CN110659773A - Flight delay prediction method based on deep learning - Google Patents
Flight delay prediction method based on deep learning Download PDFInfo
- Publication number
- CN110659773A CN110659773A CN201910870557.3A CN201910870557A CN110659773A CN 110659773 A CN110659773 A CN 110659773A CN 201910870557 A CN201910870557 A CN 201910870557A CN 110659773 A CN110659773 A CN 110659773A
- Authority
- CN
- China
- Prior art keywords
- data
- delay
- time
- network
- flight
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 15
- 238000013135 deep learning Methods 0.000 title claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims abstract description 8
- 238000000605 extraction Methods 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 14
- 230000010006 flight Effects 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000002474 experimental method Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 5
- 239000013598 vector Substances 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 125000002252 acyl group Chemical group 0.000 claims description 2
- 238000009795 derivation Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims description 2
- 238000002372 labelling Methods 0.000 claims 1
- 238000007726 management method Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a flight delay prediction method based on deep learning. The invention comprises the following steps: 1: processing the data set and inputting the processed data set into a network model; 2: inputting the data set processed in the step 1 into an LSTM network in a time sequence matrix form to extract time sequence characteristics on a time sequence; 3: inputting the time sequence characteristics extracted by the LSTM network in the step 2 into a ResNet network for further characteristic extraction; 4: predicting flight delay states and calculating prediction accuracy; namely: no delay, delay of 0 to 45 minutes, delay of 45 to 90 minutes, delay of more than 90 minutes. The invention adds the time sequence characteristics of flight delay prediction into a model, uses samples with more characteristic quantity as input, adds a network with a residual error structure to deepen the number of network layers, and keeps the characteristics in a shallow network on the basis of deepening the network, so that the characteristics are not lost in the deep network, and the accuracy of the flight delay prediction model is improved.
Description
Technical Field
The invention relates to the technical field of flight delay prediction, in particular to a flight delay prediction method based on deep learning, which is used for flight delay prediction.
Background
With the rapid development of the civil aviation industry, the flight delay gradually becomes a hot topic. The cause of the flight delay is difficult to explain because it may be due to multiple factors, such as weather causes, departure or destination airport management causes, airline management causes, preceding flight causes, traveler causes, etc., or even a superposition of multiple causes or a linkage effect of multiple causes.
For airports, especially for large-scale airports, flight delay will cause very limited resource allocation plans such as airway, runway, airport facilities and the like to be disturbed, and passengers to be detained, so that safety, operation and scheduling pressure of the airport is remarkably increased, and the satisfaction degree of airlines or passengers is reduced; for an airline company, the operating profit of the airline company depends heavily on each airplane to operate strictly according to a schedule, and each flight delay causes the increase of operation, maintenance and labor costs, and may cause all subsequent operation plans to be disturbed, subsequent flights are delayed continuously or are cancelled forcibly, and various costs are further increased; for passengers, flight delay is the situation which is most undesirable to be met in the trip, so that time and energy are lost, and subsequent trips are affected; for insurance companies, the research and prediction of flight delay also have important significance for pricing and operation of dangerous species such as travel insurance, flight delay insurance and the like. The establishment of more accurate and efficient flight delay prediction models contributes to the profound value and significance of management, decision making and selection in the aspects of airports, airlines, insurance companies and passengers.
The study of flight delay prediction by relevant scholars is mainly carried out from the following angles:
(1) delay prediction based on an airport visual angle, namely taking statistical data as the main point for airport take-off and landing data, analyzing the take-off and landing data of flights and predicting delay conditions;
(2) delay prediction based on the view angle of the airline company, namely analyzing delay conditions of all flights subordinate to the airline company and predicting future delay;
in fact, there are many factors causing flight delay, such as departure or destination airport management reasons, airline regulatory reasons, preorder flight reasons, passenger reasons, etc., and even multiple reasons are superimposed or multiple reasons form a linkage effect. Most of the existing researches at home and abroad are only researched from the perspective of a certain or a small amount of aspects, but the research on the flight delay time sequence perspective is less, so that the situation of flight delay cannot be studied by comprehensive factors, and the prediction result is not accurate enough. The following problems still exist in the research in the field of flight delay prediction at home and abroad: (1) the existing method for predicting flight delay through a neural network has a simple network structure, the depth of the network layer is only a few layers, and the characteristics cannot be effectively extracted, so that the prediction result is not accurate enough.
