Hierarchical Determination Method and System for Flight Delay Causes Based on Multi-Task Metrics

Abstract

The present invention belongs to the technical field of air transportation management, and particularly relates to a method and system for hierarchical determination of flight delay causes based on multi-task metrics. The method includes: obtaining civil aviation flight operation data and hierarchical labels for delay causes; performing data preprocessing; constructing a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer; using multi-level nesting to construct a multi-task metric and hierarchical classification model; training to obtain a delay cause classification model based on the multi-task metric and hierarchical classification model; and performing hierarchical determination of flight delay causes through the delay cause classification model. By integrating multi-task metrics and hierarchical reasoning technologies, the stability and robustness of the delay cause determination system are significantly improved, efficient and accurate delay data analysis is achieved, and scientific decision-making support is provided for civil aviation management departments.

Description

Technical Field

[0001] The present invention belongs to the technical field of air transportation management, and particularly relates to a method and system for hierarchical determination of flight delay causes based on multi-task metrics. Background Art

[0002] The problem of flight delays is one of the key challenges in the field of air transportation management, having a profound impact on airlines, airport management departments, and passengers. With the rapid development of the air transportation industry, the flight operation environment has become increasingly complex, and delays occur frequently. The reasons usually involve multiple factors, such as weather, air traffic control, airport resource scheduling, airline operation plans, etc. These factors are intertwined, forming complex characteristics at multiple levels and scales, making it extremely difficult to intelligently determine and analyze the causes of delays.

[0003] Currently, the identification of flight delay causes mainly relies on manual experience judgment. Although these methods have a certain degree of applicability, their efficiency and accuracy are difficult to meet the actual needs. In recent years, methods for determining flight delay causes based on machine learning and deep learning have gradually emerged. Through feature extraction and modeling techniques, the intelligent level of the determination process has been improved. However, existing deep learning models mostly focus on single-level feature analysis and lack effective characterization of the multi-level and multi-scale characteristics of delay causes. At the same time, many methods use fixed feature dimensions and classification mechanisms in the inference stage and fail to dynamically adjust the classification accuracy and efficiency according to task requirements, making it difficult to provide reliable determination results in complex operating environments. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system, electronic device, and storage medium for hierarchical determination of flight delay causes based on multi-task metrics.

[0005] In a first aspect, the present invention provides a method for hierarchical determination of flight delay causes based on multi-task metrics, including:

[0006] Obtain civil aviation flight operation data and hierarchical labels for delay causes;

[0007] Data preprocessing;

[0008] Construct a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer;

[0009] Utilize multi-level nesting to construct a multi-task metric and hierarchical classification model;

[0010] Based on the multi-task metric and hierarchical classification model, train to obtain a delay cause classification model;

[0011] Perform hierarchical determination of flight delay causes through the delay cause classification model.

[0012] Second aspect, the present invention provides a hierarchical determination system for flight delay reasons based on multi-task metrics, including:

[0013] A data acquisition module, configured to obtain civil aviation flight operation data and delay hierarchical labels;

[0014] A data preprocessing module, configured to preprocess the data;

[0015] A nested feature construction module, configured to construct a nested feature vector space to achieve multi-level semantic nesting of flight delay reasons;

[0016] A multi-task metric and hierarchical classification model construction module, configured to construct a multi-task metric and hierarchical classification model by using multi-level nesting;

[0017] A delay reason classification model acquisition module, which trains and obtains a delay reason classification model based on the multi-task metric and hierarchical classification model;

[0018] A delay reason identification module, which hierarchically determines flight delay reasons through the delay reason classification model.

[0019] Third aspect, the present invention provides an electronic device, including: a processor and a memory; the processor is configured to execute computer execution instructions stored in the memory, and the processor runs the computer execution instructions to execute the above-mentioned hierarchical determination method for flight delay reasons based on multi-task metrics.

[0020] Fourth aspect, the present invention provides a storage medium, including a readable storage medium and a computer program stored in the readable storage medium, and the computer program is used to implement the above-mentioned hierarchical determination method for flight delay reasons based on multi-task metrics.

