A method and apparatus for constructing knowledge graphs for collaborative operation of airport clusters

Abstract

This invention discloses a method and apparatus for constructing a knowledge graph for collaborative operation of airport clusters. The method includes: multi-source heterogeneous data acquisition and preprocessing; ontology construction, providing a semantic framework; knowledge extraction; knowledge storage and graph construction; and dynamic updating and evolution. The advantages of this invention are as follows: it achieves semantic alignment and fusion of cross-system data, enabling computers to understand and process the business meaning behind the data; it overcomes the technical shortcomings of traditional database foreign key association methods, which require writing complex multi-table join statements and have low computational efficiency for cross-subject queries; and it implements a dynamic updating and evolution mechanism based on streaming data, ensuring the timeliness and evolution capability of knowledge.

Description

Technical Field

[0001] This invention relates to the field of intelligent information processing technology for air traffic, specifically to a method and apparatus for constructing a knowledge graph for collaborative operation of airport clusters. Background Technology

[0002] In the collaborative operation and management of airport clusters, massive amounts of information, such as flight operation messages, flight plans, air traffic control instructions, and meteorological observation data, are scattered and stored in independent systems of multiple entities, including air traffic control, airports, and airlines. Faced with this multi-source, heterogeneous data environment, existing technologies typically employ relational databases for storage and management, or abstract operational data into tabular statistical values ​​for presentation. The core principle is to organize data from different sources according to a predefined table structure, establish relationships between tables through primary and foreign keys, and utilize structured query languages ​​for data retrieval and aggregation operations. Based on usage requirements, the raw operational data is aggregated and calculated according to time or spatial dimensions to form statistical indicators, which are then displayed and analyzed in tabular form.

[0003] Existing technology 1 provides a database storage management method, including: establishing the association between data at various granularities and storage resources, periodically collecting access behavior data, generating a cold and hot data distribution list, a cross-resource access topology map, and storage performance analysis results, and executing optimization operations according to an optimization mode. The core of this method lies in optimizing storage resource allocation based on behavioral data such as access frequency. Its data organization method is still mainly based on tables and indicators, and it does not involve semantic relationship modeling between business entities. Furthermore, this method can only reflect the distribution of data popularity and cannot characterize complex business logic such as flight delay propagation and the impact of flow control measures.

[0004] Existing technology two provides a heterogeneous data storage method for a civil aviation data platform. This method includes real-time collection of metadata from various subsystems, inputting it into a unified metadata model for standardized processing to generate a global metadata view, establishing a data traceability mechanism to generate metadata change trajectories, and constructing a metadata knowledge graph using metadata quality tags. This method focuses on storage management and quality assessment at the metadata level. Its constructed knowledge graph uses metadata (such as table structure, storage location, and access logs) as nodes, but does not include business entities such as flights, airports, and airlines, or their semantic relationships. Therefore, it cannot support business semantic queries and causal reasoning in the collaborative operation and management of airport clusters.

[0005] Existing technology three provides a database retrieval method based on tabular semantic annotation, which includes: integrating multiple data files into a report, converting user-input statistical criteria into SQL statements through a pre-set dictionary and dynamic concatenation technology, and generating indicator information based on target field information. The core principle of this method is to extract field information from data files through a metadata access interface and establish relationships between fields after semantic annotation. However, the relationships established by this method are limited to simple field matching and cannot depict the complex causal relationships and transmission chains between entities such as flights, flow control, weather, airports, and airlines. Furthermore, the knowledge is stored in the form of tables and indicators, lacking the ability to perform structured modeling and reasoning of domain knowledge.

[0006] Existing technology four provides a method for constructing a knowledge graph for the public transportation domain based on multi-source heterogeneous data. This method involves acquiring structured, semi-structured, and unstructured data through a multi-source data access interface, preprocessing the data, inputting it into a multimodal feature alignment module to generate feature vectors in a unified semantic space, then using a domain-enhanced named entity recognition module and a context-aware relation extraction module to generate triples, and finally constructing an incremental knowledge graph evolution engine. While this method introduces knowledge graph technology, its application scenario is limited to the public transportation domain. It lacks an ontology model adapted to the unique data formats and business semantics of airport cluster operations, such as flight messages, flow control instructions, and meteorological observation data. Furthermore, this method does not establish a dynamic update mechanism based on streaming data, making it difficult to guarantee the timeliness of the knowledge graph.

[0007] In summary, existing technologies for knowledge representation in the context of airport cluster collaborative operations share the following common problems:

[0008] First, there is a lack of semantic meaning. Data standards vary across systems, and the same business concepts are stored with different field names and formats, making it impossible for computers to understand the business meaning of the data.

[0009] Second, the relationships are implicit. The complex causal relationships and transmission chains between entities such as flights, flow control, weather, airports, and airlines cannot be explicitly expressed in the table structure or statistical indicator tables of relational databases, which makes cross-entity association analysis difficult and multi-hop reasoning impossible.