Disclosure of Invention
The invention aims to solve the defects of the existing flight delay prediction method, and provides a flight delay prediction method based on deep learning to realize a flight delay prediction method with higher accuracy.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a flight delay prediction method based on deep learning comprises the following steps:
step 1: the data set is processed and the processed data set is input into a network model.
Step 2: and (3) inputting the data set processed in the step (1) into an LSTM network in a time sequence matrix form to extract time sequence characteristics on a time sequence.
And step 3: and (3) inputting the timing characteristics extracted by the LSTM network in the step (2) into a ResNet network for further characteristic extraction.
And 4, step 4: and predicting the flight delay state and calculating the prediction accuracy.
The invention has the following beneficial effects:
the invention adds the time sequence characteristics of flight delay prediction into a model, uses samples with more characteristic quantity as input, adds a network with a residual error structure to deepen the number of network layers, and keeps the characteristics in a shallow network on the basis of deepening the network, so that the characteristics are not lost in the deep network, and the accuracy of the flight delay prediction model is improved.
Drawings
FIG. 1 is an overall flow chart of the present invention, the overall flow is divided into 4 modules, the first module is a data input module, the module inputs the preprocessed data into the LSTM network structure of the second module, the features of flight delay in time are extracted from the LSTM network module, the extracted features are input into the ResNet network structure to continue extracting deep features, and finally, the specific prediction time of flight delay is output.
Fig. 2 is a structural diagram of the network structure of the present invention, first inputting data into the LSTM network in a time-sequential manner, wherein the used features include: origin _ airport, destination _ airport, data, flight _ number, arrival _ delay. After the time sequence characteristics are extracted through the LSTM network on the left side, the characteristics and the characteristics of airline, future _ delay, taxi _ out, areas _ off, applied _ time, air _ time, distance, areas _ on and taxi _ in are input into a ResNet network structure together, deep characteristics are extracted through a multi-layer residual block structure, and finally, a prediction result is divided into 4 modules to be output by using a softmax function.
Detailed Description
The invention is further illustrated by the following figures and examples.
Step 1: the data set is processed and the processed data set is input into a network model.
The 1-1 data set contains English, such as airport, airline company, etc., which is all converted into numbers, so that the same English letters correspond to the same numbers. The encoding mode uses an ASCII code combination form, for example, ATL in a data set represents Hartsfield-Jackson Atlanta International Airport, ATL is encoded into 658476, LAS is encoded into 766583, and 322 machine fields are encoded in the mode.
1-2, preprocessing the data in the English-processed data set, deleting redundant features, processing missing values, removing unreasonable data and normalizing features.
1-3 the processed data set is divided into three categories to be reproduced.
1-3-1, screening the data in the data set according to flight categories, and separating all the take-off and landing data of different flights in one year into different data sets. For example, flight with flight number 1173 is screened, belonging to 6876 airline, and all of it is screened out and put into a csv file for 329 pieces of data. In this way, a total of 10 flights of data were selected for the experiment.
1-3-2, screening the data in the data set according to the airport categories, classifying different airports, and screening all the take-off and landing flights of a certain airport in one year to different data sets. For example, the ATL airport code is 658476, and all flights landing at this airport are screened out for 64744 flight data. In this way a total of 10 airports of data were selected for the experiment.
1-3-3, screening the data in the data set according to the category of the airlines, and screening all flight take-off and landing data of different airlines in different data sets within one year. For example, AA airline represents American airfilines Inc, coded 6565. 712936 pieces of data are screened out from the take-off and landing data of all flights to which the airlines belong, and in this way, 10 airlines of data are selected for the experiment.
1-4 dataset tagging. Data in the dataset that "ARRIVAL _ DELAY" arrives at the DELAY column is labeled and classified into four categories:
delay time is less than 0 minute;
delay time is more than 0 minute and less than 45 minutes;
delay time is more than 45 minutes and less than 90 minutes;
and fourthly, the delay time is more than 90 minutes.
The prediction of the present invention is to predict this data.
