Runway landing event risk analysis method based on association rules and complex networks

Abstract

This invention discloses a method for runway landing event risk analysis based on association rules and complex networks, comprising the following steps: extracting risk factor keywords from runway landing accident investigation reports using Jieba word segmentation technology; constructing a runway landing event risk factor dataset based on the risk factor keywords; performing data mining processing on the runway landing event risk factor dataset using an association rule mining algorithm to obtain frequent itemsets and association rules; constructing a complex network based on frequent itemsets and association rules; the nodes of the complex network are frequent itemsets, and the edges of the complex network are the association rules; determining key risk nodes based on node degree and node centrality in the complex network, and marking them as key risk factors for runway landing events; and implementing prevention and control measures based on the key risk factors. This invention combines association rule mining and complex networks to analyze key risk factors for runway landing events and quantitatively calculates and analyzes the criticality of risk nodes.

Description

Technical Field

[0001] This invention belongs to the field of aviation safety technology, and in particular relates to a method for runway landing event risk analysis based on association rules and complex networks. Background Technology

[0002] In recent years, with the increase in passenger traffic, airports in various places have built multiple runways to increase transportation capacity, and the number of airports with multiple runways has been increasing. However, according to civil aviation accident statistics, most aviation accidents occur on runways. As an important landing point for aircraft on the ground, the runway is one of the key parts of the flight process. Therefore, a series of runway safety issues related to aircraft operation have gradually become hot issues in the aviation industry and academia.

[0003] Complex network theory plays an important role in studying the evolution of accidents. Its earliest research originated from graph theory in mathematics, which uses nodes to represent the elements of a system and lines between two points to represent the interactions between elements, thus studying complex networks from a mathematical perspective.

[0004] Existing academic research on runway landing events is substantial, but for major overrun and hard landing incidents, most studies focus on a single category rather than considering the overall correlation between overrun and hard landing events. However, practical experience shows that hard landings are often accompanied by runway overruns. Furthermore, complex networks are primarily applied in shipping and railway operations; in aviation, their application is limited to studying aviation network topology, with little focus on aviation accident or incident analysis. In addition, most research on overrun and hard landing focuses on influencing factors and risk assessment, with limited research on internal causes and evolutionary processes. Therefore, there is an urgent need to apply complex network theory to runway landing event analysis to conduct systemic risk analysis and identify key risk factors for runway landing events. Summary of the Invention

[0005] The purpose of this invention is to propose a runway landing event risk analysis method based on association rules and complex networks. Based on real domestic and international runway landing accident reports, it considers the mutual influence between multiple risk factors, combines association rule mining and complex networks to identify key risk factors, thereby enabling effective prevention.

[0006] To achieve the above objectives, this invention provides a method for runway landing event risk analysis based on association rules and complex networks, comprising the following steps:

[0007] Risk factor keywords were extracted from the runway landing accident investigation report using Jieba word segmentation technology.

[0008] A dataset of runway landing event risk factors is constructed based on the aforementioned risk factor keywords;

[0009] The dataset of runway landing event risk factors was processed using an association rule mining algorithm to obtain frequent itemsets and association rules.

[0010] A complex network is constructed based on the frequent itemsets and association rules; the nodes of the complex network are frequent itemsets, the edges of the complex network are the association rules, and the weights of the edges of the risk network are the confidence levels of the association rules.

[0011] Key risk nodes are identified based on node degree and node centrality in complex networks and marked as key risk factors for runway landing events.

[0012] Based on the key risk factors mentioned above, preventive and control measures should be implemented.

[0013] Optionally, the extraction of keywords from the runway landing accident investigation report using Jieba word segmentation technology specifically includes:

[0014] Using Jieba word segmentation technology and based on the Dictionary of Aeronautical Terms as a corpus, key paragraphs of the runway landing accident investigation report were extracted.

[0015] By comparing accident texts, removing duplicate words and meaningless function words, keywords for risk factors in runway landing accident investigation reports were selected.

[0016] Optionally, the important paragraphs may include the time of occurrence, weather, aircraft type, accident type, and cause of the accident.

[0017] Optionally, a runway landing event risk factor dataset is constructed based on the aforementioned risk factor keywords, specifically including:

[0018] The risk factor keywords are categorized according to their logical order.

