, each segment is expanded using a mean-shift clustering model.
Step four, extracting historical characteristics and selecting information of multiple road sections for historical travel time data of various road sections by using an attention mechanism model, extracting the historical characteristics of various road sections for the historical travel time data of various road sections by using the attention mechanism model according to a traffic characteristic clustering result, and selecting the information of the extracted characteristics by using the attention mechanism model, wherein the specific method comprises the following steps:
the method for extracting the historical characteristics comprises the following steps: use of
Figure 403719DEST_PATH_IMAGE040
The hidden layer output is treated as a history feature.
An RNN neuron historical time series feature extractor, described with reference to fig. 4:
Figure 450172DEST_PATH_IMAGE041
as an input layer, is a time sequence of historical features
Figure 232183DEST_PATH_IMAGE042
，
Figure 338679DEST_PATH_IMAGE043
,
Figure 471721DEST_PATH_IMAGE044
The number of neurons in the input layer is,
Figure 321865DEST_PATH_IMAGE045
in order to have a cyclic hidden layer,
Figure 958383DEST_PATH_IMAGE046
is an input layer;
specifically, the following will be described with reference to fig. 5:
Figure 235780DEST_PATH_IMAGE041
is a time series of historical characteristics
Figure 327889DEST_PATH_IMAGE042
，
Figure 981724DEST_PATH_IMAGE043
,
Figure 738327DEST_PATH_IMAGE044
The number of neurons in the input layer;
Figure 186626DEST_PATH_IMAGE045
are correspondingly hidden layers
Figure 28680DEST_PATH_IMAGE047
，
Figure 220627DEST_PATH_IMAGE048
,
Figure 566158DEST_PATH_IMAGE049
The number of neurons in the hidden layer is implied.
The hidden layer defines the state space (state space) of the whole historical feature extraction system:
Figure 450937DEST_PATH_IMAGE050
wherein the content of the first and second substances,
Figure 514708DEST_PATH_IMAGE051
is the activation function of the hidden layer,
Figure 510346DEST_PATH_IMAGE052
Figure 978892DEST_PATH_IMAGE053
wherein the common function is hyperbolic tangent function
Figure 300152DEST_PATH_IMAGE054
，
Figure 116798DEST_PATH_IMAGE055
As a weight matrix, the weight matrix is,
Figure 650548DEST_PATH_IMAGE056
is the bias vector of the hidden layer.
Corresponding output layer is
Figure 970671DEST_PATH_IMAGE057
Wherein
Figure 197253DEST_PATH_IMAGE058
,
Figure 970037DEST_PATH_IMAGE059
The number of neurons in the output layer. The history feature extractor uses the vectors of the output layer as history features to select the input of the model for the attention information, i.e.
Figure 838635DEST_PATH_IMAGE060
。
The information selection method is to use an attention mechanism model to select the information of the extracted features, and the specific formula is as follows:
Figure 278844DEST_PATH_IMAGE061
wherein the content of the first and second substances,
Figure 413678DEST_PATH_IMAGE062
for historical traffic characteristic data for that type of road segment,
Figure 939337DEST_PATH_IMAGE029
is composed of
Figure 814889DEST_PATH_IMAGE062
A matrix of the states is formed of,
Figure 109604DEST_PATH_IMAGE063
is composed of
Figure 412410DEST_PATH_IMAGE064
The number of the neuron is equal to the number of the neuron,
Figure 690944DEST_PATH_IMAGE065
the function activates it to a value of 0-1 and becomes the attention weight matrix, and finally
Figure 370187DEST_PATH_IMAGE062
Multiplying corresponding elements one by one to form feature vectors selected by attention;
Figure 253830DEST_PATH_IMAGE033
is the bias vector for that neuron.
Inputting the selected feature vector into a full-connection layer, and training the model by using a mean square error as a loss function; the specific method comprises the following steps:
Figure 993116DEST_PATH_IMAGE066
wherein the content of the first and second substances,
Figure 493367DEST_PATH_IMAGE067
for the real travel time value of the road section,
Figure 713651DEST_PATH_IMAGE068
a travel time value is predicted for the road segment.
And step six, obtaining the time of predicting the bus arrival.
The invention consists of three layers of architectures, wherein the first layer is a data input layer, the second layer is a characteristic extraction layer and the third layer is an arrival time prediction layer.
The data input layer is used for processing data, dividing traffic data into continuous variables and discrete variables and inputting the data to the feature extraction layer.
The characteristic extraction layer receives data transmitted by the data input layer and processes the data by using a Wide & Deep model and an attention mechanism model.
And the arrival time prediction layer fuses the data through the clustered traffic characteristics and obtains the predicted arrival time through calculation.
The computer device of the present invention may be a device including a processor, a memory, and the like, for example, a single chip microcomputer including a central processing unit and the like. And the processor is used for implementing the steps of the recommendation method capable of modifying the relationship-driven recommendation data based on the CREO software when executing the computer program stored in the memory.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the cellular phone, and the like. In addition, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
Computer-readable storage medium embodiments
The computer readable storage medium of the present invention may be any form of storage medium that can be read by a processor of a computer device, including but not limited to non-volatile memory, ferroelectric memory, etc., and the computer readable storage medium has stored thereon a computer program that, when the computer program stored in the memory is read and executed by the processor of the computer device, can implement the above-mentioned steps of the CREO-based software that can modify the modeling method of the relationship-driven modeling data.
The computer program comprises computer program code which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice.
While the invention has been described with respect to a limited number of embodiments, those skilled in the art, having benefit of this description, will appreciate that other embodiments can be devised which do not depart from the scope of the invention as described herein. Furthermore, it should be noted that the language used in the specification has been principally selected for readability and instructional purposes, and may not have been selected to delineate or circumscribe the inventive subject matter. Accordingly, many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the appended claims. The present invention has been disclosed in an illustrative rather than a restrictive sense, and the scope of the present invention is defined by the appended claims.

