Flight return standby identification method and system based on machine learning

Abstract

The invention belongs to the technical field of civil aviation operation safety evaluation, and discloses a machine learning-based flight return standby identification method and system. The method comprises the following steps: establishing a flight learning feature sample by comprehensively considering data in the aspects of time, course, track, barometric altitude, radio altitude and the like of flight data; large-batch training and learning are carried out on flight data actually generated in the flight process; based on a machine learning algorithm of a gradient boosting decision tree, whether a flight is a homeward voyage standby flight or not is quickly judged from mass trained and learned flight data, key quantized data of the flight standby flight is extracted, and operation safety evaluation is carried out on the flight. According to the method, automatic auxiliary analysis support is provided for airline operation safety analysis.

Description

Technical Field

[0001] The present invention belongs to the technical field of civil aviation operation safety assessment, and in particular relates to a method and system for identifying flight return and alternate landing based on machine learning. Background Art

[0002] The existing airlines’ identification of flight return diversions is mostly done by manually recording and marking key information such as flight numbers, finding the decoded flight data in the information system to mark the flight return diversion, and then conducting safety analysis such as information analysis of return diversion risks. The identification of actual flight return diversions requires comprehensive consideration and quantification of data such as time (UTC), heading (magnetic heading), track (latitude and longitude coordinates), pressure altitude, and radio altitude for evaluation.

[0003] Through the above analysis, the problems and defects of the existing technology are as follows: the existing technology cannot quickly determine whether it is a return or diversion flight in the comprehensive consideration of flight data, cannot effectively extract the key quantified data of the diversion, and has low accuracy in flight operation safety assessment. Summary of the invention

[0004] In order to overcome the problems existing in the related art, the disclosed embodiments of the present invention provide a method and system for identifying flight return and alternate landing based on machine learning.

[0005] The technical solution is as follows: A method for identifying flight return and alternate landing based on machine learning, comprising: S1, establish flight learning feature samples by analyzing the flight data of time UTC, heading, track, pressure altitude, and radio altitude; S2, using the established flight learning feature samples and the flight data actually generated during the flight process to train and learn, using the training label data set to train the model, and using the supervised training data set to contain input labels and corresponding output labels, and using the validation set to monitor the performance of the model; S3, a machine learning algorithm based on the gradient boosting decision tree (GBDT), can quickly determine whether a flight is a return or diversion flight from the flight data after training and learning, and extract the key quantified data of the diversion flight to conduct an operational safety assessment of the flight.

[0006] In step S1, a flight learning feature sample is established, including: establishing the features of the flight return and alternate landing, knowing that A set of examples and their corresponding labels, expressed as: ; In the formula, the sample Depend on Attribute description, label The value of is 1 or -1; Use supervised learning algorithms to determine the input from the data set and output Determine the prediction function , get the predicted value ; In flight risk identification, features Represents the terminal factor vector, each component takes an integer between 1 and 10, A serial number representing the flight ID, It means abnormal, which includes flight return or alternate landing. Indicates that the flight is normal.

[0007] In step S2, training and learning are performed, including: Using cost-sensitive methods or sampling-based techniques, the established flight learning feature samples and the flight data actually generated during the flight process are trained and learned for the category imbalance problem, which is used to identify flight risks based on the number of training samples of different categories in the classification task.

[0008] In step S3, the machine learning algorithm of the gradient boosting decision tree GBDT adopts a method combining oversampling of the training data set with the SMOTE function and data replication to solve the imbalance problem of the number of samples of the two categories.

[0009] In step S3, an operational safety assessment is performed on the flight, including: Step 1. Use SMOTE technology to perform oversampling to obtain a new data set newtrain; Step 2. Copy the abnormal data in newtrain and ORItrain to form the tmpabnormal data set, and compare the sizes of the normal data set and the tmpabnormal data set to get keynum; Step 3. Take keynum as the sampling number and randomly sample the tmpabnormal and normal data sets to form SYNtrain and output SYNtrain as the training data of XGBOOST; Step 4. Use the training data and xgboost to get the model; output the confusion matrix of the training data and the misclassified flight information; Step 5. Call the model to predict the test data; output the confusion matrix of the test data and record the misclassified flight numbers.

