Civil aviation emergency management status analysis and improvement method

Abstract

The invention discloses a civil aviation emergency management current situation analysis and improvement method. The method comprises the following steps: acquiring civil aviation emergency management questionnaire data; preprocessing the questionnaire data to obtain preprocessed data; performing document category division on the preprocessed data, and performing document clustering and visualization on the divided documents to form a plurality of categories; extracting document keywords from different types of documents, constructing a co-occurrence network, calculating network indexes, performing feature analysis, and identifying structural features of the network; and calculating node indexes according to the identified structural features of the network, sorting node importance based on the node indexes, and identifying key nodes. The civil aviation emergency management capability is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of civil aviation emergency management, and in particular relates to a method for analyzing and improving the current status of civil aviation emergency management. Background Art

[0002] When a major emergency occurs in civil aviation, it will have a serious impact on the regional economy, and the diversity of events greatly affects the decision-making efficiency of the emergency management system. Compared with conventional management, emergency management faces more time pressure, resource constraints and decision-making consequence pressure. Therefore, it is necessary to effectively collect, integrate, mine and apply various emergency information resources to support emergency management decisions and improve emergency management capabilities. At present, scholars at home and abroad have applied artificial intelligence and information technology to civil aviation emergency management, involving emergency response decision-making methods, complex terrain rescue, and emergency drill simulation platform construction. Text mining technology is an effective method for obtaining information and knowledge, and is a research hotspot in the field of data mining. The LDA model proposed by Blei is a basic method of text mining and has been widely used. Jingqi Dai et al. used the LDA model to analyze the relationship and development trend between major emergency topics and public sentiment. Jintao Wang first proposed a method for constructing a knowledge graph for civil aviation emergencies. Based on the BERT model, Google proposed an improved deep learning model ALBERT, which has achieved results in text classification, knowledge graph construction, sentiment analysis, and named entity recognition. Cheng Ming et al. combined association rules with complex network models to study the mechanism of the evolution process of runway landing accidents. Currently, there is little research on the combination of text mining technology and civil aviation emergency management; the text mining LDA model fails to consider contextual semantic information and the correlation between words and documents, and cannot guarantee the topic quality under large-scale data sets. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a method for analyzing and improving the current status of civil aviation emergency management, thereby improving the emergency management capabilities of civil aviation.

[0004] To achieve the above-mentioned purpose, the present invention provides a method for analyzing and improving the current status of civil aviation emergency management, comprising: obtaining civil aviation emergency management questionnaire data;

[0005] Preprocessing the questionnaire data to obtain preprocessed data;

[0006] Divide the preprocessed data into document categories, cluster and visualize the divided documents to form several categories;

[0007] Extract keywords from documents of different categories, build co-occurrence networks, calculate network indicators, perform feature analysis, and identify the structural features of the network;

[0008] The node index is calculated according to the structural characteristics of the identified network, the node importance is sorted based on the node index, and the key nodes are identified.

[0009] Technical effect of the invention: The invention discloses a method for analyzing and improving the current status of civil aviation emergency management, using ALBERT and LDA models to classify documents, combining K-means and UMAP methods to cluster and visualize documents, constructing a keyword co-occurrence network based on complex network theory, analyzing network characteristics and node connections, and applying the TOPSIS model to obtain node importance ranking. The invention improves the emergency management capabilities of civil aviation and provides data support for the construction of a civil aviation emergency system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0011] Figure 1 This is a distribution diagram of the sixth theme of the embodiment of the present invention;

[0012] Figure 2 This is a distribution diagram of the subject matter of Embodiment 7 of the present invention;

[0013] Figure 3 This is a schematic diagram of two-dimensional clustering visualization based on UMAP according to an embodiment of the present invention;

[0014] Figure 4 This is the first category keyword network diagram of the embodiment of the present invention;

[0015] Figure 5 This is the second category keyword network diagram of the embodiment of the present invention;

