CN112182059B - High-order analysis method for flight delay characteristics - Google Patents

High-order analysis method for flight delay characteristics Download PDF

Info

Publication number
CN112182059B
CN112182059B CN202010844144.0A CN202010844144A CN112182059B CN 112182059 B CN112182059 B CN 112182059B CN 202010844144 A CN202010844144 A CN 202010844144A CN 112182059 B CN112182059 B CN 112182059B
Authority
CN
China
Prior art keywords
order
flight
sequence
delay
node
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
CN202010844144.0A
Other languages
Chinese (zh)
Other versions
CN112182059A (en
Inventor
杜文博
曹先彬
张明远
李思远
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.)
Beihang University
Original Assignee
Beihang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beihang University filed Critical Beihang University
Priority to CN202010844144.0A priority Critical patent/CN112182059B/en
Publication of CN112182059A publication Critical patent/CN112182059A/en
Application granted granted Critical
Publication of CN112182059B publication Critical patent/CN112182059B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2458Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2462Approximate or statistical queries
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/18Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40Business processes related to the transportation industry

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Probability & Statistics with Applications (AREA)
  • Mathematical Physics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Mathematics (AREA)
  • Software Systems (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Evolutionary Biology (AREA)
  • Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Operations Research (AREA)
  • Computational Linguistics (AREA)
  • Algebra (AREA)
  • Fuzzy Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a high-order analysis method for flight delay characteristics, and belongs to the technical field of civil aviation delay analysis. According to the method, flight high-order characteristics are extracted from real flight data, high-order dependency items are extracted and identified through rules, high-order edges are set through edge reconstruction, a high-order network model representing the high-order dependency is constructed by combining weather and flight information, flight delay is predicted, the problem that the high-order dependency of the flight delay is lost when the existing model is constructed is solved, and prediction accuracy is improved.