[0010] In step 1, before obtaining the new data set newtrain, it is necessary to: combine the data training data received by the gradient boosting decision tree model, namely flightnormal, flightabnormal0 and flightabnormal1, into a training data set train1.

[0011] In step 1, the parameters of the training data received by the gradient boosting decision tree model are set as follows: sampling volume 1000% and using 5-nearest neighbors.

[0012] In step 2, check=0 for the abnormal data in newtrain and the abnormal data in ORItrain; The abnormal data in newtrain and the abnormal data in ORItrain are replicated 20 times; Compare the keynums obtained from the normal dataset and the tmpabnormal dataset, and take the one with the smaller number of data in the two datasets.

[0013] In step 3, a random sample is formed into SYNtrain, and the ratio of the number of 0s and 1s in SYNtrain is 1:1.

[0014] Another object of the present invention is to provide a flight return and alternate landing identification system based on machine learning, the system implements the flight return and alternate landing identification method based on machine learning, and the system includes: The flight learning feature sample establishment module is used to establish the flight learning feature sample by analyzing the flight data of time UTC, heading, track, pressure altitude, and radio altitude; The class imbalance problem training and learning module is used to train and learn using the established flight learning feature samples and the flight data actually generated during the flight process. The model is trained using the training label dataset. The supervised training dataset contains input labels and corresponding output labels, and the validation set is used to monitor the performance of the model. The gradient boosting decision tree learning module is used for the machine learning algorithm based on the gradient boosting decision tree GBDT to quickly determine whether it is a return or diversion flight from the flight data after training and learning, and to extract the key quantified data of the diversion to conduct an operational safety assessment of the flight.

[0015] Combined with all the above technical solutions, the beneficial effects of the present invention are as follows: the present invention uses time (UTC), heading (magnetic heading), track (latitude and longitude coordinates), pressure altitude, radio altitude and other aspects as automated flight return and diversion identification features to quickly determine whether the flight is returning and diverting, extract key quantified data for return and diversion, conduct operational safety assessments on flights, and provide automated auxiliary analysis support for airline operational safety analysis. Traditional return and diversion identification requires manual analysis of data to determine the compliance of return and diversion data. The present invention can quickly identify return and diversion flights, improve data reliability and availability, and reduce manual identification costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the principles of the present disclosure; Figure 1 This is a flow chart of a method for identifying flight return and alternate landing based on machine learning provided by an embodiment of the present invention; Figure 2 is an input and output relationship diagram of the i-th sample provided by an embodiment of the present invention; Figure 3 1 is a diagram of the operation principle of the Boosting model provided in an embodiment of the present invention; wherein (a) is a diagram of the Boosting model (M=1), (b) is a diagram of the Boosting model (M=2), (c) is a diagram of the Boosting model (M=3), (d) is a diagram of the Boosting model (M=6), (e) is a diagram of the Boosting model (M=10), and (f) is a diagram of the Boosting model (M=150); Figure 4 It is a process effect diagram of Boosting provided by an embodiment of the present invention; Figure 5 is a schematic diagram of a flight return and alternate landing identification system based on machine learning provided by an embodiment of the present invention; In the figure: 1. Flight learning feature sample establishment module; 2. Class imbalance problem training and learning module; 3. Gradient boosting decision tree learning module. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0018] The innovation of the present invention lies in that: based on a machine learning algorithm, the present invention automatically identifies whether a flight is to be returned or diverted by analyzing flight data, thereby solving the problem of low efficiency of traditional manual identification and authentication.