[0021] The beneficial effects of the present invention are as follows. The hierarchical determination method and system for flight delay reasons based on multi-task metrics of the present invention obtain civil aviation flight operation data and delay reason hierarchical labels; perform data preprocessing; construct a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer; use multi-level nesting to construct a multi-task metric and hierarchical classification model; train and obtain a delay reason classification model based on the multi-task metric and hierarchical classification model; hierarchically determine flight delay reasons through the delay reason classification model. It realizes efficient, automated, and accurate identification of specific flight delay reasons, can effectively guide subsequent refined management and optimization decisions, and improve the safety and orderliness of the transportation system. At the same time, by constructing a multi-task metric and hierarchical classification model, the adaptability and determination accuracy of the model at multiple scales and multiple semantic levels are significantly enhanced, ensuring the reliability and stability of the system in a complex operating environment.

[0022] Other features and advantages of the present invention will be described in the following specification, and in part will be obvious from the specification, or can be understood by implementing the present invention. The objectives and other advantages of the present invention are achieved and obtained by the structure specifically pointed out in the specification and the drawings.

[0023] In order to make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specifically presents preferred embodiments and, in conjunction with the accompanying drawings, provides a detailed description as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0025] Figure 1 is the principle flowchart of the method for hierarchical determination of flight delay causes based on multi-task measurement involved in the present invention;

[0026] Figure 2 is the schematic flowchart of the nested feature vector space model based on hypersphere involved in the present invention;

[0027] Figure 3 is the schematic flowchart of the multi-task measurement and hierarchical classification model under multi-level nesting involved in the present invention;

[0028] Figure 4 is the specific flowchart of the method for hierarchical determination of flight delay causes based on multi-task measurement involved in the present invention;

[0029] Figure 5 is the principle block diagram of the system for hierarchical determination of flight delay causes based on multi-task measurement involved in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] In order to make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions of the present invention in conjunction with the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0031] Embodiment

[0032] As Figure 1As shown in the figure, an embodiment of the first aspect of the present invention provides a hierarchical determination method for flight delay reasons based on multi-task metrics, including the following steps:

[0033] S110: Obtain civil aviation flight operation data and hierarchical labels for delay reasons;

[0034] S120: Data preprocessing;

[0035] S130: Construct a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer;

[0036] S140: Use multi-level nesting to construct a multi-task metric and hierarchical classification model;

[0037] S150: Based on the multi-task metric and hierarchical classification model, train to obtain a delay reason classification model;

[0038] S160: Perform hierarchical determination of flight delay reasons through the delay reason classification model.

[0039] In this embodiment, the civil aviation flight operation data obtained in step S110 specifically includes: flight basic information, actual takeoff and landing times, estimated and actual arrival times, delay times, hierarchical labels for delay reasons, etc.; the obtained hierarchical labels for delay reasons divide flight delay reasons into levels from coarse to fine. Among them, at the coarse level, delay reasons are classified into airport reasons, air traffic control reasons, airline reasons, traffic flow reasons, and other reasons to cover the main delay categories. On this basis, it is further refined to fine-level labels. For example, airport reasons can be specifically divided into ground service delays, taxiway congestion, etc., and airline reasons can be divided into mechanical failures, crew errors, etc., finally forming a gradually progressive label system to provide clear semantic guidance for the accurate determination and refined classification of delay reasons.

[0040] After obtaining the data, it is also possible to check the integrity of the data and supplement the data for missing time periods or routes.

[0041] In this embodiment, step S120 specifically includes: processing the collected civil aviation flight operation data, detecting and filtering outliers and duplicate data in the data to obtain cleaned data; using a dimensionality reduction method, combined with the hierarchical characteristics of flight delay reasons, to select key features related to flight delay reason judgment; annotating the hierarchical delay reason labels of the dataset samples; constructing a flight delay reason determination dataset and dividing the dataset into training and validation datasets; performing standardization or normalization processing on numerical features to eliminate the influence of dimensions, and numericalizing categorical features.

[0042] Such as Figure 2As shown, in this embodiment, step S130 specifically includes: constructing a nested feature vector space, dividing the features of flight operation data into multiple subspaces, where the feature vector of each sample is expressed as:

[0043]

[0044] where d represents the total dimension of the feature vector. A feature arrangement strategy that progresses step by step from coarser to finer is adopted. The feature is defined as the coarsest-grained feature of the sample, usually containing important global information shared among multiple categories, while the feature is located at the end of the feature vector and serves as the finest feature for capturing subtle differences between specific categories. This strategy places the features common across categories at the front of the vector, while the more discriminative and finer features are placed at the end of the vector.