Step 2: and (3) inputting the data set processed in the step (1) into an LSTM network in a time-sequence matrix form to extract time-sequence features. First determining input features to be selected into the LSTM network, the selected input features including:
there are 18 features in total. Taking step 1-3-1 as an example, 329 pieces of data in flight data file with flight number 1173 are sequentially input into the network at 20 time steps, and the data dimension input into the LSTM is xt(329, 20, 18). The number of hidden layers for the LSTM is set to 35.
2-1 Current input x Using LSTMtAnd h passed by the last statet-1And splicing to obtain four states: z, zi、zf、zo。ht-1Is the input received from the last node, where xtRepresenting the data dimension input to the LSTM; z is the input data, and the data is converted into a value between-1 and 1 through a tanh activation function; z is a radical ofi、zf、zoAfter the weight matrix is multiplied by the splicing vector of the data set, the data set is converted into a numerical value between 0 and 1 as a gating state through a sigmoid activation function. Wherein:
w is a weight value set in the z state, wiIs ziWeight value set in State, wfIs zfWeight value set in State, woIs zoWeight value set in state.
Forgetting phase of LSTM. Z calculated in step 2-1fControlling the last state c as a forgetting gate of a forgetting phaset-1Retention and forgetting of the features obtained in (1), wherein ctFor a transmission state of the LSTM network at the current moment, ct-1The transmission state at the previous moment;
2-3. selective memory phase of LSTM. Converting z in step 2-1iAs the selective memory gating of the selective memory stage, important features are recorded in an emphasized mode, and unimportant features are recorded in a small amount.
And 2-4. output stage of LSTM. Converting z in step 2-1oAnd determining the output characteristic of the current state as the output gating of the output stage.
And 2-5, outputting the data through an LSTM network to obtain a multidimensional time sequence characteristic. The vector of the timing feature is denoted as yt(118440, 6580, 6462, 360) dimensions of data.
And step 3: timing feature y extracted from LSTM networktInputting the data into a ResNet network for further feature extraction.
Inputting the timing characteristics into the residual structure of ResNet152, first y istSet to feature x of ResNet first layerlThrough xl+1＝xl+F(xl,wl) The residual structure of (2) transfers the features of the shallow network into the deep network, so that the features are not lost in the deep network, and the efficiency of the feature extractor is improved. By recursion, the expression of the deep unit Lth layer feature is:
where L denotes the L < th > layer of the deep unit of the residual network, L denotes the L < th > layer of the shallow unit of the residual network, xLIs a deep unitCharacteristic of L layer, xlIs the characteristic of the l layer of the shallow unit, i is any i layer of the residual error network, xiIs a feature of the i-th layer unit, wiIs the weight value of the i-th layer unit.
I.e. feature xLIs the product of a series of matrix vectors. Next, a back propagation calculation of the residual error network is performed, with a loss function E and a back propagation derivation as:
after calculation, the deep layer characteristic x is obtainedLIs expressed as data of (50008, 3040, 2736, 3800) dimensions.
And 4, step 4: and predicting the flight delay state and calculating the prediction accuracy.
4-1 extracting the deep layer characteristic x extracted in the step 3LThe results are classified into four categories of predictions via the Softmax function:
1) the delay time is less than 0 minute;
2) the delay time is more than 0 minute and less than 45 minutes;
3) the delay time is more than 45 minutes and less than 90 minutes;
4) the delay time is greater than 90 minutes.
The Softmax functional formula used by the invention is as follows:
wherein x is1，x2，x3，x4Are respectively the first element, x, in the four classified categoriesiIs the ith element in the category, and n is the total number of elements in each category.
And 4-2, comparing the predicted four types of results with the labels made in the step 1, and calculating the prediction accuracy acc. The accuracy is the flight delay prediction accuracy by dividing the total number of the delay prediction results predicted by the calculation model and the class equality corresponding to the label by the total number of the test sets. Wherein, y represents the total number of the test sets in the test sample, m represents the total number of the prediction results equal to the categories corresponding to the labels, and the accuracy calculation formula is as follows:
in the experiment, the hyper-parameters set by the network structure are as follows: the learning rate was set to 0.01; batch _ size is set to 100; the number of hidden layers of the LSTM structure is 35; the number of nodes in each layer is set to be 30; the input dimension of the ResNet structure is 35 consistent with the number of hidden layers of the LSTM; the number of hidden layers of ResNet is 152. The error function uses a cross entropy loss function: l- [ acyl '+ (1-y) log (1-y') ]; the optimization algorithm in the network structure is a Stochastic Gradient Descent algorithm.