[0019] Embodiment 1, as Figure 1 As shown, the method for identifying flight return and alternate landing based on machine learning provided by an embodiment of the present invention includes: S1, establish flight learning feature samples by analyzing the flight data of time UTC, heading, track, pressure altitude, and radio altitude; S2, using the established flight learning feature samples and the flight data actually generated during the flight process to train and learn, using the training label data set to train the model, and using the supervised training data set to contain input labels and corresponding output labels, and using the validation set to monitor the performance of the model; S3, a machine learning algorithm based on the Gradient Boosting Decision Tree (GBDT), can quickly determine whether a flight is a return or diversion flight from the flight data after training and learning, and extract the key quantified data of the diversion flight to conduct an operational safety assessment of the flight, providing automated auxiliary analysis support for airlines' operational safety analysis.

[0020] Example 2, as another detailed implementation of the present invention, the method for identifying flight return and alternate landing based on machine learning provided in the embodiment of the present invention includes: The first step is to identify the characteristics of the flight's return and alternate landing: Known A set of examples and their corresponding labels, expressed as: ; In the formula, the sample Depend on Attribute description, label The value of is 1 or -1; Use supervised learning algorithms to determine the input from the data set and output Determine the prediction function , get the predicted value ; In flight risk identification, features Represents the terminal factor vector, each component takes an integer between 1 and 10, A serial number representing the flight ID, It means abnormal, which includes flight return or alternate landing. Indicates that the flight is normal. The input and output relationship of each sample is shown in Figure 2.

[0021] It can be understood that according to formula (1), the prediction model is obtained from the data set ,make , for any given flight feature vector , which is the predicted output for this flight ,in, It means that the flight is abnormal (returning or making an alternate landing); otherwise, it means that the flight is normal.

[0022] Among them, the number of "abnormal" flights in the data of flight risk identification problem is very small, which belongs to the category imbalance problem.

[0023] The second step is the category imbalance problem: Common classification learning methods all have a common basic assumption, that is, the number of training examples in different categories is equal. If the number of training examples in different categories is slightly different, it usually has little impact, but if the difference is large, it will cause trouble in the learning process. For example, there are 998 counterexamples , but there are only two positive examples, then the learning method can achieve 99.8% accuracy by simply returning a learner that always predicts new samples as negative examples; however, such a learner is often worthless because it cannot predict any positive examples.

[0024] Class-imbalance refers to the situation where the number of training examples of different categories in the classification task is very different. Without loss of generality, the present invention assumes that there are fewer positive class examples and more negative class examples. For the flight risk identification problem in the present invention, there are fewer examples of return and alternate flights and more examples of normal flights, which is a typical class-imbalance problem.

[0025] There are two basic methods to solve the problem of class imbalance. One is the cost-sensitive method, such as weighted support vector machine; the other is sampling-based technology. There are generally three types of existing technologies: the first is to directly "undersample" the negative samples in the training set, that is, remove some negative examples to make the number of positive and negative samples close, and then learn; the second is to "oversample" the positive samples in the training set, that is, add some positive examples to make the number of positive and negative samples close, and then learn; the third is directly based on the combination of the two, that is, undersample the negative examples and oversample the positive examples. The representative algorithm of oversampling, SMOTE, generates additional positive examples by interpolating the positive examples in the training set. On the other hand, if the undersampling method randomly discards negative examples, some important information may be lost; the representative algorithm of the undersampling method, EasyEnsemble, uses an ensemble learning mechanism to divide the counterexamples into several sets for different learners to use, so that each learner is undersampled, but no important information is lost from a global perspective.

[0026] The third step is the gradient boosting decision tree machine learning algorithm: Model combinations, such as Gradient Boosting Decision Tree (GBDT) or Random Forest (RF), are combinations of simple models, and the prediction effect is better than a single more complex model. There are many ways to combine, and randomization (such as RF) and Boosting (such as GBDT) are typical methods. The present invention selects the GradientBoosting method. Establish an initial model as the basis for iteration. In each iteration, it is necessary to calculate the residual between the predicted value and the actual value of the current model as the target variable, train a new weak classifier, and add the prediction result of the newly trained weak classifier to the predicted value of the current model, thereby updating the model. Repeat the calculation of residuals, training new weak classifiers, and updating the model until the preset number of iterations is reached or the model performance reaches the preset accuracy.