[0010] Third, knowledge is solidified. Experienced knowledge exists in a static form and lacks a dynamic update and iteration mechanism based on operational data, making it difficult for knowledge to continuously evolve with changes in the operating environment.

[0011] The aforementioned problems have led to a high degree of reliance on manual intervention in airport cluster operation and management at three levels: data understanding, relationship analysis, and knowledge evolution. Overall, this results in issues such as high dependence on manual data understanding, difficulty in locating the root causes of operational anomalies, and slow adaptation to complex management and control scenarios, making it difficult to maximize the operational efficiency and quality improvement of airport cluster collaboration.

[0012] Therefore, there is an urgent need for a method that can achieve unified representation of multi-source heterogeneous data, explicit modeling of business relationships, and dynamic evolution of domain knowledge, so as to provide a knowledge foundation that can be directly calculated and traced by reasoning for the collaborative operation and management of airport clusters. Summary of the Invention

[0013] To address the shortcomings of the existing technologies, the present invention aims to provide a method and apparatus for constructing knowledge graphs for collaborative operation of airport clusters, thereby solving the following technical problems:

[0014] 1. Overcome the shortcomings of traditional database storage methods, such as the disconnect between data and business meaning, and the fact that computers can only perform numerical calculations but cannot understand the business logic behind the data;

[0015] 2. Overcome the shortcomings of traditional database foreign key association methods, such as the need to write complex multi-table join statements for cross-subject queries, low computational efficiency, and difficulty in implementing multi-hop inference;

[0016] 3. Overcome the shortcomings of experiential knowledge that relies on manual transmission, has high reuse costs, and is difficult to dynamically evolve with the operating environment.

[0017] To achieve the above objectives, in a first aspect, embodiments of the present invention provide a method for constructing a knowledge graph for collaborative operation of airport clusters, comprising:

[0018] Collect multi-source heterogeneous data generated during the collaborative operation of the airport cluster, and preprocess the heterogeneous data to obtain preprocessed text data;

[0019] Construct an ontology model for the collaborative operation of airport clusters to define the semantic framework of the knowledge graph;

[0020] Based on the defined knowledge graph semantic framework, entities, relations and attributes are automatically extracted from preprocessed text data to generate entity-relation-entity triple data;

[0021] Based on the triplet data, knowledge storage and graph construction are completed to obtain a knowledge graph for the collaborative operation of the airport cluster.

[0022] Furthermore, as a preferred implementation of this application, after obtaining the airport cluster collaborative operation knowledge graph, the method further includes establishing a dynamic update and evolution mechanism based on streaming data, specifically:

[0023] The system utilizes the Apache Flink streaming engine to access real-time data streams, which include flight operation status updates, flow control command change data, and weather warning information.

[0024] When the real-time streaming engine Apache Flink detects the arrival of new data, it triggers an incremental update operation.

[0025] Generate a graph snapshot of the current running state at preset time intervals or based on event triggering; the graph snapshot represents the complete state of the knowledge graph at any given time.

[0026] Secondly, embodiments of the present invention also provide a knowledge graph construction apparatus for airport cluster collaborative operation, including a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method described in the first aspect.

[0027] Thirdly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the method described in the first aspect.

[0028] The advantages of implementing the embodiments of the present invention are as follows:

[0029] (1) Achieve unified semantic representation of multi-source heterogeneous data

[0030] To address the problem of semantic gaps in multi-source heterogeneous data and the inability of computers to understand the business meaning of data in existing technologies, this invention constructs an ontology model adapted to the airport cluster operation domain, defines unified entity categories, relationship types, and attribute structures, and associates heterogeneous data scattered across multiple independent systems such as air traffic control, airports, and airlines into a unified semantic network, realizing semantic alignment and fusion of cross-system data, enabling computers to understand and process the business meaning behind the data.

[0031] (2) Achieve explicit modeling and efficient reasoning of business relationships

[0032] To address the challenges of existing relational databases in explicitly expressing business relationships, performing difficult cross-entity association analysis, and implementing multi-hop inference, this invention employs an "entity-relationship-entity" triple structure to explicitly model business relationships. It stores the complex causal relationships and transmission chains between entities such as flights, flow control, weather, airports, and airlines in a graph structure. Based on this graph database storage method, it supports efficient graph traversal and multi-hop queries, overcoming the technical shortcomings of traditional database foreign key association methods, which require complex multi-table join statements and suffer from low computational efficiency for cross-entity queries.

[0033] (3) Realize the dynamic evolution and iterative optimization of domain knowledge

[0034] To address the problems of existing technologies where experiential knowledge exists in a static form, lacks a dynamic update and iteration mechanism based on operational data, and is difficult to continuously evolve with changes in the operating environment, this invention establishes a dynamic update and evolution mechanism based on streaming data. By using a streaming computing engine to access operational data in real time, the knowledge graph is incrementally updated, and graph snapshots are generated at preset time intervals or triggered by events. This allows the knowledge graph to continuously evolve in response to changes in the operational status of random field clusters, ensuring the timeliness and evolutionary capability of the knowledge. Attached Figure Description

[0035] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below.

