CN108898838B - Method and device for predicting airport traffic jam based on LSTM model - Google Patents

Method and device for predicting airport traffic jam based on LSTM model Download PDF

Info

Publication number
CN108898838B
CN108898838B CN201810878081.3A CN201810878081A CN108898838B CN 108898838 B CN108898838 B CN 108898838B CN 201810878081 A CN201810878081 A CN 201810878081A CN 108898838 B CN108898838 B CN 108898838B
Authority
CN
China
Prior art keywords
airport
information
lstm model
road
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.)
Expired - Fee Related
Application number
CN201810878081.3A
Other languages
Chinese (zh)
Other versions
CN108898838A (en
Inventor
周芳
张波
李强
缪明月
张军
李国军
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.)
Beijing Capital International Airport Public Security Bureau
CAPITAL UNIVERSITY OF ECONOMICS AND BUSINESS
Original Assignee
Beijing Capital International Airport Public Security Bureau
CAPITAL UNIVERSITY OF ECONOMICS AND BUSINESS
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 Beijing Capital International Airport Public Security Bureau, CAPITAL UNIVERSITY OF ECONOMICS AND BUSINESS filed Critical Beijing Capital International Airport Public Security Bureau
Priority to CN201810878081.3A priority Critical patent/CN108898838B/en
Publication of CN108898838A publication Critical patent/CN108898838A/en
Application granted granted Critical
Publication of CN108898838B publication Critical patent/CN108898838B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
    • 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
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G5/00—Traffic control systems for aircraft, e.g. air-traffic control [ATC]

Landscapes

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

Abstract

The invention relates to an airport traffic jam prediction method and device based on an LSTM model, wherein the method comprises the following steps: acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport and aeronautical meteorological information in the preset range around the airport in real time; inputting the traffic condition information, flight take-off and landing information and aeronautical meteorological information into an LSTM model; and obtaining an output result of the LSTM model, wherein the output result is a predicted road congestion index of a road in a preset range around an airport in a future period. The method is based on the consideration of space and time effects, the aeronautical meteorological information is added, the traffic in the area is regarded as a space-time related system, the prediction result is obtained based on the LSTM model, and the accuracy of the road congestion index prediction in the preset range around the airport can be further improved.