[0027] like Figure 3 As shown in the figure, the idea of ​​Boosting is quite simple. It is to build M models (such as classification) for the data set. Generally, this model is relatively simple and is called a weak classifier. Each time the classification is performed, the weight of the wrongly classified data is increased a little bit before classification. In this way, the final classifier can achieve relatively good results on both the test data and the training data. Figure 3 In Figures (a) to (f), the green line represents the current model (the model is obtained by merging the models obtained M times before), the dotted line represents the current model, and the red and blue dots are the data. The larger the dot, the higher the weight.

[0028] Figure 4 As shown in the figure, this is a Boosting process. The solid line represents the current model (the model is obtained by merging the models obtained in the previous M times), and the dotted line represents the current model. Each time when classifying, more attention will be paid to the wrong data. Figure 4 In the figure, the dots and circles are the data. The larger the dots, the higher the weights. When M=150, the obtained model can almost distinguish the red and blue dots.

[0029] The fourth step is the learning method and steps of gradient boosting decision tree.

[0030] In order to solve the imbalance problem of the number of samples in the two categories in the experiment (seed=300 and seed=321), the method of oversampling the training data set with the SMOTE function and combining data replication is adopted. The detailed steps are as follows: Step 1. The training data received by the gradient boosting decision tree model, namely flightnormal, flightabnormal0 and flightabnormal1, are used to form the training data set train1. Then, the SMOTE technology is used for oversampling. K nearest neighbor samples are selected from the minority class samples. The difference between the two samples is calculated to randomly generate new samples. For each difference, a random number is multiplied, and then the result is added to the original sample to obtain a synthetic new sample to obtain a new data set newtrain. The parameter settings of this experiment are: sampling volume 1000%, using 5-nearest neighbors.

[0031] Step 2. Next, copy the abnormal data (check=0) in newtrain 20 times and the abnormal data (check=0) in ORItrain 20 times to form the tmpabnormal data set, and then compare the size of the normal data set with the tmpabnormal data set to get keynum (take the smaller one of the two data sets).

[0032] Step 3. Take keynum as the sampling number and randomly sample the tmpabnormal and normal data sets to form SYNtrain (so that the ratio of 0 to 1 in SYNtrain is 1:1) and output SYNtrain as the training data of XGBOOST.

[0033] Step 4. Use the training data and xgboost to get the gradient boosting decision tree model. Output the confusion matrix of the training data and the misclassified flight information. The confusion matrix is ​​a sample table of the gradient boosting decision tree matrix for two types of data samples.

[0034] Step 5. Call the model to predict the test data. Output the confusion matrix of the test data and record the misclassified flight numbers.

[0035] Examples of gradient boosting decision trees. Table 1 records the confusion matrices of two experiments with seed = 300 and seed = 321 on four datasets. Table 2 gives the information of misclassified flights in the corresponding experiments.

[0036] Observe the results of the models in Table 1 and Table 2 on the 8 test sets, which correspond to the confusion matrix in the second and fourth columns of Table 1 and the misclassified flight information in the second and fourth columns of Table 2, and we can summarize the following phenomena: (1) Using abnormal1 does not bring any new information to the identification of return flights, that is, the number of return flights identified by the model (the number of true samples) is not different. Combined with Table 2, it is found that the return flight abnormal0 that is incorrectly identified on the test set is exactly the same in the two cases of not using abnormal1 data (column 2) and using abnormal1 data (column 4). After using abnormal1, one sample is added to the test set before and after takeoff, and this sample is misjudged.

[0037] (2) Using abnormal1 increases the number of false negative samples, that is, the number of misclassified normal flights in column 4 of Table 2 is greater than the number of misclassified normal flights in column 2.