[0036] Figure 1 This is a flowchart of a knowledge graph construction method for airport cluster collaborative operation provided in an embodiment of the present invention;

[0037] Figure 2 yes Figure 1 Another flowchart of the method shown;

[0038] Figure 3 This is a schematic diagram of the airport cluster collaborative operation ontology structure constructed by the present invention;

[0039] Figure 4 This is the structure diagram of the BERT-BiGRU-Attention-CRF knowledge extraction model of this invention;

[0040] Figure 5 This is a schematic diagram of the dynamic update and evolution mechanism of the present invention;

[0041] Figure 6 This is a structural diagram of the knowledge graph construction device for airport cluster collaborative operation provided in an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0044] The inventive concept of this invention is to provide a knowledge graph construction method for airport cluster collaborative operation, aiming to solve the following technical problems:

[0045] (1) To address the semantic loss of multi-source heterogeneous data during storage and management, and the inability to distinguish the differences in data meaning under different business scenarios, a unified semantic framework is constructed to give clear business semantics to various types of operational data. This overcomes the defects of traditional database storage methods where data and business meaning are disconnected and computers can only perform numerical calculations but cannot understand the business logic behind the data, thus achieving semantic alignment and unified representation of multi-source heterogeneous data.

[0046] (2) To address the implicit relational problems of relational databases, which make it difficult to explicitly express business relationships and have low efficiency in multi-hop reasoning, we design a triple structure of "entity-relationship-entity" to explicitly model business relationships. This connects data stored in multiple systems into a unified semantic network, overcoming the shortcomings of traditional database foreign key association methods, which require writing complex multi-table join statements for cross-subject queries, have low computational efficiency, and are difficult to achieve multi-hop reasoning. This enables efficient querying and reasoning of business relationships.

[0047] (3) To address the problem of knowledge solidification that exists in a static form, lacks a dynamic update and iteration mechanism based on operational data, and is difficult to continuously evolve with changes in the operational environment, domain knowledge is modeled as relational patterns, rule templates, and statistical laws in a graph, and a dynamic update mechanism based on operational data is established. This overcomes the shortcomings of experiential knowledge that rely on manual transmission, has high reuse costs, and is difficult to dynamically evolve with the operational environment, thereby realizing the structured accumulation and dynamic iterative evolution of domain knowledge.

[0048] Please refer to Figure 1 and Figure 2 This invention provides a method for constructing a knowledge graph for collaborative operation of airport clusters, which may include the following steps:

[0049] Step S1: Collect and preprocess multi-source heterogeneous data to obtain text data.

[0050] S1-1: Data Acquisition. Collect multi-source heterogeneous data generated during the collaborative operation of the airport cluster, including but not limited to: flight operation messages, flight plan reports, air traffic control instructions, meteorological observation data, flight schedules, runway status information, etc.

[0051] S1-2: Data Preprocessing. Preprocess the collected multi-source heterogeneous data. Remove duplicate, outlier, and invalid data, and handle missing values ​​to ensure data integrity and consistency. For Chinese text data, use the Jieba word segmentation tool for word segmentation. For specialized terminology in the airport cluster operation field, construct a civil aviation-specific feature lexicon and add the specialized terms to the word segmentation dictionary to avoid incorrect segmentation. Normalize data from different sources and in different formats, unifying time formats, unit systems, field naming conventions, etc., to provide standardized data input for knowledge extraction in the subsequent step S3.

[0052] Furthermore, the data is standardized using the range normalization method:

[0053] For positive indicators, see equation (1):

[0054]

[0055] For negative indicators, see equation (2):

[0056]

[0057] in, N The total number of samples, x ij For the first j The first indicator i The original values ​​of each sample The value is the normalized value. The first j The maximum and minimum values ​​of each indicator across all samples.

[0058] The above preprocessing operations provide standardized, uniform-dimensional data input for subsequent knowledge extraction.

[0059] Step S2: Construct an ontology model for the collaborative operation of airport clusters to define the semantic framework of the knowledge graph.

[0060] After completing the collection and preprocessing of multi-source heterogeneous data, it is necessary to construct an ontology model for the collaborative operation of airport clusters in order to define the semantic framework of the knowledge graph.

[0061] S2-1: Definition of core entity categories.

[0062] The entity categories defined in this embodiment of the invention include flight entities, airport entities, airline entities, air traffic control unit entities, meteorological event entities, and flow control measure entities. Specifically, flight entities include attributes such as flight number, scheduled departure time, actual departure time, scheduled landing time, actual landing time, and flight status; airport entities include attributes such as airport name, geographical location, number of runways, and airport capacity; airline entities include attributes such as airline name, code, and base airport; air traffic control unit entities include attributes such as unit name and jurisdictional airspace; meteorological event entities include attributes such as meteorological type, impact range, start time, and end time; and flow control measure entities include attributes such as measure type, issuance time, implementation time, and impact range.