Description

Method and device for predicting airport traffic jam based on LSTM model
Technical Field
The invention relates to the technical field of computer science and intelligent traffic, in particular to an airport traffic jam prediction method and device based on an LSTM model.
Background
Along with the rapid promotion of the urbanization process of China, urban congestion becomes one of important factors which puzzle urban development, and how to improve the operation efficiency of urban traffic and relieve congestion pressure becomes a problem which needs to be solved for realizing sustainable health development of cities. In the face of the challenge, each large city and navigation company successively release the 'traffic jam delay index' as an important means for traffic management and guidance. For example, "traffic index" and "traffic operation analysis report" issued by the transportation commission of beijing city, and "congestion ranking of main city of china" issued by the high-end map.
The urban traffic management level can be effectively improved by utilizing the index information, but the traffic index can only reflect the current traffic condition and lacks predictability on the change of the future traffic condition. Traffic congestion not only reduces travel efficiency and increases travel cost and accident rate, but also causes energy waste and air pollution due to increase of oil consumption and incomplete combustion of fuel. How to avoid and relieve traffic congestion becomes a concern.
On the other hand, with the application of the car networking technology and the arrival of the big data era, people can monitor traffic congestion by various means, for example, [1] Andrea and the like use a GPS tracker and a smart phone to identify real-time traffic congestion and accidents, Kong and the like identify and predict urban traffic congestion through floating car track data, internet companies providing road condition services are also mainly based on floating car models, and Bauza and Gozalvez monitor road traffic congestion based on information exchange between vehicles and vehicles, and between vehicles and infrastructure nodes.
In recent decades, scholars have developed various models for traffic flow prediction problems, broadly classified into the following categories: (1) taking a time sequence method as a representative, building models such as ARIMA for traffic jam delay index, predicting traffic conditions in a period of time (such as day, hour, minute and the like) in the future, researching the application of the ARIMA model in traffic flow prediction by Voort and the like, and carrying out short-time prediction of traffic flow by utilizing seasonal ARIMA models by Williams and the like; (2) nonparametric methods such as K-nearest neighbor, support vector machine and kernel function regression, such as Smith and the like, relatively study the prediction effects of nonparametric regression and seasonal ARIMA models, carry out short-time prediction on the traffic flow based on the K-nearest neighbor nonparametric method in the shore and the like, and establish traffic flow and speed models by applying nonparametric regression in Daqingdong and the like; (3) modeling and predicting the traffic condition as a random process by using Bayes methods such as a Gaussian process and the like, such as the health and military; (4) application of Kalman filtering method in traffic flow prediction, such as XIEYUan-chang, XUDong-wei et al. (5) A short-term traffic flow prediction technology based on an artificial neural network, such as the research of people like Yao shiyao flood, and love. The methods have good application effect in specific scenes, but still need to be improved.
Disclosure of Invention
In view of the above problems, the invention provides an airport traffic congestion prediction method and device based on an LSTM model, the method considers space and time effects, increases aeronautical meteorological information, regards traffic in an area as a space-time related system, obtains a prediction result based on the LSTM model, and can further improve the accuracy of road congestion index prediction in a preset range around an airport.
In a first aspect, an embodiment of the present invention provides an airport traffic congestion prediction method based on an LSTM model, including:
acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport and aeronautical meteorological information in the preset range around the airport in real time;
inputting the traffic condition information, flight take-off and landing information and aeronautical meteorological information into an LSTM model;
and obtaining an output result of the LSTM model, wherein the output result is a predicted road congestion index of a road in a preset range around an airport in a future period.
In one embodiment, the LSTM model training process is as follows:
acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport and aeronautical meteorological information in the preset range around the airport in real time;
correlating the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information according to time;
obtaining road congestion index, aviation data and meteorological data in each hour after correlation, and merging holiday, working day and other activity time sequence data to form a data set;
and adding a hysteresis effect to the aviation data and the meteorological data to generate the LSTM model.
In one embodiment, adding a hysteresis effect to the aviation data and meteorological data to generate the LSTM model comprises:
the LSTM model is used for time series prediction to obtain an output vector htAnd then, connecting a full connection layer, thereby finally obtaining a predicted value:
Figure GDA0002546714040000031
where σ represents an activation function; woutIs a fully connected layer matrix with dimension 1 × dh；htRepresenting an output vector; boutRepresenting an intercept term;
Figure GDA0002546714040000032
the predicted value of the s-th time sequence t +1 phase is shown.
In one embodiment, the loss function of the LSTM model includes: MAE and MAPE loss functions;
the formula of the loss function of the forward prediction n steps is as follows:
Figure GDA0002546714040000033
in the formula (8), MAEsIs an index name for measuring the prediction effectWeighing; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure GDA0002546714040000034
representing an actual congestion index;
Figure GDA0002546714040000035
representing a predicted congestion index;
Figure GDA0002546714040000036
in formula (9), MAPEsIs another index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure GDA0002546714040000037
representing an actual congestion index;
Figure GDA0002546714040000038
representing the predicted congestion index.
In one embodiment, the real-time acquisition of traffic condition information of roads within a preset range around an airport includes:
and acquiring the traffic flow and the running speed of the road in the preset range around the airport in real time through a data interface, and calculating the road congestion index.
In a second aspect, an embodiment of the present invention provides an apparatus for predicting airport traffic congestion based on an LSTM model, including:
the acquisition module is used for acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time and aeronautical meteorological information in the preset range around the airport;
the input module is used for inputting the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information into an LSTM model;
and the prediction module is used for obtaining an output result of the LSTM model, wherein the output result is a congestion index of a road in a preset range around the airport in a predicted future period.
In one embodiment, the LSTM model includes:
the acquisition submodule is used for acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time and aeronautical meteorological information in the preset range around the airport;
the input submodule is used for correlating the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information according to time;
the forming submodule is used for obtaining road congestion indexes, aviation data and meteorological data in each hour according to the input submodule after correlation, and combining holiday, working day and other activity time sequence data to form a data set;
and the generation submodule is used for adding a hysteresis effect to the aviation data and the meteorological data to generate the LSTM model.
In an embodiment, the generating submodule is specifically configured to use the LSTM model for time series prediction to obtain an output vector htAnd then, connecting a full connection layer, thereby finally obtaining a predicted value:
Figure GDA0002546714040000041
where σ represents an activation function; woutIs a fully connected layer matrix with dimension 1 × dh；htRepresenting an output vector; boutRepresenting an intercept term;
Figure GDA0002546714040000042
the predicted value of the s-th time sequence t +1 phase is shown.
In one embodiment, the loss function of the LSTM model includes: MAE and MAPE loss functions;
the formula of the loss function of the forward prediction n steps is as follows:
Figure GDA0002546714040000051
formula (II)(8) In, MAEsIs the index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure GDA0002546714040000052
representing an actual congestion index;
Figure GDA0002546714040000053
representing a predicted congestion index;
Figure GDA0002546714040000054
in formula (9), MAPEsIs another index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure GDA0002546714040000055
representing an actual congestion index;
Figure GDA0002546714040000056
representing the predicted congestion index.
In one embodiment, the obtaining sub-module obtains traffic condition information of roads in a preset range around an airport in real time, and the obtaining sub-module includes:
and acquiring the traffic flow and the running speed of the road in the preset range around the airport in real time through a data interface, and calculating the road congestion index.
The technical scheme provided by the embodiment of the invention has the beneficial effects that at least:
the method for predicting the traffic jam of the airport based on the LSTM model, provided by the embodiment of the invention, comprises the steps of acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around the airport in real time and aeronautical meteorological information in the preset range around the airport; inputting the traffic condition information, flight take-off and landing information and aeronautical meteorological information into an LSTM model; and obtaining an output result of the LSTM model, wherein the output result is a predicted road congestion index of a road in a preset range around an airport in a future period. The method is based on the consideration of space and time effects, the aeronautical meteorological information is added, the traffic in the area is regarded as a space-time related system, the prediction result is obtained based on the LSTM model, and the accuracy of the road congestion index prediction in the preset range around the airport can be further improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a flowchart of an airport traffic congestion prediction method based on an LSTM model according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the RNN model structure;
FIG. 3 is a block diagram of the classical RNN model (left) and the LSTM model (right);
FIG. 4 is a diagram of an LSTM model module;
FIG. 5 is a flow chart of an LSTM model training process provided by an embodiment of the present invention;
FIG. 6 is a schematic diagram of a high-speed congestion delay index for a capital airport in 7 months in 2017;
FIG. 7 is a schematic diagram of a relationship between an airport highway congestion delay index and an airline passenger flow;
FIG. 8 is a block diagram of an apparatus for predicting airport traffic congestion based on an LSTM model according to an embodiment of the present invention;
fig. 9 is a block diagram of an LSTM model provided by an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Referring to fig. 1, an airport traffic congestion prediction method based on an LSTM model according to an embodiment of the present invention includes: s101 to S103;
s101, acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time, and aeronautical meteorological information in the preset range around the airport;
s102, inputting the traffic condition information, the flight taking-off and landing information and the aeronautical meteorological information into an LSTM model;
s103, obtaining an output result of the LSTM model, wherein the output result is a congestion index of a road in a preset range around an airport in a predicted future period.
In this embodiment, in step S101, traffic condition information of roads in a preset range around the airport, such as traffic flow on the roads and vehicle running speed, is obtained in real time, so as to further calculate the road congestion index.
Taking a Beijing capital airport as an example, acquiring traffic flow and running speed of 58 roads around the airport in real time through a data interface, and further calculating a road congestion index, wherein a calculation formula is the real-time average running speed divided by the designed hourly speed of the roads; or may be a road congestion index calculated by other conventional calculation methods. Such as: if the current traffic flow is 0, the congestion index is 1 (indicating no congestion at all), the congestion index is larger, indicating that the road is congested, and finally obtaining the following data structure:
Figure GDA0002546714040000071
acquiring flight take-off and landing information of the Beijing capital airport in real time, wherein the flight take-off and landing information comprises actual domestic arrival number, actual international arrival number, actual domestic departure number, actual international departure number, planned domestic arrival number, planned international arrival number, planned domestic departure number, planned international departure number, and planned international departure number.
Through an airport data interface, the first airport flight data is obtained in real time, particularly when air traffic control exists, the actual aviation data is 0, but the planning data is still not 0. The acquired flight information is specific flight details, such as 21 o' 03, flights from beijing to shanghai, and the acquired data is summarized every hour, namely the specific flight information is summarized into data, and the summarized data means that the sum of the number of arriving people and the sum of the number of arriving people in China between 0 and 1 of 20170809 days.
And organizes the data into the following structure:
Figure GDA0002546714040000081
acquiring aeronautical meteorological information around an airport in real time, wherein the aeronautical meteorological information comprises aeronautical influence factors such as wind speed, wind direction, temperature, rainfall, thunder, fog, dust and the like;
the method comprises the steps of obtaining real-time aviation meteorological message data through a meteorological data interface, wherein the real-time aviation meteorological message data comprises a routine weather Message (METAR) and a special weather message (SPECI), extracting meteorological information in the messages through a message analysis module, wherein the special meteorological data are message data every half hour, the meteorological condition closest to an integral point time is taken as the meteorological condition of the integral point time when modeling is carried out, if the meteorological message time is 21-point 05 minutes, the meteorological data are taken as the meteorological data of the 21-point integral point when modeling, and the aviation data are organized into the following structure:
Figure GDA0002546714040000082
inputting the traffic condition information, flight take-off and landing information and aeronautical meteorological information into an LSTM model; including model dependent variables and model independent variables; the model dependent variable is, for example, an hour-level congestion index of 58 roads around the airport, and the total number of the model dependent variables is 58. The model independent variables include meteorological data, flight data, and congestion index lag terms. Wherein the meteorological variables include: the wind speed, wind direction, temperature, precipitation, thunder, fog and dust are 7 variables; the aviation variables include: the actual domestic number of arriving at, actual international number of arriving at, actual domestic number of departing from, actual international number of departing from, planned domestic number of arriving at, planned international number of arriving at, planned domestic number of departing from, planned international number of departing from, total 16 variables; the second-order lag term of the congestion index, such as the predicted congestion index at 9 am, is the congestion index at 8 am and 7 am.
The output of the LSTM model, namely: and predicting the congestion index of the road in the preset range around the airport in the future period. When the future traffic condition is predicted, the aviation data, the meteorological data and the holiday data are taken as input and are brought into a model for operation to obtain the congestion index predicted value of 58 roads around the airport. The method is based on the consideration of space and time effects, the aeronautical meteorological information is added, the traffic in the area is regarded as a space-time related system, the prediction result is obtained based on the LSTM model, and the accuracy of the road congestion index prediction in the preset range around the airport can be further improved.
The LSTM model was first proposed by SeppHochreiter and Jurgen Schmidhuber in 1997 to address one specific variation of gradient disappearance or gradient expansion in the RNN model. By introducing multiple thresholds into the RNN, the integral can be changed at different times with fixed model parameters, thereby avoiding the problem of gradient disappearance or expansion. The LSTM model achieves surprising results in sequence data prediction.
The LSTM model is one of the most widely applied models in the RNN model, and the RNN model has the greatest characteristic that the hidden layer output of the RNN model is not only connected with the output layer, but also connected with the hidden layer at the next moment. Fig. 2 is a schematic diagram of an RNN model structure, and the right side of the equal sign is an expanded RNN network structure, which includes an input layer X, a hidden layer module a, and an output layer h, where the repeated module a has the same structure at any time in the network. Unlike a typical neural network, the output of module a includes the same two copies, one copy passed to the output layer and one copy passed to the next hidden module. Since module a transforms the input variable through only one tanh activation function in the classical RNN model, the RNN model often encounters the problems of gradient disappearance and gradient expansion when processing sequence data, especially long sequences, and finally causes the model to be non-convergent.
Referring to FIG. 3, the LSTM model is developed based on the RNN model, with the difference in the intermediate repeat modules. The LSTM model repetitive model (also called a storage unit) comprises three threshold structures consisting of four related active functions (three of which are sigmoid functions and one is tanh function), namely a forgetting gate, an input gate and an output gate, and the input and the output of the switch control information are simulated through the threshold structures, so that the forward and backward propagation of a training error is realized, and the aim of converging the training model is fulfilled.
It is now assumed that there are k endogenous time series variables
Figure GDA0002546714040000101
m exogenous variables
Figure GDA0002546714040000102
Order to
Figure GDA0002546714040000103
The goal now is to build an LSTM model to predict future endogenous variables
Figure GDA0002546714040000104
In order to make the LSTM model keep memory of the sequence, a lag term x is used for constructing a samplet-1，xt-2From this point of view, therefore, the LSTM model is used in time seriesThe VAR model can be degenerated to a VAR model in the case where the activation functions are all linear transformations, and thus is actually a special case of an LSTM model.