[0019] The risk factor keywords are divided into start nodes, intermediate nodes, and result nodes to obtain a runway landing event risk factor dataset.

[0020] Optionally, the step of using an association rule mining algorithm to perform data mining processing on the runway landing event risk factor dataset to obtain frequent itemsets and association rules specifically includes:

[0021] Scan the runway landing event risk factor dataset, count the items in all occurrence sets, and calculate the support of individual sets.

[0022] Generate a frequent 1-itemset L1 and determine whether its support meets the minimum support preset by the algorithm. If it does, obtain a frequent 2-itemset L2 from L1 according to the association rule mining algorithm, and then use L2 to find L3, and so on, until the algorithm ends.

[0023] The association rules between datasets that satisfy threshold conditions are determined based on the generated frequent itemsets.

[0024] Optionally, the node degree represents the number of edges connecting the current node directly to other nodes; the node centrality represents the relative importance of the current node to other nodes in the complex network.

[0025] Optionally, the node degree is calculated as follows:

[0026] k i =∑a ij

[0027] Where, k i Let a represent the degree of the i-th node, and j represent the nodes in the complex network other than i. When j is directly connected to i, a ij Set the value to 1, otherwise set it to 0.

[0028] Optionally, the node centrality includes: degree centrality DC. i Betweenness centrality BC i Proximity centrality CC i The calculation is as follows:

[0029]

[0030]

[0031]

[0032] Among them, DC i Describes the degree centrality of node i; k i The degree of node i is represented by N, where N is the number of nodes; BC i The s and t denote the betweenness centrality of node i; s and t denote the nodes in the network other than i. Let n represent the number of paths that pass through node i and are the shortest paths. st Represents the number of all shortest paths; CC i Indicates the proximity centrality of node i; d ij This represents the average distance from node i to node j.

[0033] Technical effects of the invention: The invention discloses a runway landing event risk analysis method based on association rules and complex networks. Based on real domestic and foreign runway landing accident reports, it considers the mutual influence between multiple risk factors, combines association rule mining and complex networks to construct a runway landing event risk network, and performs quantitative analysis through feature indicators to obtain key risk nodes, thereby preventing runway landing events in a targeted manner. Attached Figure Description

[0034] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0035] Figure 1 This is a flowchart illustrating the runway landing event risk analysis method based on association rules and complex networks, according to an embodiment of the present invention.

[0036] Figure 2 This is a flowchart illustrating the process of obtaining frequent itemsets and association rules using the association rule mining algorithm in an embodiment of the present invention.

[0037] Figure 3 This is a scatter plot showing the support and confidence of association rules in an embodiment of the present invention.

[0038] Figure 4 This is a schematic diagram of a complex network divided according to community structure, as described in an embodiment of the present invention. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0041] like Figure 1-4 As shown, this embodiment provides a runway landing event risk analysis method based on association rules and complex networks, including the following steps:

[0042] Risk factor keywords were extracted from the runway landing accident investigation report using Jieba word segmentation technology.

[0043] A dataset of runway landing event risk factors is constructed based on the aforementioned risk factor keywords;

[0044] The dataset of runway landing event risk factors was processed using an association rule mining algorithm to obtain frequent itemsets and association rules.

[0045] A complex network is constructed based on the frequent itemsets and association rules; the nodes of the complex network are frequent itemsets, the edges of the complex network are the association rules, and the weights of the edges of the risk network are the confidence levels of the association rules.

[0046] Key risk nodes are identified based on node degree and node centrality in complex networks and marked as key risk factors for runway landing events.

[0047] Based on the key risk factors mentioned above, preventive and control measures should be implemented.

[0048] Using Jieba word segmentation technology and based on the Dictionary of Aeronautical Terms as a corpus, key paragraphs of the runway landing accident investigation report were extracted, including fields such as time of occurrence, weather, aircraft type, accident type, and cause of accident. By comparing the accident text, duplicate words and meaningless function words were removed, and keywords of risk factors in the runway landing accident investigation report were selected.