Description

High-order analysis method for flight delay characteristics
Technical Field
The invention belongs to the technical field of civil aviation delay analysis, and aims to explore a high-order dependency relationship of flight delay from original flight data, further combine flight weather information, construct a high-order network model containing the weather information, and further predict the flight delay.
Background
The flight delay refers to a situation in which the flight landing time is delayed by more than 15 minutes from the scheduled landing time or the flight is cancelled. Generally, when an airplane executes continuous flight tasks, the executed flight records form a flight chain, the delay of a preceding flight can cause the delay of a subsequent flight, along with the development of economy and further globalization of economy in China, the air transportation industry develops rapidly, the problem of flight delay is increasingly highlighted, and the flight delay brings high time cost to passengers and also causes serious resource waste and huge economic loss. It is therefore necessary to investigate flight delays and make predictions.
In 2013, a queuing analysis and network decomposition model is established in Pyrgiotis, Malone and the like to research the phenomenon of delay spread in a large airport network. The model calculates the delay of a single airport through repeated iteration calculation between two main components, namely a Queuing Engine (QE) and another Delay Propagation Algorithm (DPA) for updating flight schedule and demand rate of all airports in the model. Compared with the original research, the model considers the randomness AND the time-varying characteristic of airport demands AND capacity, can accurately model the traffic jam condition of a flight transportation network, AND can describe the development trend of an airport chain from a macroscopic perspective. However, this model also has two significant disadvantages: firstly, delay caused by road congestion is not considered, and from a practical point of view, the model cannot simulate the reaction of an airline company to congestion and cannot capture other delay sources. Secondly, the model operation result is similar to the actual delay distribution, but the magnitude of the model operation result has larger difference, so that the model operation result is not suitable for practical use. Lu et al 2014 used a combined economics model, which evaluates the association degree of new york airports and surrounding airports through a large amount of historical data, and a SWAC model of FAA, which evaluates the influence degree of three airports in new york on NAS delay by tracking the working performance of a daily flight system through simulating flight. The actual value of the model and the simulation are mutually verified, and the cause of the congestion phenomenon is analyzed from different angles by combining a measurement economics model and a simulation model. However, although the model captures the delay propagation process of the airplane, due to the characteristics of the economic metrology model, the influence of factors such as airspace control on the delay of the airplane is rough, and for airports sensitive to the factors, the difference between the simulation result of the model and the actual situation is large. Kafle et al 2016 established a joint discrete-continuous metrology economic model based on delay data calculated by a classical model, and revealed the influence of various influencing factors on the initiation and progress of propagation delay by using a Heckman two-stage model. Wang et al empirically counts the American flight data in 2017 to obtain a cumulative distribution function (CCDF) and a delay transfer function of the flight data. CCDF of 14 major airlines was classified into two categories: one type of distribution follows a transferred power-law distribution (transferred power-law), the other type of distribution follows an exponential truncated power-law distribution (amplified truncated transferred power-law), and other observable factors are converted into parameters describing delayed propagation characteristics by using a mean value domain method, so that the feasibility of the model in actual dynamic flight planning is verified. However, the delay probability Q in this method is not known in advance, and needs to be determined through data mining and macroscopic statistics, so the initial effect of this model is poor, and a certain amount of data accumulation is needed to play a role.
The research results are only representative methods in scientific exploration related to delay and delay spread at home and abroad. There are also many other ways to solve the delay and delay propagation problems. In general, existing research in the field of flight delay is mainly based on theories of complex networks, regression analysis, measurement economy and the like to establish models, and many implicit assumptions are that networks are of low-order markov attributes, but in the existing research, on one hand, only specific airport conditions are researched from flight chain angles formed by connecting different airports, coupling effects of delay among different flights are ignored, on the other hand, the researched network models are more focused on describing inherent topological structures of flight transportation networks, and dynamic flight delay is less researched. With the development of civil aviation, air transportation networks are increasingly complex, and data in complex systems may exhibit up to five-order correlation due to the nature of flight delay propagation. In this case, the conventional network model may be too simplified to truly represent the actual data, which may affect the accuracy of the analysis result.