[0016] Figure 6 This is the third category keyword network diagram of the embodiment of the present invention;

[0017] Figure 7 This is the fourth category keyword network diagram of the embodiment of the present invention;

[0018] Figure 8 This is the fifth category keyword network diagram of the embodiment of the present invention;

[0019] Fig. 9 This is the sixth category keyword network diagram of the embodiment of the present invention;

[0020] Fig.10 This is a weighted degree value distribution diagram of an embodiment of the present invention;

[0021] Fig.11 This is a distribution diagram of betweenness centrality of an embodiment of the present invention;

[0022] Fig.12 It is a close centrality distribution diagram of an embodiment of the present invention;

[0023] Fig.13 The present invention is a flowchart of a method for analyzing and improving the current status of civil aviation emergency management according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] like Figure 1 As shown, this embodiment provides a method for analyzing and improving the current status of civil aviation emergency management, including: obtaining civil aviation emergency management questionnaire data;

[0027] Preprocessing the questionnaire data to obtain preprocessed data;

[0028] Divide the preprocessed data into document categories, cluster and visualize the divided documents to form several categories;

[0029] Extract keywords from documents of different categories, build co-occurrence networks, calculate network indicators, perform feature analysis, and identify the structural features of the network;

[0030] The node index is calculated according to the structural characteristics of the identified network, the node importance is sorted based on the node index, and the key nodes are identified.

[0031] Specifically, the co-occurrence network reveals the potential connections between words by analyzing the co-occurrence frequency between keywords. The indicators of the co-occurrence network include network indicators and node indicators. Network indicators reflect the structural characteristics of the entire network; node indicators reflect the importance, connectivity and influence of a single node in the overall network. Key nodes refer to several nodes that have a high degree of influence on the overall network.

[0032] Furthermore, the preprocessed data includes:

[0033] Cleaning the questionnaire data and removing invalid text;

[0034] Based on Sogou Cell Dictionary, we supplemented the professional vocabulary of civil aviation emergency management and built a custom dictionary;

[0035] The questionnaire data was preprocessed by segmenting words, removing stop words, and regularizing special symbols, spaces, and punctuation in the text.

[0036] Furthermore, the document category classification of the preprocessed data includes:

[0037] The preprocessed data is divided into document categories according to semantic features and topic features.

[0038] Furthermore, classifying the preprocessed data into document categories according to semantic features includes:

[0039] The ALBERT model is used to extract semantic feature vectors. The word embedding vector is calculated and input into the albert-base-v2 pre-trained model to vectorize the text data and extract the vector results.

[0040] Furthermore, classifying the preprocessed data into document categories according to subject features includes:

[0041] The LDA model is used to extract topic feature vectors, the PyLDAVis package is used to determine the optimal number of topics, and the topic interactive map is drawn.

[0042] Furthermore, the divided documents are clustered and visualized to form several categories including:

[0043] Using a concatenate function to concatenate the semantic feature vector and the topic feature vector in a high-dimensional vector space;

[0044] An autoencoder is used to concatenate two sets of vectors in a low-dimensional vector space of a preset dimension to obtain a low-dimensional representation with concentrated information;

[0045] The K-means algorithm is used to assign words with similar semantics and themes to the same category, determine the number of clusters, and form several categories;

[0046] The number of clusters determined by the UMAP algorithm is visualized.

[0047] Furthermore, constructing a co-occurrence network includes:

[0048] According to several categories, the co-occurrence network was constructed using Gephi, with the extracted document keywords as network nodes and the co-occurrence frequencies between document keywords as edges.

[0049] Furthermore, the network metrics calculated include:

[0050] The network density ρ is calculated as:

[0051]

[0052] Among them, M is the number of edges actually existing in the network, and n is the number of nodes in the network;

[0053] The average path length L is calculated as:

[0054]

[0055] Among them, d ij is the shortest path length between node i and node j, n is the number of nodes in the network;

[0056] The average clustering coefficient C is calculated as:

[0057]

[0058] Where N is the total number of nodes in the network, k i is the number of nodes adjacent to node i, E i is the number of edges between node i and its adjacent nodes.