[0045] Furthermore, for the semantic expression of hierarchical delay reasons, each subspace is defined to correspond to a specific level:

[0046]

[0047] where H represents the total number of semantic levels; h represents the current level; represents the feature sub-vector of the h-th level, containing the features from the 1st to the dimension.

[0048] The coarse-grained features are a subset of the fine-grained features, and the features at each level satisfy the following nested relationship:

[0049]

[0050] Through this progressive relationship, the coarse-grained features provide context support for the fine-grained features, ensuring the semantic consistency of multi-level features.

[0051] Furthermore, based on the nested feature vector, hyper-sphere mapping is used for geometric optimization. The feature vector is normalized to the surface of the hyper-sphere to obtain a unit vector:

[0052]

[0053] The feature satisfies , avoiding the influence of feature amplitude differences on classification determination. The contributions of all level features are only determined by directionality, making the features more regularly distributed in the geometric space. Through the above construction of the nested feature vector space and hyper-sphere mapping optimization, an efficient and clearly hierarchical feature input is obtained.

[0054] Such as Figure 3As shown in the figure, in step S140 of this embodiment, based on the nested feature vector space, a dynamic measurement mechanism based on cosine similarity is introduced to measure the similarity between feature vectors. Its core calculation formula is:

[0055]

[0056] Wherein, and represent the nested feature vectors after hypersphere normalization. Through the directional calculation of cosine similarity, the interference of feature amplitude on similarity measurement can be effectively avoided, and at the same time, the direction consistency of features is ensured.

[0057] Furthermore, in order to incorporate multi-level nested features into the classification model, a hierarchical classification task framework is constructed, and a loss function is independently designed for each level of the task. For each level h, the classification objective of the model is to maximize the confidence of the class to which the sample belongs. A multi-level loss function based on hypersphere normalization is adopted:

[0058]

[0059] Wherein, represents the unit vector of class c; represents the dynamic scale factor, which regulates the determination range of each level; represents the normalized feature vector of sample i at level h; represents the class weight vector; represents the set of classes at level h.

[0060] To achieve the collaborative optimization of multi-level tasks, the loss functions of each level are added and fused into the total loss function:

[0061]

[0062] As Figure 4 shown, in step S150 of this embodiment, based on the multi-task measurement and hierarchical classification model, that is, using the nested feature space constructed in the training stage, the multi-level features of flight delay reasons are gradually nested in the same vector, and combined with the hierarchical multi-task learning framework, the coarse-grained and fine-grained classification tasks are optimized respectively, and the multi-scale classification information is integrated in the final result to provide a complete determination of flight delay reasons; training the delay reason classification model on the flight operation training data set with multi-scale labels; evaluating the classification performance of the model on the independent validation set, and calculating the hierarchical classification effect of the model in judging flight delay reasons.

[0063] In this embodiment, step S160 specifically includes: constructing an independent validation set, including various feature scenarios of flight operations and delay data, covering delay reason labels at different granularities; mapping flight operation data into the constructed nested feature vector space to extract its corresponding multi-scale features; within the full classification task scope from coarse to fine granularity, verifying the classification ability of the model at each level; dynamically invoking the hierarchical classification logic to determine the delay reasons layer by layer from low dimension to high dimension; at the coarse granularity level, quickly identifying the coarse-level reasons for delays; invoking higher-dimensional feature vectors for further refined classification; ultimately realizing a hierarchical determination system for flight delay reasons that can dynamically adapt to different scenarios in an actual complex application environment.

[0064] Based on the above embodiments, the second aspect embodiment of the present invention provides a hierarchical determination system for flight delay reasons based on multi-task measurement, including: a data acquisition module for obtaining civil aviation flight operation data and delay hierarchical labels; a data preprocessing module for preprocessing the data; a nested feature construction module for constructing a nested feature vector space to achieve multi-level semantic nesting of flight delay reasons; a multi-task measurement and hierarchical classification model construction module for constructing a multi-task measurement and hierarchical classification model by using multi-level nesting; a delay reason classification model acquisition module for training to obtain a delay reason classification model based on the multi-task measurement and hierarchical classification model; and a delay reason identification module for hierarchically determining flight delay reasons through the delay reason classification model.