Disclosure of Invention
Aiming at the problems, the invention provides a high-order analysis method for flight delay characteristics, and relates to a flight high-order network (HON) model.
The invention discloses a high-order analysis method for flight delay characteristics, which comprises the following steps:
step 1: and according to the historical data of the airport flights, identifying the high-order dependency relationship in the data and finishing the extraction of the data rule.
Step 2: and (4) reconstructing the edge line on the basis of the step 1, and ensuring that the high-order node has a transmitting edge.
And step 3: the high-level network model integrates flight delay information and weather information.
And 4, step 4: and inputting continuous prediction sets containing delay information and weather information into a high-order model, wherein the high-order model gives delay conditions of future nodes.
The invention has the advantages that:
1. according to the high-order analysis method for the flight delay characteristic, a high-order model capable of describing the high-order dependency relationship in the aviation network more accurately is constructed, delay information is combined with flight information weather information, a high-order network model containing the delay information and the weather information is constructed, and then flight delay is predicted.
2. Flight delays have the characteristic of high dynamic, the flight delays are propagated along routes in a complex network, multiple factors are mutually coupled, and the propagation characteristics of the flight delays cannot be accurately described only by modeling the delays of a single airport or only by researching a static global network; compared with the traditional network model, the high-order analysis method for flight delay characteristics can more truly represent actual data characteristics, and further improve the accuracy of analysis results.
3. The high-order analysis method for flight delay characteristics uses a high-order model to predict flight delay conditions, provides a certain optimization direction for flight scheduling work, and provides a certain theoretical basis for more effectively reducing loss caused by flight delay. In addition, the high-order network model constructed by the invention describes a flight delay propagation mechanism and provides reference for the layout construction of an airport.
4. The high-order analysis method for flight delay characteristics has strong robustness to data set differences, and is still suitable for aviation networks of different scales and different network structures, such as Chinese aviation networks, European aviation networks, American aviation networks and the like.
Drawings
FIG. 1 is a flow chart of a high-level analysis method for flight delay characteristics according to the present invention.
FIG. 2 is a flow chart of a data rule extraction method in the high-order flight delay characteristic analysis method of the invention.
FIG. 3 is a flow chart of a method for reconstructing a boundary in a high-order flight delay characteristic analysis method according to the present invention.
Fig. 4a is a schematic diagram of a first-order sequence addition mode in the sideline reconstruction process of the high-order analysis method for flight delay characteristics of the invention.
Fig. 4b is a schematic diagram of a high-order sequence adding mode in the sideline reconstruction process of the high-order analysis method for flight delay characteristics according to the present invention.
Fig. 4c is a schematic diagram of a flight sequence redirection mode in the sideline reconstruction process of the high-order analysis method for flight delay characteristics.
FIG. 4d is a schematic diagram of a high-order sequence correction mode in the sideline reconstruction process of the high-order analysis method for flight delay characteristics according to the present invention.
FIG. 5 is a schematic diagram of a high-level network model established by the high-level analysis method for flight delay characteristics according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The invention relates to a high-order analysis method for flight delay characteristics, which comprises the following specific steps as shown in figure 1:
step 1: and according to the historical data of the airport flights, identifying the high-order dependency relationship in the data and finishing the extraction of the data rule.
As shown in fig. 2, in building a first order network, the order of dependence in the identification data is usually ignored, and is often just a pair of connections in the statistical data. The main purpose of rule extraction is to identify a high-order dependency relationship in data, and due to different orders of flight chains, when rule extraction is performed, a high-order dependency term needs to be accurately identified, and then a flight high-order network model can be accurately constructed. Rule extraction requires the following principles to be satisfied:
higher order dependent terms can significantly affect the probability distribution. For example: assuming that the probability distributions for flights to other airports are the same whether from beijing to tianjin or from shanghai to tianjin, then it is stated that there is no second order dependency (but there may be third order or higher dependencies). Only if a more significant probability distribution difference occurs, it can be shown that an effective high-order dependency relationship exists.
Rule extraction results should distinguish between different flight sequences containing similar flight sequence segments. In the actual flight sequence, the two sequences may have part of the same flight sequence, and rule extraction needs to identify the differences of similar flights, so as to completely characterize the network structure.
The rule extraction should identify general data, the flight high-order network model describes a more generalized network structure, and occasional flight sequences with low occurrence frequency should be filtered out, so as to improve the effectiveness of rule extraction.
Step 101: with a single aircraft as the subject, the aircraft will perform multiple flight missions in succession with only brief loading (human) and unloading (human) intervals between them. Therefore, discrete flights can be connected in series in time sequence to form a plurality of flight chains mainly taking airplanes as main bodies, the airplane flies from an airport A to an airport B and is called a flight, and the airplane takes off and lands for multiple times and flies through A, B, C, D to form a flight chain, wherein the total number of 4 airports is A, B, C, D. Flight sequences refer to a specific flight chain, for example, Beijing- > Tianjin- > Dalian and Chengdu- > Tianjin- > Shanghai are two different flight chains, which can also be referred to as two different flight sequences.
Counting the occurrence times of the flight sequence in the original flight data, for example, the original flight data is as follows: flight chain 1: beijing- > Tianjin- > Dalian, flight chain 2: chengdu- > Tianjin- > Shanghai, … …. Counting the occurrence times of flight sequences in the original flight data: tianjin- > da lian ═ 71, tianjin- > shanghai ═ 68, tianjin- > Chongqing ═ 43, Beijing- > tianjin- > da lian 31, Beijing- > tianjin- > shanghai ═ 2, Shenzhen- > tianjin- > da lian ═ 27, Shenzhen- > tianjin- > shanghai ═ 25.
Step 102: creating a flight sequence table, counting probability distribution of the flight sequence, and checking first-order node probability distribution; the probability distribution result of the flight sequence is as follows:
P(Xt+1|Xt=mi)={mj:a%
wherein, XtAn airport for which the flight sequence arrives at time t; xt+1For the airport where the flight sequence arrives at time t +1, m is the airport set, m ═ m1,m2,…,mi,…];P(Xt+1|Xt=mi)={mj: a% indicates that the flight sequence arrives at a certain airport m at time tiThen, the time t +1 arrives at the airport mjThe probability of (a%); after the flight probability is integrated and countedThe results are as follows:
Figure BDA0002642464670000041
the above formula indicates that the flight sequence arrives at a certain airport m at time tiThen, the time t +1 arrives at the airport mj、mk、mlThe probabilities of the remaining airports are a%, b%, c%, … …, respectively.
Step 103: adding flight sequences and counting the probability distribution change of subsequent nodes;
and (3) counting the probability distribution of the second-order nodes, wherein the statistical result is as follows:
Figure BDA0002642464670000042
Figure BDA0002642464670000043
wherein, Xt-1For the airport where the flight sequence arrives at time t-1,
Figure BDA0002642464670000044
indicating that the flight sequence arrived at airport m at time t-1aAnd reaches m at time tiThen, time t +1 reaches mj、mlAnd the probabilities of the remaining airports.
Depending on the data characteristics, there may be more nodes and more sequences, and the probability distribution variation of the higher order nodes can be counted in the above manner.
Step 104: and measuring the distance between the probability distributions before and after the new adding flight sequence by using Kullback-Leibler divergence, and further judging whether the new adding flight sequence is effective, wherein the relative entropy is a method for describing the difference between two probability distributions P and Q, and can be used for measuring the distance between the probability distributions before and after the new adding flight sequence. Since the addition of the flight sequence will change the order of the sequence, the flight sequence may be referred to as a high-order dependent term, the more significant the influence of the high-order dependent term is, the larger the change of the probability distribution is, and the corresponding relative entropy is, whereas if the influence of the high-order dependent term is weaker, the smaller the change of the probability distribution is, the smaller the corresponding relative entropy is, and when the probability distributions are the same, the obtained relative entropy value is 0. Therefore, a proper threshold value can be set, and when the relative entropy of the probability distribution after the high-order dependent item is added is larger than the threshold value, the high-order dependent item is judged to be more remarkable to the probability distribution and is an effective high-order dependent item. The divergence calculation formula is as follows:
Figure BDA0002642464670000051
Figure BDA0002642464670000052
where P (x), Q (x) are two probability distributions of the random variable x, and KL (P | | Q) is the relative entropy of the probability distribution P and the probability distribution Q.
And checking the probability distribution change of the subsequent nodes from the low-order flight through the judgment mode, if the probability distribution change is obvious, meeting the validity principle, namely preliminarily explaining that the flight chain has the high-order relation, then checking the universality principle and the similarity principle, if the universality principle and the similarity principle are met, continuously checking the subsequent nodes, and otherwise, explaining that the extraction of the high-order dependency relation of the current flight sequence is finished. And finally, repeating the steps through recursive iteration by the algorithm until the whole data set is extracted, and finishing rule extraction.
In the step 1, the process of combining the actual data is as follows:
in step 101, flight data is counted, and the original flight data is: flight chain 1- > Beijing- > Tianjin- > Dalian, flight chain 2- > Chengdu- > Tianjin- > Shanghai, … …. And then, counting the occurrence times of the flight sequences in the original flight data: tianjin- > da lian ═ 71, tianjin- > shanghai ═ 68, tianjin- > Chongqing ═ 43, Beijing- > tianjin- > da lian 31, Beijing- > tianjin- > shanghai ═ 2, Shenzhen- > tianjin- > da lian 27, and Shenzhen- > tianjin- > shanghai ═ 25, the statistical results are shown in table 1.
TABLE 1 partial flight chain frequency statistics
Figure BDA0002642464670000053
Figure BDA0002642464670000061
In step 102, a flight sequence table is newly created, statistics is performed on probability distribution of the flight sequence, first-order node probability distribution is firstly counted, and the statistical result is as follows:
Figure BDA0002642464670000062
in step 103, adding flight sequences and checking the probability distribution change of subsequent nodes, and counting the probability distribution of second-order nodes, wherein the counting result is as follows:
Figure BDA0002642464670000063
Figure BDA0002642464670000064
through comparison, after Beijing is added as a high-order dependent term of Tianjin, the probability distribution is obviously changed compared with that without the high-order dependent term, namely the high-order dependent term exists, and the high-order dependent term is newly added to the flight sequence after being checked according to the universality principle and the similarity principle and meeting the conditions. And after adding Shenzhen as the higher order dependency term of Tianjin, the probability distribution does not change obviously, that is, the higher order dependency term does not exist, compared with the case that the higher order dependency term is not added. And finally, traversing all data in the data set through multiple iterations to complete the high-order dependency relationship extraction of the data set, and providing a data basis for the subsequent edge reconstruction and construction of a high-order model.
Step 2: and (3) reconstructing an edge on the basis of the step 1 to ensure that the high-order node has an incoming edge, as shown in the figure 3.
According to the conventional first-order network method, although data can be directly mapped as an edge connecting two airport nodes, when a rule of table dependent variables coexists, if only the last airport node of each path is taken as a target node, each edge points to one first-order node, and there is no edge in a high-order node, so that edge reconstruction is required to ensure that the high-order node has an incoming edge, as shown in fig. 4a to 4d, the specific method is as follows:
step 201: and (3) converting all the first-order sequences extracted according to the rule in the step (1) into edges according to a traditional first-order network method, wherein each first-order sequence corresponds to a weighted edge. For example: if the flight sequence is Beijing- > Tianjin, adding Beijing nodes and Tianjin nodes in a complex network, adding continuous edges from Beijing to Tianjin, wherein each first-order sequence corresponds to a weighted edge, as shown in FIG. 4a, after the rule extraction in the step 1, forming a plurality of high-order and low-order dependency relationships, converting the first-order sequences of Beijing- > Tianjin, Chengdu- > Tianjin and the like into edges according to a traditional first-order network method, wherein each first-order sequence corresponds to a weighted edge.
Step 202: similarly converting the high-order sequence into an edge according to a traditional first-order network method, adding a high-order node to the network, similarly converting the high-order sequence such as Tianjin | Beijing- > Dalian, Tianjin | Chengdu- > Shanghai, Tianjin | Chengdu- > Guangzhou and the like into an edge according to a traditional first-order network method, and adding high-order nodes such as Tianjin | Chengdu, Tianjin | Beijing and the like to the network, as shown in FIG. 4 b;
step 203: flight sequence redirection.