[0059] Further, calculating the node index according to the structural characteristics of the identified network includes:

[0060] Calculate the node weighted degree M i for:

[0061]

[0062] Where n is the number of nodes in the network, a ij is an element in the adjacency matrix, indicating whether there is an edge between node i and node j;

[0063] Calculate betweenness centrality as:

[0064]

[0065] Among them, F sv (i) is the probability of the shortest path between nodes s and v, X sv (i) is the number of paths passing through node i in the shortest path between nodes s and v, X sv is the number of nodes s and the shortest path;

[0066] Calculate closeness centrality as:

[0067]

[0068] Where C(x) is the closeness centrality of node x, and ∑(x,y) is the sum of the shortest paths from node x to all other nodes y.

[0069] As a technical means to increase

[0070] The TOPSIS method can be used to reflect the importance of nodes and highlight key nodes. ip Represents the pth characteristic index value of the i-th node, where i = 1, 2, 3…n, p = 1, 2, 3…m. First, construct the node characteristic index matrix:

[0071]

[0072] Since the node attributes are different, the matrix is ​​standardized:

[0073]

[0074] The matrix after normalization is:

[0075]

[0076] In order to further reasonably allocate the indicator weights, the average weighted method is used to normalize the matrix:

[0077]

[0078] Finally, the TOPSIS model is used to evaluate the importance of nodes. The formula is as follows:

[0079]

[0080] are the minimum and maximum values ​​of each column of matrix Q respectively. i The size of the node can be used to determine the importance of the node. The larger the value, the more important it is to the network.

[0081] A specific application example of the present invention is as follows:

[0082] The data in this paper are derived from the feedback from civil aviation emergency management personnel collected through questionnaires, including the problems existing in the current emergency work and suggestions for improvement.

[0083] The present invention cleans the text information, removes invalid text, and obtains a total of 1,342 valid information. Based on the Sogou Cell Dictionary, the professional vocabulary of civil aviation emergency management is supplemented to build a custom dictionary. Special symbols, spaces and punctuation in the text are processed by word segmentation, stop word removal, and regularization to improve data quality and consistency.

[0084] The effective division of document categories can be carried out based on semantic features and topic features. The semantic feature vector is extracted by the ALBERT model, and the word embedding vector is calculated and loaded into the albert-base-v2 pre-trained model to vectorize the text data and extract a 1342×768-dimensional vector result; the topic feature vector is extracted by the LDA model, and the PyLDAVis package is used to determine the optimal number of topics and draw a topic interactive map.

[0085] The interactive map can intuitively show the effect of topic generation. The size and distance of the bubbles indicate the frequency and proximity of the topic. If there is no overlap of bubbles or the overlap accounts for less than 2%, it means that the distribution of topics is quite different, and the topic generation effect is good, otherwise it is poor.

[0086] like Figure 1 As shown, when the number of topics is 6, the overlapping part accounts for less than 2%, indicating that the topic distribution is highly different and the generation effect is better; Figure 2 As shown, when the number of topics is 7, the overlapping part accounts for more than 5%, indicating that the difference in topic distribution is small and the generation effect is poor. Therefore, the number of topics is determined to be 6. After determining that the number of topics is 6, based on the bag-of-words model, the LDAmodel in Python is used to vectorize the data and extract the 1342×6-dimensional topic feature vector results. The semantic and topic feature vectors have high dimensions. To ensure the integrity of the information, it is necessary to convert the high-dimensional vector space to a low-dimensional vector space representation. The present invention uses the concatenate function to splice two feature vectors in a high-dimensional vector space; and uses an autoencoder to concatenate two sets of vectors in an 8-dimensional low-dimensional vector space to obtain a low-dimensional representation with condensed information.