Referring to fig. 4, the core of the LSTM model is three threshold structures in the storage unit, and the storage unit finally obtains a state vector CtAnd output vector htThe method specifically comprises the following three processes:
1) a forgetting threshold (forget gate) is formed by formula 1, and a connection matrix W is setfTo determine the output vector h from the previous timet-1Which information is removed or retained
ft＝σ(Wf·[ht-1,xt]+bf) Formula (1)
Wherein x istIs dk+mDimension original input, ht-1Is dhDimension output, hidden layer dimension dc(in general d)h＝dc)，[ht-1,xt]Means that two vectors are spliced into a longer vector with dimension dh+dk+m，WfCoefficient matrix dimension for forgetting gate is dc×(dh+dk+m)，bfIs a bias vector with dimension dcF is obtained after point-by-point transformation of the activation function sigmatDimension d ofc。
2) Input gate (input gate) for determining the cell state vector CtHow to update is specifically as follows:
it＝σ(Wf·[ht-1,xt]+bi) Formula (2)
Figure GDA0002546714040000105
Figure GDA0002546714040000111
The vector i is obtained by the formula (2)tDimension d ofcThen, the temporary state vector is obtained by equation (3)
Figure GDA0002546714040000112
Finally, the last-phase state vector Ct-1And a temporary shape vector
Figure GDA0002546714040000113
Weighted summation, updating the state vector Ct。
3) The output gate (output gate) is formed by the equations (5) and (6), and the latest cell state vector C is obtainedtThen, using tanh activation function to obtain output vector ht
ot＝σ(Wo·[ht-1,xt]+bo) Formula (5)
ht＝ot*tanh(Ct) Formula (6)
The structures of the modules in the LSTM model are completely the same, and the weight matrix W is shared among different modulesf、Wi、WcAnd WoThis allows the scale of the model parameters to be controlled.
When the LSTM model is used for time series prediction, an output vector h is obtainedtThen, a full link layer is also connected to obtain the predicted value finally, as shown in equation (7)
Figure GDA0002546714040000114
Where σ represents an activation function; woutIs a fully connected layer matrix with dimension 1 × dh；htRepresenting an output vector; boutRepresents an intercept term, also called offset, as a parameter to be estimated;
Figure GDA0002546714040000115
the predicted value of the s-th time sequence t +1 phase is shown.
MAE (mean absolute error) and MAPE (mean absolute percentage error) are chosen herein as the loss function of the LSTM model, mainly considering two points: first, time series prediction is essentially a regression problem and thus is not suitable for selecting cross entropy as a loss function, which is commonly used in classification problems; second, MAE and MAPE are more robust than MSE (mean square error). Forward prediction of the loss function for n steps is shown in equations (8) - (9)
Figure GDA0002546714040000116
Figure GDA0002546714040000117
Represents the true congestion index of a road s, such as one of 58 roads, at time i (e.g., 9 am), and
Figure GDA0002546714040000118
representing predicted congestion index, n representing the step size of forward prediction (n hours into the prediction), MAEsIs the index name for measuring the prediction effect, and is called Mean Absolute Error in English.
Figure GDA0002546714040000121
Figure GDA0002546714040000122
Represents the true congestion index of a road s, such as one of 58 roads, at time i (e.g., 9 am), and
Figure GDA0002546714040000123
representing predicted congestion index, n representing step size of forward prediction (n hours into the prediction), MAPEsIs another index name for measuring the prediction effect, and is called Mean Absolute Percent Error in English.
The method for solving the LSTM model is a BPTT (Back Propagation Through time) algorithm, and based on the BPTT algorithm, the method can conduct gradients in different network layers and at different moments so as to update model parameters.
In one embodiment, the above LSTM model training process is as follows: referring to fig. 5, the method includes: S501-S504;
s501, acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time, and aeronautical meteorological information in the preset range around the airport;
s502, correlating the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information according to time;
s503, obtaining road congestion indexes, aviation data and meteorological data in each hour after correlation, and combining holiday, working day and other activity time sequence data to form a data set;
s504, adding a hysteresis effect to the aviation data and the meteorological data to generate the LSTM model.
In step S503, all the acquired and analyzed data are correlated according to time, and after the correlation, the road congestion index, the aviation data and the meteorological data per hour can be obtained, and meanwhile, a complete data set is formed according to information such as time-related holidays. The road congestion index is a dependent variable, and all other data are independent variables.
The step of training the model specifically comprises: and the aviation data and the meteorological data are added to increase the hysteresis effect. For example, if the current time is 10 am on 9/8/2017, aviation and meteorological data at three times of 9 am on 9/8/2017, 8 am on 9/2017, 8 am on 8/7 am, and 7 am on 9/2017 are added to the predicted independent variable, the reason is that the traffic condition at the current time is influenced by factors such as aviation at the previous times, for example, after a flight arrives, a traveler needs to take a baggage or take a ground vehicle for a certain time, which is often a significant time lag.
The accuracy of the prediction method of the present invention is demonstrated by the following specific examples.
1 data description and descriptive analysis
For example, the traffic jam delay index (provided by Gaode) and the flight inbound and outbound data (provided by China civil aviation science and technology research institute) of 58 roads in the first-capital international airport district of Beijing from 1: 00:00:00 hour at 8 months and 1 st day of 2016 to 31 days 23:00:00 at 31 months and 7 years are used as the data, the time granularity is hour level, and 8756 time points are counted. The variables thus collated and derived are shown in Table 1. The congestion delay index is a ratio of an average one-trip actual travel time of urban residents to a free flow state, and is used as an evaluation index of congestion degree.
Table 1: description of variables
Figure GDA0002546714040000131
Through data description analysis, the following rules were found:
(1) the difference of the congestion conditions of different roads is obvious
Modeling and predicting 58 road congestion delay indexes in the district of the Beijing capital international airport1,…,y58The congestion delay index time-series data representing 58 roads. The congestion conditions of all roads have great difference, wherein an airport freight road, an airport high speed and an airport high speed auxiliary road are three busiest roads, the average congestion delay index 2 is more than 1.3, and the rest roads are rarely congested. The congestion delay index of the freight road of the airport is the highest, and the average congestion delay index reaches 1.9, but the reason is that taxis wait in line instead of real traffic congestion. The congestion conditions of three main roads connecting the Beijing urban area and the airport are greatly different, the airport high speed is the shortest high speed trunk road connecting the airport and the Beijing urban area, the congestion delay index is highest, and secondly, the airport high speed auxiliary road is provided, although the airport freight road has the effect of shunting the passenger flow of the airport, the road has no obvious traffic congestion even in the peak time of the morning and evening. Aiming at the characteristic that the congestion condition difference of each road is large, each road is modeled respectively, and possible mutual influence among the roads is considered during modeling.
(2) The periodic characteristics of road congestion are obvious
FIG. 6 shows 2017Moon cakeCapital machineField(s)High speed Congestion delay index, from FIG. 6SeeThe high-speed congestion delay index of the capital airport presents obvious periodicity, the difference between working days, saturday and sunday is obvious, the peak time period of the working days usually appears at 7: 00-11: 00, the congestion delay indexes are all above 2, and the saturday baseThe congestion does not occur all day long, and the congestion in weekdays generally occurs at 13: 00-17: 00 in the afternoon. Therefore, the modeling introduces time factors of period, week and whether or not to work day. It can be observed from fig. 6 that the congestion delay index and the period have a relationship similar to a sine curve, and are considered to be introduced into the harmonic analysis
Figure GDA0002546714040000141
And
Figure GDA0002546714040000142
two terms are used for representing time interval effect, and the like is considered to be introduced
Figure GDA0002546714040000143
The four items are to represent week and month effects, where t represents time period, w represents week, and m represents month. The benefit of this process is that not only the number of parameters to be estimated can be reduced, but also the periodicity is taken into account, compared to introducing virtual variables for the period, week, month.
(3) Ground traffic is closely related to aviation factors
