An adaptive dynamic knowledge update method and system combining spatiotemporal data

Abstract

This invention targets the field of data resource construction, aiming to provide an adaptive dynamic knowledge update method and system that combines spatiotemporal data to solve the problem of associating and integrating spatiotemporal data (updated at the millisecond level) with static entities in knowledge graphs (updated at the minute / hour level). In practical applications, knowledge graphs containing only static data lack timeliness, are easily out of touch with the real-world environment, and have limited predictive and reasoning capabilities. Knowledge graphs integrated with spatiotemporal data better meet the needs of real-world scenarios, solve the problem of outdated knowledge graph information, and provide dynamic cognitive capabilities. This invention adaptively selects an appropriate update mechanism based on data update frequency, data importance, and environmental resource status, manages the entire update process, and integrates spatiotemporal data with graph entities for use, meeting the knowledge application needs under high timeliness requirements.

Description

Technical Field

[0001] This invention belongs to the field of data resource construction, and in particular relates to an adaptive dynamic knowledge update method and system that combines spatiotemporal data. Background Technology

[0002] With the digitalization and intelligentization of various industries, the construction of knowledge graphs is receiving increasing attention. These graphs accumulate domain rules and historical experience, providing support for analysis and decision-making. However, in practical applications, the knowledge within the graph is dynamically changing, and the timeliness of this knowledge is a crucial factor in the quality of the relation graph.

[0003] In existing technologies, the concept of dynamic knowledge graphs has been proposed frequently, but most of them start from the perspective of event graphs, considering the problems of ontology updates and entity updates. Research on how to associate and integrate spatiotemporal data with millisecond-level update frequency with graph entities and ensure the timeliness of data needs to be improved. Summary of the Invention

[0004] Purpose of the Invention: This invention aims to provide an adaptive dynamic knowledge update method and system that combines spatiotemporal data to solve the problems of outdated knowledge of static entities (updated every minute or hour) in knowledge graphs and the difficulty of associating and integrating it with frequently updated spatiotemporal data (updated every millisecond). Its applications are threefold: First, it addresses the issues of outdated knowledge and low-quality graph data in practical applications, which fail to effectively support business system applications; second, it expands the knowledge scope of knowledge graph management, associating and integrating spatiotemporal data with graph entities to support dynamic reasoning and real-time decision-making; and third, it adopts a reasonable data storage mechanism, adaptively adjusts the update mechanism, and monitors the execution status of update tasks, which helps ensure the overall stable operation of the system while meeting data timeliness requirements.

[0005] Technical solution: This invention provides an adaptive dynamic knowledge update method combining spatiotemporal data, comprising:

[0006] S1. Based on the timeliness requirements of data, knowledge data is divided into static knowledge data and real-time knowledge data, and the corresponding data storage mechanisms are used to store the knowledge data and the relationship between knowledge data of the same type.

[0007] S2. For real-time knowledge data that is updated frequently, a stream-batch integrated architecture is used to create knowledge update tasks by receiving data change notifications from information sources; for static knowledge data that is updated every minute or hour, a timed knowledge update mechanism is used.

[0008] S3. Adaptively select a knowledge update mechanism based on the timeliness requirements of the data, the importance of the data, and the status of environmental resources;

[0009] S4. Implement full-process management of knowledge update tasks, monitor the execution status of knowledge update tasks, and initiate compensation measures as needed for tasks that fail to execute.

[0010] S5. Using entity alignment, context association, and semantic association, obtain the relationship between static knowledge data and real-time knowledge data, and then integrate and apply the real-time knowledge data with the static knowledge data to obtain complete and accurate entity information.

[0011] Furthermore, the data storage mechanism described in S1 includes:

[0012] For static knowledge data, a graph database is used to store the static knowledge data and the relationships between static knowledge data.

[0013] For real-time knowledge data, use a big data platform or time-series database to store the real-time knowledge data and the relationships between real-time knowledge data.

[0014] Furthermore, the creation of the knowledge update task described in S2 includes:

[0015] Use Kafka as a message middleware, and for data sources that are updated frequently, subscribe to the message notification topic of that data source;

[0016] When the data source that is frequently updated changes, a data change message is sent to Kafka;

[0017] Upon receiving a data change message, the message is parsed to obtain the operation type, topic name, and data content. The primary key information of the data to be updated is extracted from the data content. A knowledge update task is created based on the operation type, topic name, and primary key attribute information, and the knowledge update task is placed in the knowledge update thread pool queue.

[0018] Furthermore, the timed knowledge update mechanism described in S2 includes:

[0019] For data updated on a minute or hourly basis, a timed knowledge update mechanism is set up, using fixed time points or fixed time intervals for updates; the timed knowledge update mechanism includes either non-incremental updates or incremental updates.

[0020] The non-incremental update refers to clearing the original data and then reloading the updated data each time a scheduled knowledge update task is executed;

[0021] The incremental update refers to the process of adding, deleting, or modifying data by comparing the differences between existing data in the knowledge graph and its corresponding source data each time an update task is executed, based on the differences. The comparison between existing data and source data uses a hash value verification method, including either MD5 or SHA-256.