[0087] In order to further obtain structured information expression, the present invention uses the K-means algorithm to assign words with similar semantics and themes to the same category. According to relevant literature, the number of clusters is determined to form 6 categories. In order to clearly and intuitively illustrate the clustering effect, the present invention uses the UMAP (Uniform Manifold Approximation and Projection) algorithm to visualize the topic results. The coordinate axis in the UMAP diagram represents the projection position of the data point in the two-dimensional space, and the uniformity of the data point distribution can indicate the quality of the clustering effect. Figure 3 As shown, the number of data points in categories 1 to 6 is not much different, with the smallest accounting for 13.19% and the largest accounting for 20.12%. This shows that the density distribution of themes in space is relatively uniform, avoiding the phenomenon of over-concentrated clustering, and the clustering effect is good.

[0088] Building a co-occurrence network is helpful for in-depth mining of text information and quantitative analysis of research results. Based on the clustering results in the previous article, we use keywords as network nodes and the co-occurrence frequency between keywords as edges, and use Gephi to build a network, such as Figure 4-Figure 9 shown.

[0089] In network parameter analysis, Figure 4-9 As shown in Table 1, the second type has the largest number of edges, and the overall network density is less than 0.598, which means the connection is relatively tight. The average clustering coefficient exceeds 0.76, indicating that there are strong local connections between nodes, forming a tight community structure. The average path length is between 1.4 and 3, indicating that on average each node can affect other nodes through at least 1.4 units, reflecting the good connectivity of the network. The network covers civil aviation units such as "air traffic control" and "feeder airports"; it covers emergency management elements such as "special emergency plans", "professional training", and "emergency rescue equipment", providing rich information for key node analysis. Emergency management elements refer to the key components that constitute the overall function and operation of the emergency management system.

[0090] Table 1

[0091]

[0092] Node parameter analysis:

[0093] (1) Weighted degree, such as Fig.10 As shown, the larger the weighted value, the stronger the influence of the node. The results show that category 2 has the highest weighted value (718), and the node label is "Plan", indicating that this node has a strong influence on other nodes and needs to be focused on. If the plan is not well managed, it is easy to cause subsequent problems such as poor exercise results and handling errors, increasing the possibility of risk events. 2024 Civil Aviation Emergency Management Forum

[24] It emphasizes the necessity of focusing on compiling institutional documents for emergency plans and improving the industry plan system, indicating that the research results are consistent with the actual situation.

[0094] (2) Betweenness centrality, such as Fig.11 As shown in the figure, the higher the betweenness centrality, the more important the node's intermediary role. The results show that the "airport" node in category 5 has the highest betweenness centrality, which is 0.225, indicating that this node has a high intermediary effect in network node conduction. In actual operations, the civil aviation management department regards airports as an important link in information exchange and resource allocation, and incorporates them into the local emergency rescue system, which shows that the analysis results are consistent with the actual situation.

[0095] (3) Close centrality

[0096] like Fig.12As shown in the figure, close centrality reflects the position of a node in the network, and the larger the value, the closer it is to the center of the network. In practice, this type of node has the most connections with other nodes and can reach other parts of the network faster. The results show that the "part-time" node in category 2 has the highest close centrality of 0.86, which indicates that it is at the core of the network. In emergency management work, part-time work of personnel is likely to cause conflicts in performance of duties and reduce work quality, so "part-time" nodes are particularly important in the network.

[0097] (4) Order of importance

[0098] According to the TOPSIS model, the three parameters of weighted degree, betweenness centrality and closeness centrality are considered, and the comprehensive value is calculated to describe the importance of the node. As shown in Table 2, the "part-time" node has the highest importance, which is 0.673, indicating that it has a decisive influence in the civil aviation emergency management network, which is consistent with the closeness centrality analysis and is the most important content.