Referring to fig. 7, since there may be inconsistency between the actual time of arrival (departure) and the planned time of arrival (departure), the inbound and outbound flight information is processed to obtain inbound and outbound passenger flow data in two dimensions, one per hour, the planned time of the company and the actual time of arrival (departure). According to empirical judgment, people usually arrive at an airport 1-2 hours ahead of the scheduled departure time of a flight and leave the airport 1-2 hours after the flight actually enters the airport. Fig. 7 shows a time chart of the planned departure number and the actual arrival number in the lag phase and the airport high-speed congestion delay index, wherein the airport high-speed congestion delay index has a strong correlation with the planned departure number and the actual arrival number in the lag phase, and the correlation coefficient is calculated to be about 0.6. Therefore, 4 variables related to aviation factors are introduced into the planned departure number and the actual arrival number in the lag 1 and lag 2 stages for predicting the ground traffic congestion delay index.
3.2 data set construction
The modeling herein includes the following variables: the total of 174 variables of 58 roads and 2-order lag items thereof, 4 variables of 1-order lag items and 2-order lag items of the departure number according to the planned time and the arrival number according to the actual time and 7 time variables, and the total of 185 variables, wherein the sample size is 8754. (the correlation coefficient of the planned departure number of people in the lag 3 stage and the road congestion delay index is only 0.23, and the correlation coefficient of the actual arrival number of people in the lag 3 stage and the road congestion delay index is also only 0.15, so that only the aviation passenger flow data in the lag 1 stage and the lag 2 stage are considered, and the sample size is lost 2.)
To construct a robust predictive model, prevent overfitting, and compare the predictive effects of the model, the dataset was divided into three parts, training, validation, and test sets, with the first 80% of the sequence data as the training set (2016 (1/00: 00:00) and the last 20% of the data as the cross-validation set (20/00: 00/2017 (5/20/00: 00) and the 24/23: 00: 00/2017 (7/24/00: 00) and the last week of the dataset (00/00: 00: 00/2017/31/2017/7 (25/7/31/2017)) as the test sets. The training set is used for establishing a model and estimating model parameters, the verification set is used for selecting the model parameters and preventing the model from being over-fitted, and the test set is used for comparing final prediction results. The MAE and the MAPE are selected as evaluation indexes of the model prediction effect, and the smaller the value of the MAE and the MAPE is, the better the model prediction effect is.
3.3 comparative model
Linear models ARMA and VAR were chosen as control models. Unit root tests are performed on all time series involved in the study, and unit root primitive hypothesis is rejected under the significance level of 0.05, so that the time series are stable, and ARMA and VAR models can be established. The ARMA model and the VAR model are shown in formulas (10) to (11).
Figure GDA0002546714040000161
Figure GDA0002546714040000162
Wherein the content of the first and second substances,
Figure GDA0002546714040000163
the congestion delay index of the kth road in the t-th period is represented, DS represents the departure number of people according to planned time, AR represents the arrival number of people according to actual time, and WE represents a virtual variable of whether to work or not.
Unlike the LSTM model, both the ARMA and VAR models are linear models. The ARMA model only considers the influence of the lag phase of each road, and the VAR model not only considers the influence of the lag phase of each road, but also adds the influence of the lag phase of other roads, namely considers the possible spatial correlation.
4 model results and analysis
For example, a first airport expressway bidirectional road of the 58 roads is selected, namely the direction from the first airport to the east airport and the direction from the east airport to the first airport, and the first airport bidirectional road is used for comparing the prediction effect of the model. (see tables 2-3)
TABLE 2 high speed (first airport to east straight door direction) congestion delay index prediction comparison for first airport
Figure GDA0002546714040000164
Figure GDA0002546714040000171
TABLE 3 high speed (east-straight gate to first-all airport directions) Congestion delay index prediction comparison for first-all airports
Figure GDA0002546714040000172
From the perspective of predictive effect, the LSTM model is consistently superior to the ARMA and VAR models over almost all prediction periods. For the direction from the high-speed capital airport to the eastern orthopaedics of the airport, the MAPE mean value of the LSTM model is only 8.6 percent in the prediction period, and the prediction effect evaluated by MAE is 42 percent higher than ARMA and 22 percent higher than VAR. Compared with the LSTM model without considering aviation factors, the prediction effect of the LSTM model considering aviation factors is improved by 13%, and the particularity of airport road traffic prediction is verified, namely the influence of aviation factors is large.
For the direction from the east-west high speed airport to the capital airport, the MAPE mean value of the LSTM model is only 9.0% in the prediction period, the prediction effect is improved by 20% and 10% compared with that of the ARMA and the VAR, and the prediction effect is improved by 9% compared with that of the LSTM model without considering aviation factors.
From the stability of the two indexes of MAE and MAPE, the two indexes do not increase significantly with the time, and the model prediction capability is good in robustness.
The embodiment predicts the airport road congestion delay index based on the Beijing capital international airport air-ground data. The method proves that the aviation passenger flow in the lag phase has obvious influence on the improvement of the road traffic jam prediction effect, and the MAE is improved by about 10%; secondly, the prediction effect of the LSTM model based on the deep learning algorithm is consistent better than that of linear models ARMA and VAR, and the prediction precision is obviously improved. According to the embodiment of the invention, when the traffic jam is predicted, different roads have different road attributes and different factors influencing traffic conditions, and the effect of the prediction model is obviously improved if the reasons causing the traffic jam can be fully analyzed.
The method provided by the embodiment of the invention is beneficial to improving the prediction effect of the airport road traffic jam delay index, thereby providing more accurate guidance for the work of related management departments, such as the deployment of police force of traffic police and the like. In future research, more factors influencing traffic conditions, such as weather conditions, holiday factors, traffic accidents and the like, can be further considered, so that the accuracy of airport road prediction is further improved.
Based on the same inventive concept, the embodiment of the invention also provides an airport traffic jam prediction device based on the LSTM model, and as the principle of the problem solved by the device is similar to the airport traffic jam prediction method based on the LSTM model, the implementation of the device can refer to the implementation of the method, and repeated details are omitted.
The embodiment of the invention also provides an airport traffic jam prediction device based on the LSTM model, which is shown in figure 8,
the acquisition module 81 is used for acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time and aeronautical meteorological information in the preset range around the airport;
the input module 82 is used for inputting the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information into an LSTM model;
and the prediction module 83 is configured to obtain an output result of the LSTM model, where the output result is a congestion index of a road in a preset range around an airport in a future time period.
In one embodiment, the LSTM model includes:
the obtaining submodule 91 is used for obtaining traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time and aeronautical meteorological information in the preset range around the airport;
the input submodule 92 is used for correlating the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information according to time;
the composition submodule 93 is used for obtaining road congestion indexes, aviation data and meteorological data in each hour after the input submodules are correlated, and merging holiday, working day and other activity time sequence data to form a data set;
a generation submodule 94 for adding a hysteresis effect to the aerial data and the meteorological data to generate the LSTM model.
In an embodiment, the generating sub-module 94 is specifically configured to use the LSTM model for time series prediction to obtain an output vector htAnd then, connecting a full connection layer, thereby finally obtaining a predicted value:
Figure GDA0002546714040000191
where σ represents an activation function; woutIs a fully connected layer matrix with dimension 1 × dh；htRepresenting an output vector; boutRepresenting an intercept term;
Figure GDA0002546714040000192
the predicted value of the s-th time sequence t +1 phase is shown.
In one embodiment, the loss function of the LSTM model includes: MAE and MAPE loss functions;
the formula of the loss function of the forward prediction n steps is as follows:
Figure GDA0002546714040000193
in the formula (8), MAEsIs the index name for measuring the prediction effect; t represents time; l represents a congestion index; n represents the step size of the forward prediction;
Figure GDA0002546714040000194
represents a road s;
Figure GDA0002546714040000195
representing a predicted congestion index;
Figure GDA0002546714040000196
in formula (9), MAPEsIs another index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure GDA0002546714040000197
representing an actual congestion index;
Figure GDA0002546714040000198
representing the predicted congestion index.
In one embodiment, the obtaining sub-module 91 obtains traffic condition information of roads within a preset range around an airport in real time, and includes:
and acquiring the traffic flow and the running speed of the road in the preset range around the airport in real time through a data interface, and calculating the road congestion index.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (8)