The structure provided by the invention adds the characteristic of flight delay on the time sequence, and keeps the characteristic information in the shallow network while deepening the network layer number through the residual structure network, thereby improving the accuracy of flight delay prediction compared with the traditional method.
Claims (5)
1. A flight delay prediction method based on deep learning is characterized by comprising the following steps:
step 1: processing the data set and inputting the processed data set into a network model;
step 2: inputting the data set processed in the step 1 into an LSTM network in a time sequence matrix form to extract time sequence characteristics on a time sequence;
and step 3: inputting the time sequence characteristics extracted by the LSTM network in the step 2 into a ResNet network for further characteristic extraction;
and 4, step 4: predicting flight delay states and calculating prediction accuracy;
the step 1 is specifically realized as follows:
1-1, converting all English contained in the data set into numbers, enabling the same English letters to correspond to the same numbers, and using an ASCII code combination mode for coding;
1-2, preprocessing data in the English processed data set: deleting redundant features, processing missing values, removing unreasonable data and normalizing features;
1-3, dividing the processed data set into three categories to be reproduced;
1-3-1, screening the data in the data set according to flight types, respectively establishing the data set for all the take-off and landing data of each flight within one year according to different flights, and then randomly selecting the data of 10 flights to perform an experiment;
1-3-2, screening the data in the data set according to the airport categories, respectively establishing the data set for all the take-off and landing flight data of each airport within one year according to different airports, and then randomly selecting the data of 10 airports to carry out experiments;
1-3-3, screening the data in the data set according to the category of the airlines, and respectively screening all flight take-off and landing data of the same airline within one year to different data sets according to different airlines;
1-4, labeling a data set; data in the dataset that "ARRIVAL _ DELAY" arrives at the DELAY column is labeled and classified into four categories:
delay time is less than 0 minute;
delay time is more than 0 minute and less than 45 minutes;
delay time is more than 45 minutes and less than 90 minutes;
and fourthly, the delay time is more than 90 minutes.
2. The flight delay prediction method based on deep learning of claim 1, wherein the step 2 is implemented as follows:
first, input features selected for input into the LSTM network are determined, the selected input features being 18 in number, including:
F-FLIGHT _ NUMBER FLIGHT NUMBER; G-TAIL _ NUMBER airplane NUMBER; H-ORIGIN _ AIRPORT takeoff AIRPORT; I-DESTINATION _ AIRPORT arrives at the AIRPORT; J-SCHEDULED _ DEPARTURE planned takeoff time; K-DEPARTURE _ TIME DEPARTURE TIME; L-DEPARTURE _ DELAY DEPARTURE DELAY; M-TAXI _ OUT coast time; N-WHEELS _ OFF aircraft landing gear retraction time; O-SCHEDULED _ TIME planned flight TIME; P-ELAPSED _ TIME ELAPSED TIME; Q-AIR _ TIME flight TIME; R-DISTANCE flight DISTANCE; S-WHEELS _ ON aircraft landing gear down time; the T-TAXI _ IN airplane is stopped and positioned according to the instruction; U-SCHEDULED _ ARRIVAL projected ARRIVAL time; V-ARRIVAL _ TIME actual ARRIVAL TIME; W-ARRIVAL _ DELAY ARRIVAL DELAY time;
2-1 Current input x Using LSTMtAnd h passed by the last statet-1And splicing to obtain four states: z, zi、zf、zo；ht-1Is the input received from the last node, where xtRepresenting the data dimension input to the LSTM; z is the input data, and the data is converted into a value between-1 and 1 through a tanh activation function; z is a radical ofi、zf、zoAfter a splicing vector of a data set is multiplied by a weight matrix, a sigmoid activation function is used for converting the value between 0 and 1 to be used as a gating state; wherein:
w is a weight value set in the z state, wiIs ziWeight value set in State, wfIs zfWeight value set in State, woIs zoA weight value set in a state;
2-2. forgetting stage of LSTM; z calculated in step 2-1fControlling the last state c as a forgetting gate of a forgetting phaset-1Retention and forgetting of the features obtained in (1), wherein ctFor the current time of the LSTM networkA transmission state, ct-1The transmission state at the previous moment;