Claims (7)

1. A bus arrival time prediction method based on deep neural network calculation is characterized by comprising the following steps:
s1, preprocessing data; the specific method comprises the following steps: dividing a bus route into a pair of road sections between front and rear stops, and extracting traffic information of each road section; the traffic information specifically comprises a discrete variable and a continuous variable;
s2, extracting segmented traffic information to obtain road traffic characteristics; the method comprises the following steps:
s2.1, establishing a traffic characteristic extraction framework of the bus line;
s2.2, extracting segmented traffic information between two stations on a line;
s2.3, extracting and strengthening the relation between the discrete features;
s2.4, extracting and strengthening the relation between the continuous features;
s2.5, fusing the discrete features and the continuous features to obtain the extracted traffic features of the road section;
s3, expanding the sample size of the road section traffic characteristics; mean-shift clustering is carried out according to the traffic characteristics of each road section, k cluster groups are divided according to clustering results, and training samples of each cluster group are expanded;
s4, extracting historical characteristics and selecting information of multiple road sections by using an attention mechanism model for historical driving time data of various road sections;
s5, inputting the selected feature vectors into a full-connection layer, and training an attention mechanism model by using a mean square error as a loss function;
and S6, obtaining the time for predicting the bus arrival.
2. The method according to claim 1, wherein the step S4 is implemented by using an attention mechanism model to select information from various types of historical travel time data of road segments:
the information selection method is to use an attention mechanism model to select the information of the extracted features, and the specific formula is as follows:
Figure DEST_PATH_IMAGE001
wherein the content of the first and second substances,
Figure 710243DEST_PATH_IMAGE002
for historical traffic characteristic data for that type of road segment,
Figure DEST_PATH_IMAGE003
is composed of
Figure 402256DEST_PATH_IMAGE002
A matrix of the states is formed of,
Figure 608109DEST_PATH_IMAGE004
is composed of
Figure DEST_PATH_IMAGE005
The number of the neuron is equal to the number of the neuron,
Figure 803598DEST_PATH_IMAGE006
the function activates it to a value of 0-1 and becomes the attention weight matrix, and finally
Figure 702284DEST_PATH_IMAGE002
Multiplying corresponding elements one by one to form feature vectors selected by attention;
Figure DEST_PATH_IMAGE007
is the bias vector for that neuron.
3. The method of claim 2, wherein the step S5 is to input the selected feature vectors into the fully-connected layer and train the attention mechanism model using the mean square error as the loss function by:
Figure 383058DEST_PATH_IMAGE008
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE009
for the real travel time value of the road section,
Figure 392602DEST_PATH_IMAGE010
a travel time value is predicted for the road segment.
4. A method according to claim 3, wherein step S2.3 the specific method of extracting and enhancing the relationship between discrete features is:
Figure DEST_PATH_IMAGE011
wherein the content of the first and second substances,
Figure 973756DEST_PATH_IMAGE003
is that