1. An airport traffic jam prediction method based on an LSTM model is characterized by comprising the following steps:
acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport and aeronautical meteorological information in the preset range around the airport in real time; the traffic condition information includes: the method comprises the following steps of obviously obtaining difference information of congestion conditions of different roads, obviously obtaining periodic characteristic information of the congestion of the roads and closely related information of ground traffic and aviation factors;
inputting the traffic condition information, flight take-off and landing information and aeronautical meteorological information into an LSTM model;
obtaining an output result of the LSTM model, wherein the output result is a congestion index of a road in a preset range around an airport in a predicted future period;
the LSTM model training process is as follows:
acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport and aeronautical meteorological information in the preset range around the airport in real time;
correlating the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information according to time;
obtaining road congestion index, aviation data and meteorological data in each hour after correlation, and merging holiday, working day and other activity time sequence data to form a data set;
and adding a hysteresis effect to the aviation data and the meteorological data to generate the LSTM model.
2. The method of claim 1, adding a hysteresis effect to the aviation data, meteorological data, and generating the LSTM model, comprising:
will LSTThe M model is used for time series prediction to obtain an output vector htAnd then, connecting a full connection layer, thereby finally obtaining a predicted value:
Figure FDA0002546714030000011
where σ represents an activation function; woutIs a fully connected layer matrix with dimension 1 × dh；htRepresenting an output vector; boutRepresenting an intercept term;
Figure FDA0002546714030000012
the predicted value of the s-th time sequence t +1 phase is shown.
3. The method of claim 1, wherein the loss function of the LSTM model comprises: MAE and MAPE loss functions;
the formula of the loss function of the forward prediction n steps is as follows:
Figure FDA0002546714030000013
in the formula (8), MAEsIs the index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure FDA0002546714030000021
representing an actual congestion index;
Figure FDA0002546714030000022
representing a predicted congestion index;
Figure FDA0002546714030000023
in formula (9), MAPEsIs another index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents forward advanceMeasuring the step length;
Figure FDA0002546714030000024
representing an actual congestion index;
Figure FDA0002546714030000025
representing the predicted congestion index.
4. The method according to any one of claims 1 to 3, wherein the real-time acquisition of traffic condition information of roads within a predetermined range around an airport comprises:
and acquiring the traffic flow and the running speed of the road in the preset range around the airport in real time through a data interface, and calculating the road congestion index.
5. An apparatus for predicting traffic congestion at an airport based on an LSTM model, comprising:
the acquisition module is used for acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time and aeronautical meteorological information in the preset range around the airport; the traffic condition information includes: the method comprises the following steps of obviously obtaining difference information of congestion conditions of different roads, obviously obtaining periodic characteristic information of the congestion of the roads and closely related information of ground traffic and aviation factors;
the input module is used for inputting the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information into an LSTM model;
the prediction module is used for obtaining an output result of the LSTM model, and the output result is used for predicting the congestion index of the road in the preset range around the airport in the future period;
the LSTM model comprises:
the acquisition submodule is used for acquiring traffic condition information and airport flight taking-off and landing information of roads in a preset range around an airport in real time and aeronautical meteorological information in the preset range around the airport;
the input submodule is used for correlating the traffic condition information, the flight take-off and landing information and the aeronautical meteorological information according to time;
the forming submodule is used for obtaining road congestion indexes, aviation data and meteorological data in each hour according to the input submodule after correlation, and combining holiday, working day and other activity time sequence data to form a data set;
and the generation submodule is used for adding a hysteresis effect to the aviation data and the meteorological data to generate the LSTM model.
6. The apparatus of claim 5, wherein the generation submodule, in particular for using an LSTM model for time series prediction, results in an output vector htAnd then, connecting a full connection layer, thereby finally obtaining a predicted value:
Figure FDA0002546714030000031
where σ represents an activation function; woutIs a fully connected layer matrix with dimension 1 × dh；htRepresenting an output vector; boutRepresenting an intercept term;
Figure FDA0002546714030000032
the predicted value of the s-th time sequence t +1 phase is shown.
7. The apparatus of claim 5, wherein the loss function of the LSTM model comprises: MAE and MAPE loss functions;
the formula of the loss function of the forward prediction n steps is as follows:
Figure FDA0002546714030000033
in the formula (8), MAEsIs the index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure FDA0002546714030000034
representing an actual congestion index;
Figure FDA0002546714030000035
representing a predicted congestion index;
Figure FDA0002546714030000036
in formula (9), MAPEsIs another index name for measuring the prediction effect; s represents a road; t represents time; l represents a time series; n represents the step size of the forward prediction;
Figure FDA0002546714030000037
representing an actual congestion index;
Figure FDA0002546714030000038
representing the predicted congestion index.
8. The apparatus according to any one of claims 5-7, wherein the obtaining sub-module obtains traffic condition information of roads within a predetermined range around the airport in real time, and comprises:
and acquiring the traffic flow and the running speed of the road in the preset range around the airport in real time through a data interface, and calculating the road congestion index.
CN201810878081.3A 2018-08-03 2018-08-03 Method and device for predicting airport traffic jam based on LSTM model Expired - Fee Related CN108898838B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810878081.3A CN108898838B (en) 2018-08-03 2018-08-03 Method and device for predicting airport traffic jam based on LSTM model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810878081.3A CN108898838B (en) 2018-08-03 2018-08-03 Method and device for predicting airport traffic jam based on LSTM model