2-3. the selective memory stage of LSTM; converting z in step 2-1iSelecting memory gating as a selection memory stage;
2-4. output stage of LSTM; converting z in step 2-1oAs output gating of an output stage, determining the output characteristics of the current state;
2-5, outputting the data through an LSTM network to obtain a multi-dimensional time sequence characteristic yt。
3. The flight delay prediction method based on deep learning of claim 2, wherein the step 3 is implemented as follows:
will time sequence characteristic ytInput to the residual structure of ResNet152, y is first inputtSet to feature x of ResNet first layerlThrough xl+1＝xl+F(xl，wl) The residual structure of (a) transfers the features of the shallow network into the deep layer; by recursion, the expression of the deep unit Lth layer feature is:
wherein L denotes the L < th > level of the deep unit of the residual network, 1 denotes the 1 < th > level of the shallow unit of the residual network, xLCharacteristic of the L-th layer of the deep unit, xlIs the characteristic of the 1 st layer of the shallow cell, i is any i-th layer of the residual network, xiIs a feature of the i-th layer unit, wiIs the weighted value of the unit of the ith layer;
i.e. feature xLIs the product of a series of matrix vectors; next, a back propagation calculation of the residual error network is performed, with a loss function E and a back propagation derivation as:
4. the flight delay prediction method based on deep learning of claim 3, wherein the step 4 is implemented as follows:
4-1, extracting the deep layer characteristic x extracted in the step 3LThe results are classified into four categories of predictions via the Softmax function:
1) the delay time is less than 0 minute;
2) the delay time is more than 0 minute and less than 45 minutes;
3) the delay time is more than 45 minutes and less than 90 minutes;
4) the delay time is more than 90 minutes;
the Softmax function used is:
wherein x is1，x2，x3，x4Are respectively the first element, x, in the four classified categoriesiIs the ith element in the category, and n is the total number of elements in each category;
4-2, comparing the predicted four types of results with the labels made in the step 1, and calculating the prediction accuracy acc; wherein, y represents the total number of the test sets in the test sample, m represents the total number of the prediction results equal to the categories corresponding to the labels, and the accuracy calculation formula is as follows:
5. the flight delay prediction method based on deep learning of claim 4, wherein the hyper-parameters set by the LSTM network structure are as follows: the learning rate was set to 0.01; batch _ size is set to 100; the number of hidden layers of the LSTM structure is 35; the number of nodes in each layer is set to be 30; the input dimension of the ResNet structure is 35 consistent with the number of hidden layers of the LSTM; the number of hidden layers of ResNet is 152; the error function uses a cross entropy loss function: l- [ acyl '+ (1-y) log (1-y') ]; the optimization algorithm in the network structure is a Stochastic Gradient Descent algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910870557.3A CN110659773A (en) | 2019-09-16 | 2019-09-16 | Flight delay prediction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910870557.3A CN110659773A (en) | 2019-09-16 | 2019-09-16 | Flight delay prediction method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110659773A true CN110659773A (en) | 2020-01-07 |
Family
ID=69037343
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910870557.3A Pending CN110659773A (en) | 2019-09-16 | 2019-09-16 | Flight delay prediction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110659773A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381336A (en) * | 2020-12-11 | 2021-02-19 | 中国民航科学技术研究院 | Flight delay duration prediction method and system |
| CN113128769A (en) * | 2021-04-22 | 2021-07-16 | 上海工程技术大学 | Intelligent flight delay prediction method based on deep learning |
| CN115115312A (en) * | 2022-07-11 | 2022-09-27 | 智航信息服务(深圳)有限公司 | Flight delay early warning method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299401A (en) * | 2018-07-12 | 2019-02-01 | 中国海洋大学 | Metropolitan area space-time stream Predicting Technique based on deep learning model LSTM-ResNet |