[0038] One possible explanation for the above phenomenon is that the abnormal1 data is added. As can be seen from the third and fifth columns of Table 2, a small number of abnormal1s are indeed identified in the training, and the recognition rate of normal flights in the training set is also improved. However, the price paid here is overfitting, which causes some normal flights in the test data to be misclassified as abnormal flights.

[0039] Table 1 Confusion matrix of two sets of experiments (seed=300) and (seed=321) for four-time learning

[0040] Table 2 corresponds to the flight numbers predicted incorrectly in Table 1

[0041] Embodiment 2, as Figure 5 As shown, the flight return and alternate landing identification system based on machine learning provided by the embodiment of the present invention includes: The flight learning feature sample establishment module 1 is used to establish the flight learning feature sample by analyzing the time UTC, heading, track, pressure altitude, and radio altitude of the flight data; Class imbalance problem training and learning module 2 is used to train and learn using the established flight learning feature samples and the flight data actually generated during the flight process, using the training label data set to train the model, and using the supervised training data set to contain input labels and corresponding output labels, and using the validation set to monitor the performance of the model; The gradient boosting decision tree learning module 3 is used for the machine learning algorithm based on the gradient boosting decision tree GBDT to quickly determine whether it is a return or alternate flight from the flight data after training and learning, and to extract the key quantified data of the return or alternate flight to conduct an operational safety assessment of the flight.

[0042] Exemplarily, the flight learning feature sample establishment module 1 includes: The feature recognition module of flight return and diversion is used for known The set of examples and their corresponding labels determines the input and output The dependencies between them and determine the prediction function , get the predicted value .

[0043] The category imbalance problem training and learning module 2 includes: a category imbalance problem training and learning module, which is used to identify the number of training samples of different categories in the classification task through a cost-sensitive method or a sampling-based technology.

[0044] Gradient boosting decision tree learning module 3 includes: The new data set newtrain acquisition submodule is used to train the data received by the model, namely, flightnormal, flightabnormal0 and flightabnormal1, to form a training data set train1, and then oversample using the SMOTE technology to obtain a new data set newtrain.

[0045] The keynu acquisition submodule is used to copy the abnormal data in newtrain 20 times and the abnormal data in ORItrain 20 times to form the tmpabnormal data set, and then compare the sizes of the normal data set and the tmpabnormal data set to obtain keynum.

[0046] The SYNtrain composition submodule is used to randomly sample the tmpabnormal and normal data sets with keynum as the sampling number to form SYNtrain and output SYNtrain as the training data of XGBOOST.

[0047] The confusion matrix and misclassified flight information output submodule is used to obtain the model using the training data and xgboost. The confusion matrix and misclassified flight information of the training data are output.

[0048] The prediction submodule is used to call the model to predict the test data. It outputs the confusion matrix of the test data and records the misclassified flight numbers.

[0049] Currently, airlines still use manual interpretation to analyze flight data for return and diversion, verifying whether the flight data meets the conditions for return and diversion, which is very inefficient. The present invention can automatically identify the characteristics of flight return and diversion based on the time (UTC), heading (magnetic heading), track (latitude and longitude coordinates), pressure altitude, radio altitude and other aspects of the flight data, quickly determine whether the flight is returning and diverting, extract key quantified data for return and diversion, conduct operational safety assessments on flights, and provide automated auxiliary analysis support for airline operational safety analysis.

[0050] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for identifying flight return and alternate landing based on machine learning, characterized in that: The method includes: S1, establish flight learning feature samples by analyzing the flight data of time UTC, heading, track, pressure altitude, and radio altitude; S2, using the established flight learning feature samples and the flight data actually generated during the flight process to train and learn, using the training label data set to train the model, and using the supervised training data set to contain input labels and corresponding output labels, and using the validation set to monitor the performance of the model; S3, a machine learning algorithm based on the gradient boosting decision tree (GBDT), can quickly determine whether a flight is a return or diversion flight from the flight data after training and learning, and extract the key quantified data of the diversion flight to conduct an operational safety assessment of the flight.