Publications (2)

Publication Number Publication Date
CN108898838A CN108898838A (en) 2018-11-27
CN108898838B true CN108898838B (en) 2020-08-28

Family

ID=64353052

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810878081.3A Expired - Fee Related CN108898838B (en) 2018-08-03 2018-08-03 Method and device for predicting airport traffic jam based on LSTM model

Country Status (1)

Country Link
CN (1) CN108898838B (en)

Families Citing this family (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109190948B (en) * 2018-08-20 2020-06-12 北京航空航天大学 Correlation analysis method for operation of large-scale aviation hub and urban traffic jam
CN109559512B (en) * 2018-12-05 2021-08-24 北京掌行通信息技术有限公司 Regional traffic flow prediction method and device
CN109635246B (en) * 2018-12-06 2021-07-13 西南交通大学 Multi-attribute data modeling method based on deep learning
CN109887272B (en) * 2018-12-26 2021-08-13 创新先进技术有限公司 Traffic pedestrian flow prediction method and device
CN110223517B (en) * 2019-06-20 2021-05-18 青岛科技大学 Short-term traffic flow prediction method based on space-time correlation
CN110428613A (en) * 2019-07-09 2019-11-08 中山大学 A kind of intelligent transportation trend prediction method of machine learning
CN110991698B (en) * 2019-11-07 2023-09-22 南通大学 Seasonal traffic flow grey prediction method based on mixed processing
CN110991913B (en) * 2019-12-09 2023-04-07 南京航空航天大学 Busy airport peak time congestion risk analysis method
CN111080110B (en) * 2019-12-09 2023-06-20 南京航空航天大学 Airport congestion risk analysis system
CN111259537B (en) * 2020-01-14 2023-03-31 中交第二公路勘察设计研究院有限公司 Road surface performance prediction method based on VAR (variable-value response) multivariate time sequence
CN111340068A (en) * 2020-02-10 2020-06-26 北京中旖新能源有限公司 Road traffic state prediction method and device
CN112712695B (en) * 2020-12-30 2021-11-26 桂林电子科技大学 Traffic flow prediction method, device and storage medium
CN112967013B (en) * 2021-02-20 2024-03-19 海南太美航空股份有限公司 Method, system and electronic device for determining take-off time of pre-opened flight
CN112836905B (en) * 2021-04-22 2021-08-03 中航信移动科技有限公司 Flight event occurrence time prediction method, device, equipment and storage medium
CN114063150B (en) * 2021-11-16 2023-05-23 西南科技大学 ML-KNN algorithm-based 'seismic source-station' speed model selection method
CN114997506B (en) * 2022-06-17 2024-05-14 西北师范大学 Atmospheric pollution propagation path prediction method based on link prediction
CN115346401B (en) * 2022-08-01 2023-09-12 南京航空航天大学 Low-altitude unmanned aerial vehicle monitoring and track prediction method
CN115620526B (en) * 2022-12-21 2023-03-07 中国民航大学 Airport land side traffic management system based on big data analysis and optimization method
CN116631196B (en) * 2023-07-25 2023-11-28 南京农业大学 Traffic road condition prediction method and device based on big data