[0065] In this embodiment, the hierarchical determination system for flight delay reasons based on multi-task measurement can be implemented by using the above-mentioned method for hierarchical determination of flight delay reasons based on multi-task measurement; the implementation steps of each module can be referred to as above and will not be elaborated here.

[0066] Based on the above embodiments, the third aspect embodiment of the present invention provides an electronic device, including: a processor and a memory; the processor is used to execute the computer execution instructions stored in the memory, and the processor runs the computer execution instructions to execute the above-mentioned method for hierarchical determination of flight delay reasons based on multi-task measurement.

[0067] Based on the above embodiments, the third aspect embodiment of the present invention provides a storage medium, including a readable storage medium and a computer program stored in the readable storage medium, and the computer program is used to implement the above-mentioned method for hierarchical determination of flight delay reasons based on multi-task measurement.

[0068] In summary, the method and system for hierarchical determination of flight delay causes based on multi-task metric of the present invention obtain civil aviation flight operation data and hierarchical labels of delay causes; perform data preprocessing; construct a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer; utilize multi-level nesting to construct a multi-task metric and a hierarchical classification model; and based on the multi-task metric and the hierarchical classification model, realize the hierarchical automatic determination of flight delay causes by training a delay cause classification model. It realizes the efficient, automatic, and accurate identification of specific flight delay causes, can effectively guide subsequent refined management and optimization decisions, and improve the safety and orderliness of the transportation system. At the same time, by constructing a multi-task metric and a hierarchical classification model, the adaptability and determination accuracy of the model at multi-scale and multi-semantic levels are significantly enhanced, ensuring the reliability and stability of the system in a complex operating environment.

[0069] In several embodiments provided by the present application, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, the program segment, or the part of code includes one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the block may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0070] In addition, each functional module in various embodiments of the present invention may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0071] When the above-mentioned functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0072] Inspired by the above ideal embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A hierarchical determination method for flight delay reasons based on multi-task metrics, characterized in that Including: Obtain civil aviation flight operation data and hierarchical labels for delay reasons; Data preprocessing; Construct a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer; Utilize multi-level nesting to construct a multi-task metric and hierarchical classification model; Based on the multi-task metric and hierarchical classification model, train to obtain a delay reason classification model; Through the delay reason classification model, conduct hierarchical determination of flight delay reasons; Wherein The method for obtaining civil aviation flight operation data and hierarchical labels for delay reasons includes: Determine the required data types, including flight basic information, actual takeoff and landing times, estimated and actual arrival times, delay times, and hierarchical labels for delay reasons; the obtained hierarchical labels for delay reasons divide flight delay reasons into coarser to finer levels. At the coarser level, delay reasons are classified into airport reasons, air traffic control reasons, airline reasons, traffic flow reasons, and other reasons to cover the main delay categories; further refined to finer-level labels, the airport reasons are specifically divided into ground service delays and taxiway congestion, and the airline reasons are divided into mechanical failures and crew errors, finally forming a progressive label system to provide clear semantic guidance for the accurate determination and refined classification of delay reasons; check the integrity of the data and supplement missing time periods or route data; The method for constructing a nested feature vector space based on hyperspheres to express multi-scale semantics layer by layer includes: Construct a nested feature vector space, divide the features of flight operation data into multiple sub-spaces, and the feature vector of each sample is represented as: v i = [v i,1 , v i,2 , …, v i,d ​ Among them, d represents the total dimension of the feature vector; v i Adopted a feature arrangement strategy that progresses step by step from coarse to fine; the feature v i,1 Is defined as the coarsest feature of the sample, containing important global information shared among multiple categories; the feature v i,d Is located at the end of the feature vector and serves as the finest feature for capturing subtle differences between specific categories; For the semantic expression of hierarchical delay reasons, define that each sub-space corresponds to a specific level: Among them, H represents the total number of semantic levels, and h represents the current level. represents the eigenvector of the h-th level, which contains the features from the 1st to the d-th h dimension. The coarse-grained features are a subset of the fine-grained features, and the features at each level satisfy the following nested relationship: Through this progressive relationship, the coarse-grained features provide context support for the fine-grained features; Based on the nested feature vector, use hypersphere mapping for geometric optimization, normalize the feature vector to the surface of the hypersphere to obtain a unit vector: The feature satisfies ||v i || = 1; The method for constructing a multi-task metric and hierarchical classification model based on multi-level nesting includes: Based on the nested feature vector space, introduce a dynamic metric mechanism based on cosine similarity to measure the similarity between feature vectors; its core calculation formula is: Among them, and represent the nested feature vectors after hypersphere normalization; In the multi-level nested feature space, cluster similar category samples closer and at the same time widen the distance between different category samples; the constraint conditions for multi-level nested features in classification are: This formula indicates that in the feature space of the h-th level, for samples belonging to the same category and their similarity is higher than that of samples belonging to different categories and Integrate the multi-level nested features into the classification model, construct a hierarchical classification task framework, and independently design a loss function for each level of the task; for each level h, the classification objective of the model is to maximize the confidence of the sample belonging to the category; adopt a multi-level loss function based on hypersphere normalization: Among them, represents the unit vector of category c; σ h represents the dynamic scale factor, which regulates the determination range of each level; represents the normalized feature vector of sample i at level h; represents the weight vector of category y h ; c h represents the category set at level h; Add and fuse the loss functions of each level into a total loss function: The method for training to obtain a delay reason classification model based on the multi-task metric and hierarchical classification model includes: Use a multi-task metric and hierarchical classification model, that is, utilize the nested feature space constructed in the training phase to gradually nest the multi-level features of flight delay reasons into the same vector, and combine a hierarchical multi-task learning framework to optimize the coarse-grained and fine-grained classification tasks respectively, and integrate the multi-scale classification information in the final result to provide a complete determination of flight delay reasons; Train a delay reason classification model on a flight operation training data set with multi-scale labels; Evaluate the classification performance of the model on an independent validation set, and calculate the hierarchical classification effect of the model in judging flight delay reasons.