2. The method for identifying flight return and alternate landing based on machine learning according to claim 1, characterized in that: In step S1, a flight learning feature sample is established, including: establishing the features of the flight return and alternate landing, knowing that A set of examples and their corresponding labels, expressed as: ； In the formula, the sample Depend on Attribute description, label The value of is 1 or -1; Use supervised learning algorithms to determine the input from the data set and output The dependency relationship between them determines the prediction function , get the predicted value ; In flight risk identification, features Represents the terminal factor vector, each component takes an integer between 1 and 10, A serial number representing the flight ID, It means abnormal, which includes flight return or alternate landing. Indicates that the flight is normal.

3. The method for identifying flight return and alternate landing based on machine learning according to claim 1, characterized in that: In step S2, training and learning are performed, including: Using cost-sensitive methods or sampling-based techniques, the established flight learning feature samples and the flight data actually generated during the flight process are trained and learned for the category imbalance problem, which is used to identify flight risks based on the number of training samples of different categories in the classification task.

4. The method for identifying flight return and alternate landing based on machine learning according to claim 1, characterized in that: In step S3, the machine learning algorithm of the gradient boosting decision tree GBDT adopts a method combining oversampling of the training data set with the SMOTE function and data replication to solve the imbalance problem of the number of samples of the two categories.

5. The method for identifying flight return and alternate landing based on machine learning according to claim 1, characterized in that: In step S3, an operational safety assessment is performed on the flight, including: Step 1. Use SMOTE technology to perform oversampling to obtain a new data set newtrain; Step 2. Copy the abnormal data in newtrain and ORItrain to form the tmpabnormal data set, and compare the sizes of the normal data set and the tmpabnormal data set to get keynum; Step 3. Take keynum as the sampling number and randomly sample the tmpabnormal and normal data sets to form SYNtrain and output SYNtrain as the training data of XGBOOST; Step 4. Use the training data and xgboost to get the model; output the confusion matrix of the training data and the misclassified flight information; Step 5. Call the model to predict the test data; output the confusion matrix of the test data and record the misclassified flight numbers.

6. The method for identifying flight return and alternate landing based on machine learning according to claim 5, characterized in that: In step 1, before obtaining the new data set newtrain, it is necessary to: combine the data training data received by the gradient boosting decision tree model, namely flightnormal, flightabnormal0 and flightabnormal1, into a training data set train1.

7. The method for identifying flight return and alternate landing based on machine learning according to claim 5, characterized in that: In step 1, the parameters of the training data received by the gradient boosting decision tree model are set as follows: sampling volume 1000% and using 5-nearest neighbors.

8. The method for identifying flight return and alternate landing based on machine learning according to claim 5, characterized in that: In step 2, check=0 for the abnormal data in newtrain and the abnormal data in ORItrain; The abnormal data in newtrain and the abnormal data in ORItrain are replicated 20 times; Compare the keynums obtained from the normal dataset and the tmpabnormal dataset, and take the one with the smaller number of data in the two datasets.

9. The method for identifying flight return and alternate landing based on machine learning according to claim 5, characterized in that: In step 3, a random sample is formed into SYNtrain, and the ratio of the number of 0s and 1s in SYNtrain is 1:

1.

10. A flight return and alternate landing identification system based on machine learning, characterized in that: The system implements the flight return and alternate landing identification method based on machine learning as described in any one of claims 1 to 9, and the system includes: A flight learning feature sample establishment module (1) is used to establish a flight learning feature sample by analyzing the flight data including the time UTC, heading, track, pressure altitude, and radio altitude; The class imbalance problem training and learning module (2) is used to train and learn using the established flight learning feature samples and the flight data actually generated during the flight process, using the training label data set to train the model, and using the supervised training data set to include input labels and corresponding output labels, and using the validation set to monitor the performance of the model; The gradient boosting decision tree learning module (3) is used for the machine learning algorithm based on the gradient boosting decision tree GBDT to quickly determine whether it is a return flight from the flight data after training and learning, and to extract the key quantitative data of the return flight to conduct an operational safety assessment of the flight.

Images