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE3855163D1 (en) * 1988-05-11 1996-05-02 Alex Frauchiger METHOD FOR RESOLVING OR PREVENTING UNWANTED TRAFFIC JAMS
CN101154317A (en) * 2006-09-27 2008-04-02 株式会社查纳位资讯情报 Traffic state predicting apparatus
CN101438334A (en) * 2006-03-03 2009-05-20 因瑞克斯有限公司 Dynamic time series prediction of future traffic conditions
JP5214759B2 (en) * 2011-03-16 2013-06-19 三菱電機インフォメーションシステムズ株式会社 Congestion degree estimation device and congestion degree estimation program
CN104157139A (en) * 2014-08-05 2014-11-19 中山大学 Prediction method and visualization method of traffic jam
CN106355879A (en) * 2016-09-30 2017-01-25 西安翔迅科技有限责任公司 Time-space correlation-based urban traffic flow prediction method

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007140745A (en) * 2005-11-16 2007-06-07 Sumitomo Electric Ind Ltd Traffic congestion prediction system, traffic congestion factor estimation system, traffic congestion prediction method and traffic congestion factor estimation method
CN102346964A (en) * 2010-08-05 2012-02-08 王学鹰 Real-time jam prediction and intelligent management system for road traffic network area
CN204791180U (en) * 2015-07-17 2015-11-18 四川精工伟达智能技术股份有限公司 Wisdom tourism management system
CN105389980B (en) * 2015-11-09 2018-01-19 上海交通大学 Short-time Traffic Flow Forecasting Methods based on long short-term memory recurrent neural network
CN107230351B (en) * 2017-07-18 2019-08-09 福州大学 A kind of Short-time Traffic Flow Forecasting Methods based on deep learning
CN108229724B (en) * 2017-12-06 2020-12-22 华南理工大学 Short-term traffic data flow prediction method based on temporal-spatial information fusion

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE3855163D1 (en) * 1988-05-11 1996-05-02 Alex Frauchiger METHOD FOR RESOLVING OR PREVENTING UNWANTED TRAFFIC JAMS
CN101438334A (en) * 2006-03-03 2009-05-20 因瑞克斯有限公司 Dynamic time series prediction of future traffic conditions
CN101154317A (en) * 2006-09-27 2008-04-02 株式会社查纳位资讯情报 Traffic state predicting apparatus
JP5214759B2 (en) * 2011-03-16 2013-06-19 三菱電機インフォメーションシステムズ株式会社 Congestion degree estimation device and congestion degree estimation program
CN104157139A (en) * 2014-08-05 2014-11-19 中山大学 Prediction method and visualization method of traffic jam
CN106355879A (en) * 2016-09-30 2017-01-25 西安翔迅科技有限责任公司 Time-space correlation-based urban traffic flow prediction method

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Parking Traffic Jam Forecast System;Arunas Marma et al.;《IEEE》;20091123;全文 *
Transferability of a neuro-fuzzy river ice jam flood forecasting model;C. Mahabir et al.;《EL Sevier》;20070630;第48卷(第2007期);全文 *
基于深度学习的短时交通拥堵预测模型;高晓波;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20180415;全文 *
大型繁忙机场场面离场交通流拥堵特征分析;杨磊 等;《航空学报》;20151021;第37卷(第6期);全文 *
高速路网交通状态判别与预测的研究;崔玮;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20170215;全文 *

Also Published As

Publication number Publication date
CN108898838A (en) 2018-11-27

Similar Documents

Publication Publication Date Title
CN108898838B (en) Method and device for predicting airport traffic jam based on LSTM model
Zhao et al. Truck traffic speed prediction under non-recurrent congestion: Based on optimized deep learning algorithms and GPS data
Chen et al. Uncertainty in urban mobility: Predicting waiting times for shared bicycles and parking lots
Chen et al. A dynamic bus‐arrival time prediction model based on APC data
De Palma et al. METROPOLIS: Modular system for dynamic traffic simulation
Sundaram et al. Simulation-based dynamic traffic assignment for short-term planning applications
CN111653088A (en) Vehicle driving quantity prediction model construction method, prediction method and system
Zhao et al. A comparative study of parking occupancy prediction methods considering parking type and parking scale
Choi et al. Artificial neural network models for airport capacity prediction
Li et al. Deep learning based parking prediction on cloud platform
Xu et al. Trip-oriented travel time prediction (TOTTP) with historical vehicle trajectories
Gao et al. Aggressive driving saves more time? Multi-task learning for customized travel time estimation.
Seliverstov et al. Modeling of megalopolis traffic flows with the introduction of a new line of water intercity passenger transport
Zhu et al. Prediction of daily entrance and exit passenger flow of rail transit stations by deep learning method
CN107220724A (en) Passenger flow forecast method and device
Paliwal et al. To each route its own ETA: A generative modeling framework for ETA prediction
Hu et al. Mass rapid transit system passenger traffic forecast using a re-sample recurrent neural network
Yang et al. Dynamic origin-destination matrix estimation based on urban Rail transit AFC data: deep optimization framework with forward passing and backpropagation techniques
Penenko et al. Numerical study of a direct variational data assimilation algorithm in Almaty city conditions
Pang et al. Replicating urban dynamics by generating human-like agents from smartphone GPS data
Xu et al. A taxi dispatch system based on prediction of demand and destination
Kumar et al. Analytical approach to identify the optimum inputs for a bus travel time prediction method
Shen et al. A short-term traffic flow intelligent hybrid forecasting model and its application
Ma et al. Short-term subway passenger flow prediction based on gcn-bilstm
Lingqiu et al. A LSTM based bus arrival time prediction method

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
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20200828

Termination date: 20210803

CF01 Termination of patent right due to non-payment of annual fee