[0049] The extracted risk factor keywords were classified according to their logical order; the risk factor keywords were divided into starting nodes, intermediate nodes, and result nodes to obtain the runway landing event risk factor dataset, as shown in Table 1.

[0050] Table 1

[0051]

[0052] The keyword data of risk factors contained in each accident case are compiled and summarized separately.

[0053] Association rules discover unknown relationships within a dataset by associating data points. They reflect these relationships using three factors: support, confidence, and lift. The calculation formula is as follows:

[0054] 1. Support: The frequency of itemset {A, B} in the total itemset. The higher the support, the higher the frequency of A and B appearing together in the association rule.

[0055] Support(A→B) = P(A∪B)

[0056] 2. Confidence: The conditional probability of B occurring given that A has occurred. A higher confidence level indicates that B is very likely to occur after A has occurred, and the association rule is more reliable.

[0057]

[0058] 3. Lift: The ratio of the probability of having A under the condition of having B to the probability of having A under the condition of not having B, that is, the degree of influence of the occurrence of B on the occurrence of A.

[0059]

[0060] The specific steps of the association rule mining algorithm are as follows:

[0061] 1. Scan the entire dataset, count the items in all occurrences of the set of items, and calculate the support of each individual set of items;

[0062] 2. Generate a frequent 1-itemset L1 and determine if its support meets the minimum support preset by the algorithm. If it does, use the association rule mining algorithm to obtain a frequent 2-itemset L2 from L1. Then use L2 to find L3, and so on, until the algorithm ends;

[0063] 3. Based on the generated frequent itemsets, association rules that satisfy threshold conditions between datasets can be obtained.

[0064] Specifically:

[0065] Data import: The risk factor keyword data table, which is compiled and summarized by accident case, is stored as input data in a file (.csv) and imported into R for association rule mining.

[0066] Setting minimum thresholds: In order to obtain association rules with higher credibility, the thresholds of the two dynamic variables, minimum support and minimum confidence, are adjusted. In this example, the minimum support is 25% and the minimum confidence is 20%.

[0067] Generate frequent itemsets; filter out frequent itemsets that meet the above thresholds using an association rule mining algorithm. Generate association rules; based on the obtained frequent itemsets, use an association rule mining algorithm to filter and obtain association rules, and calculate the corresponding support and confidence of the association rules; the specific process of obtaining association rules using the algorithm is as follows: Figure 3 .

[0068] like Figure 4 As shown, the frequent itemsets that meet the minimum threshold are used as nodes of the network, each association rule is used as an edge of the network, and the confidence of each association rule is used as the weight of the edge to construct the runway landing event risk network. The runway landing event risk network is divided according to the community structure.

[0069] Calculate the node degree and node centrality in complex networks, rank these indicators, and identify the nodes with the highest rankings as key risk nodes.

[0070] Node degree represents the number of nodes directly connected to that node in the network, and is calculated using the following formula:

[0071] k i =∑a ij

[0072] Where, k i Let a represent the degree of the i-th node, and j represent the nodes in the complex network other than i. When j is directly connected to i, a ij Set the value to 1, otherwise set it to 0.

[0073] Node centrality includes degree centrality (DC) i Betweenness centrality BC i Proximity centrality CC i This reflects the relative importance of nodes in the network, and the calculation formula is as follows:

[0074]

[0075]

[0076]

[0077] Among them, DC i Describes the degree centrality of node i; k i The degree of node i is represented by N, where N is the number of nodes; BC i The s and t denote the betweenness centrality of node i; s and t denote the nodes in the network other than i. Let n represent the number of paths that pass through node i and are the shortest paths. st Represents the number of all shortest paths; CC i Indicates the proximity centrality of node i; d ij This represents the average distance from node i to node j.

[0078] Using the complex network analysis software Gephi, the evolution network of runway landing event risks was analyzed, and the community partitioning results were obtained as follows: Figure 4 Community structure refers to the grouping of nodes in a network, with relatively dense connections within groups and relatively sparse connections between groups. Community structure reflects the properties of networks at the mesoscale, and studying community structure is an important way to understand the overall network structure and function. Figure 4 It can be seen that the network has a relatively obvious community structure, with nodes within the community being more closely connected and more likely to trigger each other, thus leading to the occurrence of events.