Figure 512185DEST_PATH_IMAGE012
The matrix of (a) is,
Figure DEST_PATH_IMAGE013
the number of the discrete variables is the same as the number of the discrete variables,
Figure 178789DEST_PATH_IMAGE014
to input and use the one-hot coded discrete variables,
Figure 224981DEST_PATH_IMAGE007
is the bias vector for that neuron.
5. The method according to claim 4, wherein step S2.4 extracts and strengthens the relationship between successive features by:
Figure DEST_PATH_IMAGE015
wherein the content of the first and second substances,
Figure 457379DEST_PATH_IMAGE003
is that
Figure 697867DEST_PATH_IMAGE016
The matrix of (a) is,
Figure DEST_PATH_IMAGE017
in order to continuously change the number of the variables,
Figure 586189DEST_PATH_IMAGE018
in order to be compressed to the number of dimensions,
Figure DEST_PATH_IMAGE019
is a continuous variable of the input and is,
Figure 140798DEST_PATH_IMAGE007
is the bias vector for that neuron.
6. A computer device, characterized by: the method comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the method for predicting the bus arrival time based on the deep neural network calculation according to any one of claims 1 to 5 when executing the computer program.
7. A computer-readable storage medium having stored thereon a computer program, characterized in that: the computer program is used for realizing the bus arrival time prediction method based on the deep neural network calculation in any one of claims 1 to 5 when being executed by a processor.
CN202110922683.6A 2021-08-12 2021-08-12 Bus arrival time prediction method based on deep neural network calculation Active CN113380043B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110922683.6A CN113380043B (en) 2021-08-12 2021-08-12 Bus arrival time prediction method based on deep neural network calculation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110922683.6A CN113380043B (en) 2021-08-12 2021-08-12 Bus arrival time prediction method based on deep neural network calculation

Publications (2)

Publication Number Publication Date
CN113380043A CN113380043A (en) 2021-09-10
CN113380043B true CN113380043B (en) 2022-01-14

Family

ID=77576894

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110922683.6A Active CN113380043B (en) 2021-08-12 2021-08-12 Bus arrival time prediction method based on deep neural network calculation

Country Status (1)

Country Link
CN (1) CN113380043B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113984078B (en) * 2021-10-26 2024-03-08 上海瑾盛通信科技有限公司 Arrival reminding method, device, terminal and storage medium
CN115291508B (en) * 2022-06-16 2023-08-29 扬州大学 Dynamic bus control system and method based on distributed deep reinforcement learning
CN116596126A (en) * 2023-04-27 2023-08-15 苏州大学 Bus string prediction method and system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0219859A2 (en) * 1985-10-25 1987-04-29 Mitsubishi Denki Kabushiki Kaisha Route bus service controlling system
JP2008204361A (en) * 2007-02-22 2008-09-04 Nec Corp Sending/fetching management system, and method for it, program, and storage medium thereof
CN104217605A (en) * 2013-05-31 2014-12-17 张伟伟 Bus arrival time estimation method and device
CN105096639A (en) * 2014-05-23 2015-11-25 中国电信股份有限公司 Method, device and system used for predicting bus arrival time
CN110570678A (en) * 2019-10-23 2019-12-13 厦门大学 Method and device for predicting total travel time of bus from starting point to end point
CN111554118A (en) * 2020-04-24 2020-08-18 深圳职业技术学院 Dynamic prediction method and system for bus arrival time