[0099] Table 2

[0100]

[0101] Based on the analysis of node weighted degree, betweenness centrality, and closeness centrality, the present invention discusses the research results as follows:

[0102] The weighted value of the plan node is the largest and has the strongest influence in the network. In practice, the plan plays an important role in preventing and controlling operational risks and is highly valued by managers. Therefore, attention and investment in the plan should continue to be strengthened to improve response capabilities and ensure operational safety.

[0103] Airports have a high intermediary effect. In civil aviation emergency management, maintaining effective links with airlines and local governments is an important link in ensuring the effectiveness of emergency response and resource allocation. Therefore, management agencies should give priority to airport emergency work to enhance systematic emergency management capabilities.

[0104] Part-time workers are at the core of the network, with the largest number of nodes connected to them, and can affect the overall network the fastest. In reality, a good division of labor can clarify personnel responsibilities and ensure reasonable resource allocation. Improving the degree of specialization of corporate emergency management agencies will help improve emergency work efficiency and handling capabilities.

[0105] According to the order of node importance and in combination with actual conditions, the key contents of civil aviation emergency management are divided into five categories, with node labels in brackets:

[0106] Part-time staff (part-time)

[0107] At the enterprise level: Emergency management personnel are usually part-time business personnel and have not received systematic training. Emergency work at small and medium-sized airports is mainly done part-time by the aviation department and the control room, which can easily lead to conflicts in performance of duties and poor work quality, restricting the improvement of emergency management capabilities.

[0108] Emergency planning and response (plans, emergencies)

[0109] Enterprise level: There are too many emergency plans, but the quality and feasibility are not guaranteed. Some units do not pay enough attention to emergency plan management, resulting in insufficient personnel and funds, which affects the implementation of related work.

[0110] Organization and coordination agencies (bureaus, local governments)

[0111] System level: There are certain difficulties in coordination between the regulatory sub-bureau and the local government emergency management department. Once an emergency occurs to an aircraft, such as a fall during agricultural chemical operations, it is difficult for the local regulatory bureau to effectively connect with the local government in terms of information sharing and coordination.

[0112] Industry participants (enterprises, institutions, airlines, regional airports, air traffic control)

[0113] System level: Insufficient coordination among emergency management participants, inconsistent priorities between emergency rescue and emergency response. Insufficient coordination between enterprises and the government leads to unclear management responsibilities for complex emergencies.

[0114] Standard Specification (Specification)

[0115] Industry norms: There are relatively few laws and regulations on civil aviation emergency response, and there is a lack of specific guidance on risk assessment standards, plan coordination and liaison mechanisms, which affects the implementation of emergency response work.

[0116] In response to the above problems, the following suggestions are put forward:

[0117] From the perspective of the enterprise: Establish a full-time emergency management department and clarify its responsibilities; regularly review and evaluate existing plans, organize drills to test their effectiveness; regularly conduct personnel training; introduce information technology, establish an emergency management information platform, monitor and evaluate risks in real time, reduce dependence on part-time personnel, and improve emergency management efficiency.

[0118] From a system perspective: Establish a cross-departmental coordination mechanism, hold regular coordination meetings, ensure information sharing and work coordination among departments and units; establish an emergency training platform, strengthen government-enterprise cooperation; refine the leading and cooperating departments for emergencies, clarify the operating procedures, and ensure effective docking and coordinated handling.

[0119] Industry perspective: Promote the improvement of laws and regulations related to civil aviation emergency management, formulate industry standards and operational guidelines for risk assessment, plan coordination, etc., and provide reference for various units.