[0063] S2-2: Relation type definition.

[0064] Based on the defined entity categories, the semantic relationship types between entities are determined, including locateTo, belongTo, controlBy, affect, and cause relationships. Specifically, locateTo represents the takeoff and landing positioning relationship between a flight and an airport; belongTo represents the affiliation relationship between a flight and an airline; controlBy represents the control relationship between a flight and air traffic control; affect represents the influence relationship between entities; and cause represents the causal relationship.

[0065] S2-3: Attribute structure definition.

[0066] Define a corresponding set of attributes and attribute types for each entity category, including numeric, character, and time types, to ensure the structured storage of entity attributes and the efficiency of subsequent queries.

[0067] Please refer to Figure 3 This is a schematic diagram of the airport cluster collaborative operation ontology structure constructed in this invention. The ontology defines core entity categories such as flight entities, airport entities, airline entities, air traffic control unit entities, meteorological event entities, and flow control measure entities. Relationship types include: belongTo (belong to), locateTo (locate to), controlBy (control), affect (affect), and cause (cause). Each entity category contains a corresponding set of attributes.

[0068] Step S3: Knowledge extraction, generating entity-relationship-entity triplet data.

[0069] After the ontology is constructed, entities, relations, and attributes are automatically extracted from the preprocessed text data to generate "entity-relation-entity" triples. Knowledge extraction is divided into two subtasks: entity recognition and relation extraction.

[0070] S3-1: Entity recognition.

[0071] The goal of entity recognition is to identify entities such as flight numbers, airport codes, time information, names of air traffic control measures, and types of weather events from text, and to label their boundaries and types. This invention uses the BERT-BiGRU-Attention-CRF model for entity recognition, which consists of four parts: a BERT embedding layer, a BiGRU feature layer, an Attention layer, and a CRF output layer. Please refer to [reference needed]. Figure 4 This is a schematic diagram of the knowledge extraction module of the present invention, which adopts a four-layer architecture of BERT-BiGRU-Attention-CRF. The input text sequence first enters the embedding layer, where it is converted into an embedding vector sequence by the pre-trained model of the BERT layer; then it enters the feature layer, where context features are extracted by bidirectional gated recurrent units, and the hidden state sequence is output; next, it enters the attention layer, where the attention weights at each time step are calculated and weighted summation is performed to obtain the context vector; finally, it enters the CRF output layer, where global sequence labeling is performed by a conditional random field, and the entity boundaries and type labels are output.

[0072] Among them, the BERT embedding layer: uses the BERT pre-trained language model to convert the input text into context-dependent dynamic word vectors, solving the problem of polysemy in civil aviation professional terms.

[0073] BiGRU Feature Layer: Contextual features of the text are extracted using bidirectional gated recurrent units. Let the hidden state output by BiGRU at time step [value missing]. It captures historical and future information of the sequence through forward and backward GRUs, respectively, and concatenates the two as the output of the current time step.

[0074] Attention layer: An attention mechanism is introduced to enhance the weights of key semantic information. The attention weights at each time step are calculated using equations (3) and (4).

[0075]

[0076] in, For the first t Attention weights at each time step This is the weight matrix. For bias vectors, For attention vectors, d a Let be the dimension of the attention layer, d be the dimension of the BiGRU hidden state vector, and T be the length of the input sequence. For time step t Attention weights, satisfying For the first kAttention weights at each time step k This is a time measure in the sequence.

[0077] The attention representation of the sequence is obtained by summing the attention weights with the hidden states: This allows the model to focus on features that contribute significantly to entity recognition. Among these features, This is the context vector after incorporating the attention mechanism.

[0078] The CRF output layer uses a Conditional Random Field (CRF) to globally optimize the sequence labels, considering the transition relationships between labels and outputting the optimal entity boundary and type annotation results. The CRF works in a given input sequence... x Under the given conditions, output the label sequence The probability calculation is shown in equation (5):

[0079]

[0080] in, W yt For tags y t The launch score weight vector, For from the tag y t-1 Transfer to label y t The transfer score, Y The set of all possible label sequences.

[0081] The final optimal label sequence y* is obtained by maximizing the log-likelihood, as shown in equation (6):

[0082]

[0083] S3-2: Relation extraction.

[0084] The goal of relation extraction is to identify the semantic relationship types between identified entities. This invention employs a BiGRU-Attention model for relation extraction, whose structure is similar to the feature layer and attention layer in entity recognition. Specifically, the identified entity pairs are used as input, and the contextual information of the sentence containing the entities is encoded by a bidirectional GRU to capture the semantic dependencies between entities. An attention mechanism is introduced to focus on semantic features that play a key role in relation judgment. Finally, a Softmax classifier outputs the relationship type between entity pairs, such as affect, cause, belong to, etc.