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0219859A2 (en) * 1985-10-25 1987-04-29 Mitsubishi Denki Kabushiki Kaisha Route bus service controlling system
JP2008204361A (en) * 2007-02-22 2008-09-04 Nec Corp Sending/fetching management system, and method for it, program, and storage medium thereof
CN104217605A (en) * 2013-05-31 2014-12-17 张伟伟 Bus arrival time estimation method and device
CN105096639A (en) * 2014-05-23 2015-11-25 中国电信股份有限公司 Method, device and system used for predicting bus arrival time
CN110570678A (en) * 2019-10-23 2019-12-13 厦门大学 Method and device for predicting total travel time of bus from starting point to end point
CN111554118A (en) * 2020-04-24 2020-08-18 深圳职业技术学院 Dynamic prediction method and system for bus arrival time

Also Published As

Publication number Publication date
CN113380043A (en) 2021-09-10

Similar Documents

Publication Publication Date Title
CN113380043B (en) Bus arrival time prediction method based on deep neural network calculation
CN111768618B (en) Traffic jam state propagation prediction and early warning system and method based on city portrait
CN113487066B (en) Long-time-sequence freight volume prediction method based on multi-attribute enhanced graph convolution-Informer model
CN112418236B (en) Automobile drivable area planning method based on multitask neural network
CN113221911B (en) Vehicle weight identification method and system based on dual attention mechanism
CN111178213A (en) Aerial photography vehicle detection method based on deep learning
Faro et al. Evaluation of the traffic parameters in a metropolitan area by fusing visual perceptions and CNN processing of webcam images
CN113763700B (en) Information processing method, information processing device, computer equipment and storage medium
CN110210664B (en) Deep learning method for short-term prediction of using behaviors of multiple individual vehicles
CN110956341B (en) Parking lot flow prediction method and device, electronic equipment and storage medium
CN113537626B (en) Method for predicting neural network combined time sequence by aggregating information difference
US20240143999A1 (en) Multi-modal data prediction method based on causal markov model
CN109919112B (en) Method for detecting distribution and counting of flowing crowds in complex scene
CN112633602B (en) Traffic congestion index prediction method and device based on GIS map information
CN113673749B (en) Vehicle-pedestrian map track prediction method and device
CN111414719B (en) Method and device for extracting peripheral features of subway station and estimating traffic demand
CN114005085A (en) Dense crowd distribution detection and counting method in video
Zhang et al. Short‐Term Passenger Flow Forecast of Rail Transit Station Based on MIC Feature Selection and ST‐LightGBM considering Transfer Passenger Flow
CN110021161B (en) Traffic flow direction prediction method and system
Zhong et al. Bus travel time prediction based on ensemble learning methods
CN113298314A (en) Rail transit passenger flow prediction method considering dynamic space-time correlation
CN115985104A (en) Traffic flow prediction device, prediction method and prediction model construction method
Zheng et al. A deep learning–based approach for moving vehicle counting and short-term traffic prediction from video images
Xia et al. Link-based traffic estimation and simulation for road networks using electronic registration identification data
CN116612633A (en) Self-adaptive dynamic path planning method based on vehicle-road cooperative sensing

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
CB03 Change of inventor or designer information
CB03 Change of inventor or designer information

Inventor after: Tian Feng

Inventor after: Deng Puyang

Inventor after: Zhang Xiaoyong

Inventor after: Liu Yuming

Inventor after: Zhang Bingzhen

Inventor after: Chen Zhenwu

Inventor after: Wang Yu

Inventor after: Zhou Yong

Inventor before: Lin Tao

Inventor before: Chen Zhenwu

Inventor before: Zhou Yong

Inventor before: Li Peng

Inventor before: Wang Yu

Inventor before: Peng Yahong

Inventor before: Zhang Xiaoyong

Inventor before: Wang Jinyun

Inventor before: Qian Yuqing

GR01 Patent grant
GR01 Patent grant