[0120] The present invention discloses a method for analyzing and improving the current status of civil aviation emergency management, using ALBERT and LDA models to classify documents, combining K-means and UMAP methods to cluster and visualize documents, constructing a keyword co-occurrence network based on complex network theory, analyzing network characteristics and node connections, and applying the TOPSIS model to obtain node importance ranking. The average clustering coefficient of the network in the present invention is 0.76 to 0.971, covering different departments, positions and environments, representing the general situation of civil aviation emergency management; part-time staff, emergency plans and responses, and organization and coordination agencies are key contents that need to be paid attention to, and it is recommended that the civil aviation system improve its emergency management capabilities by establishing a full-time emergency management agency, strengthening system coordination, and improving laws and regulations. The present invention improves the emergency management capabilities of civil aviation and provides data support for the construction of a civil aviation emergency system.

[0121] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for analyzing and improving the current status of civil aviation emergency management, characterized in that: include: Obtain civil aviation emergency management questionnaire data; Preprocessing the questionnaire data to obtain preprocessed data; Divide the preprocessed data into document categories, cluster and visualize the divided documents to form several categories; Extract keywords from documents of different categories, build co-occurrence networks, calculate network indicators, perform feature analysis, and identify the structural features of the network; The node index is calculated according to the structural characteristics of the identified network, the node importance is sorted based on the node index, and the key nodes are identified.

2. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 1, characterized in that: The preprocessed data include: Cleaning the questionnaire data and removing invalid text; Based on Sogou Cell Dictionary, we supplemented the professional vocabulary of civil aviation emergency management and built a custom dictionary; The questionnaire data was preprocessed by segmenting words, removing stop words, and regularizing special symbols, spaces, and punctuation in the text.

3. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 1, characterized in that: Document classification of preprocessed data includes: The preprocessed data is divided into document categories according to semantic features and topic features.

4. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 3, characterized in that: Classifying the preprocessed data into document categories according to semantic features includes: The ALBERT model is used to extract semantic feature vectors. The word embedding vector is calculated and input into the albert-base-v2 pre-trained model to vectorize the text data and extract the vector results.

5. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 4, characterized in that: Classifying the preprocessed data into document categories according to subject features includes: The LDA model is used to extract topic feature vectors, the PyLDAVis package is used to determine the optimal number of topics, and the topic interactive map is drawn.

6. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 5, characterized in that: The divided documents are clustered and visualized to form several categories including: Using a concatenate function to concatenate the semantic feature vector and the topic feature vector in a high-dimensional vector space; An autoencoder is used to concatenate two sets of vectors in a low-dimensional vector space of a preset dimension to obtain a low-dimensional representation with concentrated information; The K-means algorithm is used to assign words with similar semantics and themes to the same category, determine the number of clusters, and form several categories; The number of clusters determined by the UMAP algorithm is visualized.

7. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 1, characterized in that: Constructing a co-occurrence network involves: According to several categories, the co-occurrence network was constructed using Gephi, with the extracted document keywords as network nodes and the co-occurrence frequencies between document keywords as edges.

8. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 1, characterized in that: Calculated network metrics include: The network density ρ is calculated as: Among them, M is the number of edges actually existing in the network, and n is the number of nodes in the network; The average path length L is calculated as: Among them, d ij is the shortest path length between node i and node j, n is the number of nodes in the network; The average clustering coefficient C is calculated as: Where N is the total number of nodes in the network, k i is the number of nodes adjacent to node i, E i is the number of edges between node i and its adjacent nodes.

9. The method for analyzing and improving the current status of civil aviation emergency management as claimed in claim 1, characterized in that: Calculate node indicators based on the structural characteristics of the identified network. Calculate node indicators including: Calculate the node weighted degree M i for: Where n is the number of nodes in the network, a ij is an element in the adjacency matrix, indicating whether there is an edge between node i and node j; Calculate betweenness centrality as: Among them, F sv (i) is the probability of the shortest path between nodes s and v, X sv (i) is the number of paths passing through node i in the shortest path between nodes s and v, X sv is the number of nodes s and the shortest path; Calculate closeness centrality as: in, C(x) is the closeness centrality of node x, and ∑(x,y) is the sum of the shortest paths from node x to all other nodes y.

Images