[0085] S3-3: Triple generation.

[0086] The results obtained from entity recognition are combined with the results obtained from relation extraction to generate "entity-relation-entity" triples. For example, for the text "Flow control measure A caused flight B to be delayed", the triples (flow control measure A, cause, flight B delayed) can be generated. The generated triples will serve as the building blocks of the knowledge graph for subsequent knowledge storage.

[0087] Step S4: Complete knowledge storage and graph construction based on triple data to obtain the knowledge graph of airport cluster collaborative operation.

[0088] After knowledge extraction and triple generation are completed, the triples are stored in a graph database to complete the construction of the knowledge graph.

[0089] S4-1: Graph database storage.

[0090] The triples generated in step S3 are stored in the graph database. Using the ontology model constructed in step S2 as the semantic framework, entities are mapped to nodes in the graph database, relations are mapped to edges between nodes, and entity attributes are mapped to attribute key-value pairs of nodes. The graph structure storage method of the graph database can naturally express complex relationships between entities and supports efficient graph traversal and multi-hop query operations.

[0091] S4-2: Map Construction.

[0092] After storing all triples, a collaborative knowledge graph for airport cluster operations is formed, covering multiple subjects and relationship types, including flights, airports, airlines, air traffic control units, meteorological events, and flow control measures. This knowledge graph uses "entity-relationship-entity" triples as its basic unit, linking heterogeneous data scattered across multiple independent systems of air traffic control, airports, and airlines into a unified semantic network. This achieves unified semantic representation of multi-source heterogeneous data and explicit modeling of business relationships.

[0093] Step S5: Establish a dynamic update and evolution mechanism based on streaming data.

[0094] To ensure that the knowledge graph can continuously evolve with changes in the operating environment and avoid the problem of knowledge gradually becoming outdated and invalid after being built once, this invention establishes a dynamic update and evolution mechanism based on streaming data.

[0095] S5-1: Real-time access to streaming data.

[0096] Streaming computing engines (such as Apache Kafka and Apache Flink) are used to access real-time data streams such as flight operation status updates, flow control command changes, and weather warning information, ensuring that the knowledge graph can promptly perceive changes in the operational situation.

[0097] S5-2: Incremental update.

[0098] When the streaming engine detects new data arrival, it triggers an incremental update operation. Specifically, this includes: for existing entity nodes (such as a flight), updating their attribute values ​​(such as actual departure time and flight status) based on the new data; for newly emerging entities (such as newly generated flights or newly released flow control measures), creating corresponding nodes in the graph and populating their attributes; for the influence relationships between entities, dynamically calculating relationship weights and updating relationship attributes based on real-time data within a sliding time window; and creating corresponding edges in the graph for newly discovered relationships between entities.

[0099] The formula for calculating the weight of dynamic relationships between entities is as follows, assuming entities u and entity v There is a relationship r ,exist t At any given moment, relational weights are based on historical data within the sliding window. The calculation is as shown in equation (7):

[0100]

[0101] in, L The length of the sliding window. For entities u exist Embedding vector at time step, The cosine similarity function is, i.e. For entities u The set of neighboring entities, The sigmoid activation function maps weight values ​​to the (0, 1) interval. Using the above formula, the relation weights can dynamically evolve with real-time data, accurately reflecting the changing strength of the influence relationship between entities.

[0102] S5-3: Graph snapshot generation.

[0103] Generate a graph snapshot of the current running state at preset time intervals or based on event triggers. The graph snapshot is the complete state of the knowledge graph at a certain moment and can be used as input for downstream applications.

[0104] The time interval for generating image snapshots can be dynamically adjusted according to system load, and the calculation formula is as follows:

[0105]

[0106] in, Δt base Based on the time interval, Δt min For the minimum time interval, N eventλ represents the number of events per unit time, and λ is the smoothing coefficient. When the event frequency is high, the snapshot generation interval is automatically shortened to improve the real-time performance of situational awareness; when the event frequency is low, the generation interval is automatically extended to reduce system load.

[0107] Please refer to Figure 5 This diagram illustrates the dynamic update and evolution mechanism of the present invention. First, a streaming computing engine accesses real-time data streams from multiple sources, including flight operation status updates, flow control command changes, and weather warnings. Then, incremental update operations are triggered, including node attribute updates, node additions, dynamic updates of relationship weights, and relationship additions. Next, at preset time intervals or based on event triggers, graph snapshots of the current operating state are generated. Each snapshot contains a set of nodes, a set of edges, and an attribute snapshot. Finally, the generated graph snapshots can be used by downstream applications, achieving dynamic evolution and timeliness assurance of the knowledge graph.