And (3) re-wiring the high-order nodes, and re-orienting the high-order flight sequence extracted by the first-order nodes according to the rule in the step (1) so that the high-order sequence nodes and the low-order sequence nodes meet the same rule, namely, edges containing the same airport node are not directly connected with incoming edges from the same airport node, as shown in fig. 4 c. For example: reorienting the first-order nodes such as Beijing- > Tianjin, Chengdu- > Tianjin and the like according to the high-order flight sequence extracted according to the rule, orienting the Beijing- > Tianjin into Beijing- > Tianjin | Beijing- > Dalian, and orienting the Chengdu- > Tianjin into Chengdu- > Tianjin | Chengdu- > Shanghai, so that the high-order sequence nodes and the low-order sequence nodes meet the same rule.
Step 204: and (5) correcting a high-order sequence.
After connecting airport nodes are extracted according to rules, in order to retain high-order information as much as possible, edges constructed according to flight sequences extracted according to the rules should preferentially point to nodes with the highest sequence length, as shown in fig. 4d, because of the limitation of the rule extraction algorithm, the last stage of rule extraction must be the actual airport as a node, so the last node of rule extraction must be a stage node, but in order to retain high-order information as much as possible, without losing important information of previous steps, the last stage needs to be relocated to an existing high-order node, for example: the guangzhou of tianjin | chengdu- > guangzhou is a first-order node, but if there is a high-order node of the guangzhou involved, the guangzhou can be labeled as guangzhou | tianjin chengzhou, and more high-order information is retained than if the guangzhou node alone is used as a last-order node.
And (3) extracting the data rule in the step (1), and reconstructing the data rule with the edge in the step (2) to finally obtain the high-order network model shown in the figure 5.
And step 3: the high-level network model integrates flight delay information and weather information.
Step 301: adding H (high) and L (low) to the tail of the airport node in the step 1 to express the delay degree of the airport, inputting the node containing the delay into a high-order network, for example, if the original airport node is A, the arrival delay is heavy delay, and then the new node is AL to replace the original node, thereby forming a high-order network model with delay information.
Step 302: dividing flight data (Weather enhanced Traffic Index) WITI Weather influence Traffic factors into four types I, II, III, IV and the like according to Weather categories, combining the Weather influence Traffic factors with original airport nodes, for example, if the original airport node fusing delay information is AL, the Weather influence Traffic factor is III, then the new node is AL III, and further inputting a high-order network model to construct the high-order network model containing the Weather information, wherein the specific method comprises the following steps:
the WITI calculation formula is as follows:
Figure BDA0002642464670000071
wherein n and m respectively represent the number of rows and columns of the longitude and latitude network, k represents the time k, and Ti,j(k) Aircraft flow, W, representing a grid of i rows and j columnsi,j(k) Representing the weather category of the grid of i rows and j columns.
The flight delay is predicted for two hours by inputting a linear autoregressive model in combination with past delay information, an airspace is processed according to a longitude and latitude grid, weather category of radar data of a weather bureau is divided into sixteen categories, four categories and the like, wherein the sixteen categories are as follows: wind, rainstorm, snowstorm, cold tide, strong wind, sand storm, high temperature, drought, thunder, hail, frost, fog, haze, road icing, thunderstorm strong wind, forest fire; the fourth and the like are respectively: the weather type is further divided into the following seven categories of 1 to 6 which respectively represent light rain, light rain to medium rain, medium rain to large rain, heavy rain, extra heavy rainstorm and hail on the basis of dividing the weather bureau, wherein the seven categories comprise a blue early warning signal-IV (general), a yellow early warning signal-III (serious), an orange early warning signal-II (serious) and a red early warning signal-I (particularly serious), and 0 represents no rainfall.
The seven weather categories are used as weights to weight the aircraft flow, the WITI of any grid point can be finally obtained, obviously, the larger the WITI is, the larger the influence of the weather on the flight is, therefore, the WITI on the way can be counted and averaged on the basis of the known start and end airports, then, the influence of the weather on the route can be used, and after the WITI counting and the normalization processing are completed, the weather categories are divided into four categories I, II, III and IV by taking 0.00, 0.25, 0.50, 0.75 and 1.00 as thresholds. And finally, inputting the node containing the WITI information into a high-order network, and reconstructing a high-order network model with weather data.
And 4, step 4: and predicting flight delay conditions by the high-level network model.
And inputting the continuous prediction sets containing delay information and weather information into the high-order network model, so that the delay condition of the future node can be given by the high-order network model.