[0079] The degree and node centrality of each node in the network are calculated. Table 2 shows the top 10 nodes in degree distribution; Table 3 shows the top 10 nodes in betweenness centrality; and Table 4 shows the top 10 nodes in proximity centrality. Based on the calculated degree centrality, betweenness centrality, and proximity centrality, their combined value is defined as the average of the three centrality values. The combined value of all nodes is then obtained, which is the importance of risky nodes. Table 5 shows the node combined value, and Table 6 shows the relevant parameters of the community.

[0080] Table 2

[0081]

[0082] Table 3

[0083]

[0084] Table 4

[0085]

[0086] Table 5

[0087]

[0088] Table 6

[0089]

[0090] Based on the key risk factors identified in the examples, corresponding preventive measures are proposed.

[0091] This invention is the first to combine association rule mining and complex networks to analyze key risk factors in runway landing events; it takes into account the mutual influence of multiple risk factors and performs quantitative calculation and analysis on the criticality of risk nodes.

[0092] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A runway landing event risk analysis method based on association rules and complex networks, characterized in that, Includes the following steps: Risk factor keywords were extracted from the runway landing accident investigation report using Jieba word segmentation technology. A dataset of runway landing event risk factors is constructed based on the aforementioned risk factor keywords; The dataset of runway landing event risk factors was processed using an association rule mining algorithm to obtain frequent itemsets and association rules. A complex network is constructed based on the frequent itemsets and association rules; the nodes of the complex network are frequent itemsets, the edges of the complex network are the association rules, and the weights of the edges of the complex network are the confidence scores of the association rules. Key risk nodes are identified based on node degree and node centrality in complex networks and marked as key risk factors for runway landing events. Based on the aforementioned key risk factors, prevention and control measures should be implemented. The extraction of keywords from the runway landing accident investigation report using Jieba word segmentation technology specifically includes: Using Jieba word segmentation technology and based on the Dictionary of Aeronautical Terms as a corpus, key paragraphs of the runway landing accident investigation report were extracted. By comparing accident texts, removing duplicate words and meaningless function words, keywords for risk factors in runway landing accident investigation reports were selected. The key information includes the time of occurrence, weather conditions, aircraft type, accident type, and the determined cause of the accident. A runway landing event risk factor dataset was constructed based on the aforementioned risk factor keywords, specifically including: The risk factor keywords are categorized according to their logical order. The risk factor keywords are divided into start nodes, intermediate nodes, and result nodes to obtain a runway landing event risk factor dataset.

2. The runway landing event risk analysis method based on association rules and complex networks according to claim 1, wherein, The process of using association rule mining algorithms to mine the runway landing event risk factor dataset to obtain frequent itemsets and association rules specifically includes: Scan the runway landing event risk factor dataset, count the items in all occurrence sets, and calculate the support of individual sets. Generate a frequent 1-itemset L1 and determine whether its support meets the minimum support preset by the algorithm. If it does, obtain a frequent 2-itemset L2 from L1 according to the association rule mining algorithm, and then use L2 to find L3, and so on, until the algorithm ends. The association rules between datasets that satisfy threshold conditions are determined based on the generated frequent itemsets.

3. The runway landing event risk analysis method based on association rules and complex networks as described in claim 1, characterized in that, The node degree represents the number of edges connecting the current node directly to other nodes; the node centrality represents the relative importance of the current node to other nodes in the complex network.

4. The runway landing event risk analysis method based on association rules and complex networks as described in claim 3, characterized in that, The node degree is calculated as follows: in, Indicates the first The degree of a node. In addition to complex networks Other nodes, when and When directly connected, Set the value to 1, otherwise set it to 0.

5. The runway landing event risk analysis method based on association rules and complex networks as described in claim 3, characterized in that, The node centrality includes: degree centrality. Betweenness centrality Proximity centrality The calculation is as follows: in, Represents a node Degree centrality; Represents nodes The degree, It is the number of nodes; Represents a node Betweenness centrality; s and t represent nodes in the network other than i. Indicates passing through nodes And this is the number of paths that represent the shortest path. This represents the number of all shortest paths; Represents a node Proximity centrality; Represents a node The average distance to node j.

Images