[0108] Through the aforementioned dynamic update and evolution mechanism, the knowledge graph constructed by this invention can be continuously updated in response to the real-time changes in random field group operation data, ensuring the timeliness and evolution capability of knowledge, and overcoming the technical defects of traditional knowledge graphs that are constructed in one go and cannot adapt to dynamic changes in the operating environment.

[0109] The advantages of implementing the method provided in the embodiments of the present invention are as follows:

[0110] (1) Achieve unified semantic representation of multi-source heterogeneous data

[0111] To address the problem of semantic gaps in multi-source heterogeneous data and the inability of computers to understand the business meaning of data in existing technologies, this invention constructs an ontology model adapted to the airport cluster operation domain, defines unified entity categories, relationship types, and attribute structures, and associates heterogeneous data scattered across multiple independent systems such as air traffic control, airports, and airlines into a unified semantic network, realizing semantic alignment and fusion of cross-system data, enabling computers to understand and process the business meaning behind the data.

[0112] (2) Achieve explicit modeling and efficient reasoning of business relationships

[0113] To address the challenges of existing relational databases in explicitly expressing business relationships, performing difficult cross-entity association analysis, and implementing multi-hop inference, this invention employs an "entity-relationship-entity" triple structure to explicitly model business relationships. It stores the complex causal relationships and transmission chains between entities such as flights, flow control, weather, airports, and airlines in a graph structure. Based on this graph database storage method, it supports efficient graph traversal and multi-hop queries, overcoming the technical shortcomings of traditional database foreign key association methods, which require complex multi-table join statements and suffer from low computational efficiency for cross-entity queries.

[0114] (3) Realize the dynamic evolution and iterative optimization of domain knowledge

[0115] To address the problems of existing technologies where experiential knowledge exists in a static form, lacks a dynamic update and iteration mechanism based on operational data, and is difficult to continuously evolve with changes in the operating environment, this invention establishes a dynamic update and evolution mechanism based on streaming data. By using a streaming computing engine to access operational data in real time, the knowledge graph is incrementally updated, and graph snapshots are generated at preset time intervals or triggered by events. This allows the knowledge graph to continuously evolve in response to changes in the operational status of random field clusters, ensuring the timeliness and evolutionary capability of the knowledge.

[0116] To better understand the knowledge graph construction method provided by this invention, specific examples are described below.

[0117] This embodiment selects the Pearl River Delta airport cluster in my country as the research object, including the three core airports—Guangzhou Baiyun International Airport, Shenzhen Bao'an International Airport, and Zhuhai Jinwan International Airport—and their surrounding small and medium-sized airports. Operational data from January 1, 2023 to December 31, 2024 (24 months) was collected as the training dataset, and operational data from January 1, 2025 to January 31, 2025 (31 days) was collected as the test dataset. Data sources include flight regularity statistics systems, the national air traffic management system, and meteorological observation systems. The implementation process of this invention will be described in detail below with specific data.

[0118] (I) Data Acquisition and Preprocessing

[0119] Following step S1, approximately 480,000 flight operation reports, 520,000 flight plan reports, 12,000 air traffic control instructions, and 35,000 meteorological observation data were collected from the flight regularity statistics system, the national air traffic management system, and the meteorological observation system from January 2023 to January 2025. The collected data underwent preprocessing: approximately 300 duplicate reports and 50 abnormal data were removed; Jieba word segmentation was used to construct a civil aviation characteristic vocabulary library containing over 120 professional terms such as "overflow," "time slot," "time adjustment," "CTOT," and "TOBT"; all time fields were uniformly converted to the "YYYY-MM-DD HH:MM:SS" format, speed units were standardized to "km / h," and distance units were standardized to "km."

[0120] (II) Ontology Construction

[0121] Following step S2, construct an ontology model for the collaborative operation of the Pearl River Delta airport cluster. Based on the operational needs of the Pearl River Delta airport cluster, define entity categories including: flight entities (including attributes such as flight number, planned departure time, actual departure time, planned landing time, actual landing time, and flight status), airport entities (including attributes such as airport name, geographical location, number of runways, and airport capacity), airline entities (including attributes such as airline name, code, and base airport), air traffic control unit entities (including attributes such as unit name and jurisdictional airspace), meteorological event entities (including attributes such as meteorological type, impact range, start time, and end time), and flow control measure entities (including attributes such as measure type, release time, execution time, and impact range). Define relationship types including: locateTo (takeoff and landing positioning relationship), belongTo (attribution relationship), controlBy (control relationship), affect (impact relationship), and cause (causal relationship).

[0122] (III) Knowledge Extraction

[0123] According to step S3, entities, relationships and attributes are automatically extracted from the preprocessed text data.

[0124] First, an entity recognition model using BERT-BiGRU-Attention-CRF is employed. Preprocessed text is input into the BERT embedding layer to generate dynamic word vectors (768 dimensions); contextual features are extracted through the BiGRU feature layer (256 hidden units); after weighted fusion through the attention layer, the output is fed into the CRF layer for global sequence labeling. Testing shows that the entity recognition accuracy reaches 92.7%, and the recall rate reaches 89.4%.