2. The hierarchical determination method for flight delay reasons based on multi-task metric according to claim 1, characterized in that, The method for data preprocessing includes: Process the collected flight operation data, detect and filter outliers and duplicate data in the data to obtain the cleaned data; Use a dimensionality reduction method, combined with the hierarchical characteristics of flight delay reasons, to select key features related to flight delay reason judgment; Label the hierarchical delay reason labels of the data set samples; Construct a flight delay reason determination data set, and divide the data set into training and validation data sets; Perform standardization or normalization processing on numerical features to eliminate the influence of dimension, and numericalize categorical features.

3. The hierarchical determination method for flight delay reasons based on multi-task metric according to claim 2, characterized in that, The method for hierarchically determining flight delay reasons through a delay reason classification model includes: Use a delay reason classification model, that is Construct an independent validation set, including various feature scenarios of flight operation and delay data, covering delay reason labels of different granularities; Map the flight operation data to the constructed nested feature vector space, and extract its corresponding multi-scale features; Dynamically call the hierarchical classification logic to complete the determination of flight delay reasons layer by layer from low dimension to high dimension; at the coarse-grained level, quickly identify the coarse-level reasons for the delay; call higher-dimensional feature vectors for further refined classification; Finally, realize the hierarchical determination of flight delay reasons that can dynamically adapt to different scenarios in the actual complex application environment.

4. A hierarchical determination system for flight delay causes based on multi-task metrics, which adopts the hierarchical determination method for flight delay causes based on multi-task metrics as described in any one of claims 1-3, characterized in that It includes: A data acquisition module for obtaining civil aviation flight operation data and delay hierarchical labels; A data preprocessing module for preprocessing the data; A nested feature construction module for constructing a nested feature vector space to realize the multi-level semantic nesting of flight delay reasons; A multi-task metric and hierarchical classification model construction module for constructing a multi-task metric and hierarchical classification model by using multi-level nesting; A delay reason classification model acquisition module, based on the multi-task metric and hierarchical classification model, trains to obtain a delay reason classification model; A delay reason identification module for hierarchically determining flight delay reasons through a delay reason classification model.

5. An electronic device, characterized in that, It includes: A processor and a memory; The processor is used to execute the computer execution instructions stored in the memory, and the processor runs the computer execution instructions to execute the hierarchical determination method for flight delay reasons based on multi-task metric according to any one of claims 1-3.

6. A storage medium, characterized in that, Comprising a readable storage medium and a computer program stored in the readable storage medium, the computer program being used to implement the method for hierarchical determination of flight delay causes based on multitask metrics according to any one of claims 1-3.

Images