Claims (5)

1. A high-order analysis method for flight delay characteristics is characterized by comprising the following steps: the method comprises the following steps:
step 1: according to the airport flight historical data, identifying a high-order dependency relationship in the data, and finishing data rule extraction:
the data rule extraction method comprises the following steps:
A. counting the number of times of flight sequence in the original flight data;
B. newly building a flight sequence table, counting the probability distribution of the flight sequence, and counting the probability distribution of first-order nodes;
C. adding flight sequences, checking the probability distribution change of subsequent nodes, and counting the probability distribution of second-order nodes;
D. measuring the distance between probability distributions before and after the new high-order dependent item by using Kullback-Leibler divergence, and further judging whether the new flight sequence is effective;
step 2: and (3) reconstructing a boundary line on the basis of the step 1 to ensure that a high-order node has an incoming boundary:
the edge line reconstruction method comprises the following steps:
step a: converting all the first-order sequences into edges according to a traditional first-order network method, wherein each first-order sequence corresponds to a weighted edge;
step b: the high-order sequence is also converted into an edge, and when needed, a high-order node is added to the network;
step c: redirecting the flight sequence, namely rewiring the high-order nodes, and redirecting the flight sequence extracted by the first-order nodes according to the rule, so that the high-order sequence nodes and the low-order sequence nodes meet the same rule; that is, the edge containing the same airport node is not directly connected with the incoming edge from the same airport node;
step d: correcting a high-order sequence; after the connecting airport node is extracted according to the rule, the edge constructed according to the flight sequence extracted according to the rule preferentially points to the node with the highest sequence length;
and step 3: the high-order network model integrates flight delay information and weather information;
and 4, step 4: and inputting continuous prediction sets containing delay information and weather information into a high-order model, wherein the high-order model gives delay conditions of future nodes.
2. A method for high-level analysis of flight delay characteristics as claimed in claim 1, wherein: the first-order node and second-order node probability distribution statistical method in the step B is as follows:
first-order node probability distribution statistics:
Figure FDA0003597710030000011
equation (1) indicates that the flight sequence arrives at a certain airport m at time tiThen, the time t +1 arrives at the airport mj、mk、mlThe probabilities of the other airports are a%, b%, c%, … … respectively;
and (3) second-order node probability distribution statistics:
Figure FDA0003597710030000021
Figure FDA0003597710030000022
wherein, Xt-1For an airport where a flight sequence arrives at time t-1, equations (2) and (3) respectively represent the arrival of the flight sequence at the airport m at time t-1aAnd mbAnd reaches m at time tiThen, time t +1 reaches mj、mkProbabilities of the remaining airports.
3. A method for high-level analysis of flight delay characteristics as claimed in claim 1, wherein: the divergence calculation formula in the step D is as follows:
Figure FDA0003597710030000023
Figure FDA0003597710030000024
where P (x), Q (x) are two probability distributions of random variable x, and KL (P | | Q) is the relative entropy of probability distribution P and probability distribution Q.
4. A method for high-level analysis of flight delay characteristics as claimed in claim 1, wherein: the specific method of the step 3 comprises the following steps:
step 1): adding H (high) and L (low) to the tail of the original airport node to express the delay degree of the airport, and inputting the node containing delay into a high-order network to form a high-order network model with delay information;
step 2): dividing the WITI Weather influence Traffic factors of flight data (Weather Impacted Traffic Index) into four equal parts, combining the four equal parts with the original airport nodes, and further inputting a high-order network model to construct a high-order network model containing Weather information;
step 3): and predicting flight delay conditions by the high-level network model.
5. The method of claim 4, wherein the flight delay characteristic is analyzed by the following steps: the specific method of the step 3) is as follows: the weather categories are divided into the following seven categories from 1 to 6, which are respectively expressed as light rain, light rain to medium rain, medium rain to heavy rain, extra heavy rain and hail, and 0 represents no rainfall; weighting the aircraft flow by taking seven types of weather categories as weights, and finally obtaining the WITI of any lattice point, wherein obviously, the larger the WITI is, the larger the influence of the weather on the flight is, therefore, on the basis of the known start and end airports, the WITI on the way can be counted and averaged, then the air route can be influenced by the weather, and after the WITI counting and the normalization processing are completed, the weather categories are divided into four categories I, II, III and IV by taking 0.00, 0.25, 0.50, 0.75 and 1.00 as thresholds; and finally, inputting the node containing the WITI information into a high-order network, and reconstructing a high-order network model with weather data.
CN202010844144.0A 2020-08-20 2020-08-20 High-order analysis method for flight delay characteristics Active CN112182059B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010844144.0A CN112182059B (en) 2020-08-20 2020-08-20 High-order analysis method for flight delay characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010844144.0A CN112182059B (en) 2020-08-20 2020-08-20 High-order analysis method for flight delay characteristics