| CN109448445A (en) * | 2018-09-05 | 2019-03-08 | 南京航空航天大学 | Flight based on shot and long term Memory Neural Networks is delayed grading forewarning system method |
| CN109492334A (en) * | 2018-12-11 | 2019-03-19 | 青岛心中有数科技有限公司 | Delayed method for establishing model, prediction technique and device |
| CN109767759A (en) * | 2019-02-14 | 2019-05-17 | 重庆邮电大学 | End-to-end speech recognition methods based on modified CLDNN structure |
| CN109948716A (en) * | 2019-03-25 | 2019-06-28 | 中国民航大学 | A kind of airport delay prediction technique based on region residual sum LSTM network |
-
2019
- 2019-09-16 CN CN201910870557.3A patent/CN110659773A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299401A (en) * | 2018-07-12 | 2019-02-01 | 中国海洋大学 | Metropolitan area space-time stream Predicting Technique based on deep learning model LSTM-ResNet |
| CN109448445A (en) * | 2018-09-05 | 2019-03-08 | 南京航空航天大学 | Flight based on shot and long term Memory Neural Networks is delayed grading forewarning system method |
| CN109492334A (en) * | 2018-12-11 | 2019-03-19 | 青岛心中有数科技有限公司 | Delayed method for establishing model, prediction technique and device |
| CN109767759A (en) * | 2019-02-14 | 2019-05-17 | 重庆邮电大学 | End-to-end speech recognition methods based on modified CLDNN structure |
| CN109948716A (en) * | 2019-03-25 | 2019-06-28 | 中国民航大学 | A kind of airport delay prediction technique based on region residual sum LSTM network |
Non-Patent Citations (1)
| Title |
|---|
| 谢海涛 等: "基于LSTM的媒体网站用户流量预测与负载均衡方法", 《网络空间安全》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381336A (en) * | 2020-12-11 | 2021-02-19 | 中国民航科学技术研究院 | Flight delay duration prediction method and system |
| CN113128769A (en) * | 2021-04-22 | 2021-07-16 | 上海工程技术大学 | Intelligent flight delay prediction method based on deep learning |
| CN115115312A (en) * | 2022-07-11 | 2022-09-27 | 智航信息服务(深圳)有限公司 | Flight delay early warning method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yu et al. | Flight delay prediction for commercial air transport: A deep learning approach | |
| CN107086935B (en) | People flow distribution prediction method based on WIFI AP | |
| CN111401601B (en) | Delay propagation-oriented flight take-off and landing time prediction method | |
| Ding | Predicting flight delay based on multiple linear regression | |
| CN110659773A (en) | Flight delay prediction method based on deep learning | |
| Olive et al. | Identifying anomalies in past en-route trajectories with clustering and anomaly detection methods | |
| CN111178628B (en) | Luggage arrival time prediction method and device | |
| Dhief et al. | Predicting aircraft landing time in extended-TMA using machine learning methods | |
| Srinivasan et al. | Mining and classifying aviation accident reports | |
| Ballakur et al. | Empirical evaluation of gated recurrent neural network architectures in aviation delay prediction | |
| CN116468186B (en) | Flight delay time prediction method, electronic equipment and storage medium | |
| Ai et al. | A deep learning approach to predict the spatial and temporal distribution of flight delay in network | |
| Hu et al. | The study on hard landing prediction model with optimized parameter SVM method | |
| CN110443448A (en) | A kind of aircraft seat in the plane classification prediction technique and system based on two-way LSTM | |
| Subramanian et al. | Deep-learning based time series forecasting of go-around incidents in the national airspace system | |
| Basturk et al. | Prediction of aircraft estimated time of arrival using machine learning methods | |
| CN112132366A (en) | Prediction system for flight clearance rate | |
| Belkoura et al. | Beyond linear delay multipliers in air transport | |
| Mirmohammadsadeghi et al. | Enhancements to the runway capacity simulation model using the asde-x data for estimating airports throughput under various wake separation systems | |
| Chou et al. | A machine learning application for predicting and alerting missed approaches for airport management | |
| Cheng et al. | Data mining for air traffic flow forecasting: a hybrid model of neural network and statistical analysis | |
| Zhang et al. | Runway visual range prediction based on ensemble learning | |
| Ayaydın et al. | Deep learning based forecasting of delay on flights | |
| Wang et al. | A review of flight delay prediction methods | |
| Suryaman et al. | Application of data mining for slot time prediction at international airports in indonesia: J48 algorithm |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200107 |