[0125] Secondly, a BiGRU-Attention model is used for relation extraction. The identified entity pairs are used as input, and contextual information is encoded through a bidirectional GRU. An attention mechanism is introduced to focus on key features, and a Softmax classifier outputs the relation type. Testing showed that the relation extraction accuracy reached 88.3%.

[0126] Finally, "entity-relationship-entity" triples are generated. For example, from the flow control instruction text "Due to thunderstorms, flow control is now implemented in the Guangzhou terminal area, affecting flight CZ3506," the triples (thunderstorms, cause, flow control measures) and (flow control measures, affect, CZ3506) are generated. Approximately 1.258 million triples are generated in total.

[0127] (iv) Knowledge storage and graph construction

[0128] Following step S4, the generated triples are stored in the Neo4j graph database. Using the ontology model as the semantic framework, entities are mapped to nodes (approximately 42,000 nodes in total), relations are mapped to edges (approximately 126,000 edges in total), and entity attributes are mapped to attribute key-value pairs of nodes. This forms a knowledge graph for the collaborative operation of the Pearl River Delta airport cluster, covering multiple subjects and multiple relationship types, including flights, airports, airlines, air traffic control units, meteorological events, and flow control measures.

[0129] (V) Dynamic updates and evolution

[0130] Following step S5, the Apache Kafka stream processing engine is used to access the operational data stream from January 2025 in real time. When new data is detected, an incremental update operation is triggered: for existing entity nodes (such as flight CZ3506), their attribute values ​​(such as actual departure time and flight status) are updated based on the new data; for newly emerging entities (such as new flights and new flow control measures), corresponding nodes are created in the graph and their attributes are populated; for the influence relationships between entities, the relationship weights are calculated and updated according to the dynamic relationship weight formula based on real-time data within a sliding time window (window length L = 30 minutes). Graph snapshots of the current operational status are generated every 15 minutes, for a total of 2976 graph snapshots (31 days × 24 hours × 4 snapshots / hour). The latency of graph snapshot generation is controlled within 200 milliseconds, and the average processing latency of incremental updates is less than 5 seconds.

[0131] (vi) Application effect

[0132] Through the above steps, this embodiment successfully constructed a knowledge graph for the collaborative operation of the Pearl River Delta airport cluster. This knowledge graph supports multi-hop queries across entities, such as: "Query all flow control measures affected by thunderstorms and a list of affected flights." Testing showed that the response time for a single multi-hop query is less than 100 milliseconds. Furthermore, through a dynamic update mechanism, the knowledge graph can promptly reflect real-time operational status changes such as flight status updates and adjustments to flow control measures, demonstrating significantly better timeliness than static knowledge graphs.

[0133] In summary, the knowledge graph construction method for airport cluster collaborative operation provided by this invention has been verified on real operational data of the Pearl River Delta airport cluster. It can realize unified semantic representation of multi-source heterogeneous data, explicit modeling of business relationships, and dynamic evolution of domain knowledge, thus verifying the effectiveness and practicality of this invention.

[0134] Based on the same inventive concept, such as Figure 6As shown, this embodiment of the invention also provides a knowledge graph construction apparatus for airport cluster collaborative operation, which may include: one or more processors 101, one or more input devices 102, one or more output devices 103, and a memory 104. The processors 101, input devices 102, output devices 103, and memory 104 are interconnected via a bus 105. The memory 104 is used to store a computer program, which includes program instructions. The processors 101 are configured to invoke the program instructions to execute the method described in the above-described method embodiment.

[0135] It should be understood that, in this embodiment of the invention, the processor 101 may be a central processing unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0136] Input device 102 may include a keyboard, etc., and output device 103 may include a display (LCD, etc.), a speaker, etc.

[0137] The memory 104 may include read-only memory and random access memory, and provides instructions and data to the processor 101. A portion of the memory 104 may also include non-volatile random access memory. For example, the memory 104 may also store device type information.

[0138] In specific implementations, the processor 101, input device 102, and output device 103 described in the embodiments of the present invention can execute the implementation methods described in the embodiments of the knowledge graph construction method for airport cluster collaborative operation provided by the present invention, which will not be repeated here.

[0139] Accordingly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, implement the above-described knowledge graph construction method for airport cluster collaborative operation.

[0140] The computer-readable storage medium can be an internal storage unit of the system described in any of the foregoing embodiments, such as the system's hard disk or memory. The computer-readable storage medium can also be an external storage device of the system, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the system. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0141] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0142] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, or may be electrical, mechanical or other forms of connection.

[0143] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0144] Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units. When using each module, user information is collected and stored only with the user's full authorization and in compliance with relevant laws and regulations, protecting the security and privacy of user data, and strictly prohibiting unauthorized access; data processing will be conducted within the scope stipulated by law and will not exceed the purpose and scope authorized by the user; at the same time, users have the rights to access, correct, delete, restrict processing, and refuse their personal data; and must strictly comply with applicable laws and regulations and conduct compliance reviews.