Publications (2)

Publication Number Publication Date
CN112182059A CN112182059A (en) 2021-01-05
CN112182059B true CN112182059B (en) 2022-05-27

Family

ID=73923955

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010844144.0A Active CN112182059B (en) 2020-08-20 2020-08-20 High-order analysis method for flight delay characteristics

Country Status (1)

Country Link
CN (1) CN112182059B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113554899B (en) * 2021-07-30 2022-06-03 中国民用航空总局第二研究所 Weather influence air traffic degree analysis method, device, equipment and storage medium
CN115689073B (en) * 2023-01-04 2023-05-23 北京航空航天大学 Flight delay prediction method and system based on high-order network

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107038211A (en) * 2017-02-28 2017-08-11 大连理工大学 A kind of paper impact factor appraisal procedure based on quantum migration
CN108519988A (en) * 2018-02-27 2018-09-11 北京航空航天大学 Aviation based on co-integration relationship is delayed causality network construction method
CN109325062A (en) * 2018-09-12 2019-02-12 哈尔滨工业大学 A kind of data dependence method for digging and system based on distributed computing
CN111144631A (en) * 2019-12-19 2020-05-12 南京航空航天大学 Flight delay real-time probability prediction method based on Bayesian network algorithm

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11024182B2 (en) * 2018-12-17 2021-06-01 The Boeing Company System and method to forecast flight delay based on real-time data

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107038211A (en) * 2017-02-28 2017-08-11 大连理工大学 A kind of paper impact factor appraisal procedure based on quantum migration
CN108519988A (en) * 2018-02-27 2018-09-11 北京航空航天大学 Aviation based on co-integration relationship is delayed causality network construction method
CN109325062A (en) * 2018-09-12 2019-02-12 哈尔滨工业大学 A kind of data dependence method for digging and system based on distributed computing
CN111144631A (en) * 2019-12-19 2020-05-12 南京航空航天大学 Flight delay real-time probability prediction method based on Bayesian network algorithm

Also Published As

Publication number Publication date
CN112182059A (en) 2021-01-05

Similar Documents

Publication Publication Date Title
Ding Predicting flight delay based on multiple linear regression
Lieder et al. Scheduling aircraft take-offs and landings on interdependent and heterogeneous runways
CN112182059B (en) High-order analysis method for flight delay characteristics
Chen et al. Chained predictions of flight delay using machine learning
Delgado et al. Cruise speed reduction for ground delay programs: A case study for San Francisco International Airport arrivals
CN105844346A (en) Flight delay prediction method based on ARIMA model
Baspinar et al. Analysis of delay characteristics of European air traffic through a data-driven airport-centric queuing network model
CN104299454B (en) Multi-airport collaborative delivery system flight sort method under a kind of inclement weather
CN110570693A (en) Flight operation time prediction method based on reliability
CN116468186B (en) Flight delay time prediction method, electronic equipment and storage medium
CN104573873A (en) Airport terminal departure passenger traffic volume prediction method based on fuzzy decision-making tree
CN111160612A (en) Off-site flight delay analysis and prediction method based on weather influence
Ai et al. A deep learning approach to predict the spatial and temporal distribution of flight delay in network
CN110796315B (en) Departure flight delay prediction method based on aging information and deep learning
Yang Practical method for 4-dimentional strategic air traffic management problem with convective weather uncertainty
CN110991913B (en) Busy airport peak time congestion risk analysis method
CN113706931B (en) Airspace flow control strategy recommendation method and device, electronic equipment and storage medium
CN114781704A (en) Flight delay prediction method based on station-passing flight guarantee process
Schultz et al. Classification of weather impacts on airport operations
Kleinman et al. SPSA/SIMMOD optimization of air traffic delay cost
Yang et al. Stochastic scheduling of ground movement problem integrated with taxiway routing and gate/stand allocation
Sheth et al. Development of a probabilistic convective weather forecast threshold parameter for flight-routing decisions
Baspinar et al. Large scale data-driven delay distribution models of european air traffic flow network
Nazeri et al. Mining aviation data to understand impacts of severe weather on airspace system performance
CN114742458A (en) Airport flight operation toughness assessment and prediction method in severe weather

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