[0145] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it 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 all or part of the 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 cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0146] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for constructing a knowledge graph for collaborative operation of airport clusters, characterized in that, include: Collect multi-source heterogeneous data generated during the collaborative operation of the airport cluster, and preprocess the heterogeneous data to obtain preprocessed text data; Construct an ontology model for the collaborative operation of airport clusters to define the semantic framework of the knowledge graph; Based on the defined knowledge graph semantic framework, entities, relations and attributes are automatically extracted from preprocessed text data to generate entity-relation-entity triple data; Based on the triplet data, knowledge storage and graph construction are completed to obtain a knowledge graph for the collaborative operation of the airport cluster.

2. The method as described in claim 1, characterized in that, The heterogeneous data is preprocessed as follows: Remove duplicate, outlier, and invalid data, and handle missing values; For the Chinese text data in the multi-source heterogeneous data, the Jieba word segmentation tool is used for word segmentation processing; For the professional terms in the field of airport cluster operation in the multi-source heterogeneous data, a civil aviation feature library is constructed, and the professional terms are added to the word segmentation dictionary; The multi-source heterogeneous data is normalized.

3. The method as described in claim 1, characterized in that, Define the semantic framework of the knowledge graph as follows: Define the core entity categories; these entity categories include flight entities, airport entities, airline entities, air traffic control entities, meteorological event entities, and flow control measure entities. Define semantic relationship types between multiple entities; the semantic relationship types include locateTo relationship, belongTo relationship, controlBy relationship, affect relationship and cause relationship; wherein, the locateTo relationship represents the take-off and landing positioning relationship between a flight and an airport, the belongTo relationship represents the affiliation relationship between a flight and an airline, the controlBy relationship represents the control relationship between a flight and an air traffic control unit, the affect relationship represents the influence relationship between entities, and the cause relationship represents the causal relationship; Define the set of attributes and attribute types for each entity category; the attribute types include numeric, character, and time types.

4. The method as described in claim 3, characterized in that, Generate entity-relation-entity triplet data, specifically: The BERT-BiGRU-Attention-CRF model was used to perform entity recognition on the preprocessed text data, and the recognition results were labeled with boundaries and types. The recognition results included multiple entities, namely flight number, airport code, time information, flow control measure name and meteorological event type. The BiGRU-Attention model is used to extract relations from the recognition results to obtain the extraction results; the extraction results include the relationship types between entities. Based on the identification and extraction results, generate entity-relationship-entity triplet data.

5. The method as described in claim 4, characterized in that, The BERT-BiGRU-Attention-CRF model consists of a BERT embedding layer, a BiGRU feature layer, an Attention layer, and a CRF output layer. The BERT embedding layer uses a BERT pre-trained language model to convert input text into context-sensitive dynamic word vectors; The BiGRU feature layer employs bidirectional gated recurrent units to extract contextual features of the text. The Attention layer introduces an attention mechanism to enhance the weight of key semantic information; The CRF output layer employs a conditional random field for global optimization of sequence labels.

6. The method as described in claim 4, characterized in that, The recognition results are subjected to relation extraction to obtain the extraction results, specifically: The identified entity pairs are used as input. The context information of the sentence in which the entity is located is encoded by bidirectional GRU to capture the semantic dependencies between entities. An attention mechanism is introduced to focus on semantic features that play a key role in relation judgment. The relationship type between entity pairs is output by a Softmax classifier.

7. The method as described in claim 6, characterized in that, The knowledge graph of airport cluster collaborative operation was obtained, specifically: Using the constructed ontology model as a semantic framework, entities are mapped to nodes in the graph database, relations are mapped to edges between nodes, and entity attributes are mapped to attribute key-value pairs of nodes to complete the storage of the triple data. After storing all the triplet data, a knowledge graph of airport cluster collaborative operation covering multiple subjects and multiple relationship types, including flights, airports, airlines, air traffic control units, meteorological events, and flow control measures, is formed.

8. The method as described in claim 7, characterized in that, After obtaining the knowledge graph of airport cluster collaborative operation, the method also includes establishing a dynamic update and evolution mechanism based on streaming data, specifically: The system utilizes the Apache Flink streaming engine to access real-time data streams, which include flight operation status updates, flow control command change data, and weather warning information. When the real-time streaming engine Apache Flink detects the arrival of new data, it triggers an incremental update operation. Generate a graph snapshot of the current running state at preset time intervals or based on event triggering; the graph snapshot represents the complete state of the knowledge graph at any given time.

9. A knowledge graph construction device for collaborative operation of airport clusters, characterized in that, The system includes a processor, an input device, an output device, and a memory, which are interconnected. The memory is used to store a computer program, which includes program instructions. The processor is configured to invoke the program instructions to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-8.

Images