[0022] Furthermore, the adaptive selection knowledge update mechanism described in S3 includes:

[0023] Based on the data characteristics of the original information source and the user's requirements for the timeliness of knowledge data, an initial knowledge update mechanism is manually set, including: for knowledge data for which the original information source has already sent change notification messages via Kafka and the user's timeliness requirement is at the second level, the update mechanism is set to a message notification-based knowledge update mechanism; for knowledge data for which the user's timeliness requirement is at the minute or hour level, the update mechanism is set to a scheduled knowledge update mechanism; for real-time knowledge data, either a message notification-based knowledge update mechanism or a scheduled knowledge update mechanism can be used; for static knowledge data, either a no-update mechanism or a scheduled update mechanism can be used.

[0024] The system automatically adjusts the update mechanism and priority settings based on the importance of data in the knowledge graph and the status of environmental resources. This includes: counting the daily access frequency of different types of data in the knowledge graph; setting the importance of different types of data in the knowledge graph based on the daily access frequency threshold over a continuous period of time, where importance is categorized as important, average, and unused; setting priority weights for knowledge update tasks of different types of data in the knowledge graph based on their importance; monitoring the server's resource status; issuing a server alarm when memory usage exceeds a preset memory threshold; and adjusting the knowledge update mechanism settings according to a degradation rule, whereby the time interval for updating knowledge data is doubled if its importance is average.

[0025] Furthermore, the end-to-end management of knowledge update tasks described in S4 includes:

[0026] The system periodically outputs the number of active worker threads, the number of completed tasks, the historical maximum thread count, and the number of tasks waiting in the queue in the knowledge update task thread pool.

[0027] Output the execution status of each task;

[0028] Based on log monitoring results, promptly grasp the status of knowledge update tasks in the system, adaptively adjust the update mechanism, or reset failed update tasks; when the number of waiting tasks in the queue exceeds the task threshold, adjust the settings of the knowledge update mechanism in the knowledge graph, double the time interval for data resources with general knowledge importance, or increase the resource status of the deployment environment; for tasks that fail to execute, manually select whether to re-add them to the task resource pool.

[0029] Furthermore, S5 describes using entity alignment, context association, and semantic association methods to obtain the association relationship between static knowledge data and real-time knowledge data, including:

[0030] The method of obtaining the association relationship using entity alignment is as follows: based on the attribute information in the real-time knowledge data in the spatiotemporal information, the inherent attribute corresponding to the attribute information is identified, the corresponding entity in the knowledge graph is found based on the inherent attribute, and the spatiotemporal information and the entity in the knowledge graph are aligned to obtain the association relationship between the static knowledge data of the entity in the knowledge graph and the real-time knowledge data in the spatiotemporal information.

[0031] The method of obtaining the relationship using context association is as follows: within a fixed scope of static knowledge data of an entity, spatiotemporal data within a certain time and space range are statistically analyzed, and the statistical analysis results are integrated with the static knowledge data of entities in the knowledge graph to obtain the relationship between static knowledge data and real-time knowledge data.

[0032] Obtaining associations using semantic association involves calculating the association between the semantic information of entities in the knowledge graph and spatiotemporal data based on preset state conditions and mapping conditions between object entities, thereby obtaining the association between static knowledge data and real-time knowledge data.

[0033] This invention also provides an adaptive dynamic knowledge update system that combines spatiotemporal data, comprising:

[0034] The storage module is used to divide knowledge data into static knowledge data and real-time knowledge data according to the timeliness requirements of the data, and to use the corresponding data storage mechanism to store the knowledge data and the relationship between knowledge data of the same type.

[0035] The update module is used for real-time knowledge data that is updated frequently. It uses a stream-batch integrated architecture and creates knowledge update tasks by receiving data change message notifications from information sources. For static knowledge data that is updated every minute or hour, a timed knowledge update mechanism is used.

[0036] The adaptive update mechanism module is used to adaptively select the knowledge update mechanism based on the timeliness requirements of the data, the importance of the data, and the status of environmental resources.

[0037] The monitoring and compensation module is used to manage the entire process of knowledge update tasks, monitor the execution status of knowledge update tasks, and initiate compensation measures as needed for tasks that fail to execute.

[0038] The fusion application module is used to obtain the relationship between static knowledge data and real-time knowledge data by using entity alignment, context association, and semantic association. It then associates and fuses the real-time knowledge data with the static knowledge data to obtain complete and accurate entity information.

[0039] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0040] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0041] Beneficial Effects: This invention provides an adaptive dynamic knowledge update method and system that combines spatiotemporal data. It utilizes different data storage mechanisms based on the timeliness requirements of the data and employs a stream-batch integrated architecture. For frequently updated data, knowledge update tasks are initiated by receiving data change notifications from information sources. For data updated every minute or hour, a timed data update mechanism is used. Based on the above data update mechanisms, the update mechanism is adaptively selected according to the timeliness requirements, data importance, and environmental resource status. To ensure the effective operation of the update mechanism, the entire knowledge update task process is managed, and the execution status of the knowledge update task is monitored. Entity alignment, context association, and semantic association are used to integrate spatiotemporal data with graph entities, enhancing the system's reasoning and decision-making capabilities.

[0042] Compared to existing knowledge update methods, this method has the following advantages: First, it proposes an architecture that physically distributes spatiotemporal data and graph entities for integrated logical application, using different knowledge storage mechanisms according to the timeliness requirements of the data, thus expanding the scope of traditional graph management knowledge. Second, it refines the real-time data update mechanism and the timed data update mechanism based on message notification, and adaptively selects the appropriate data update mechanism according to the characteristics of data types, managing the entire process of data update tasks to ensure the normal execution of update tasks, demonstrating strong engineering practicality. Third, it proposes spatiotemporal data and graph entity association fusion methods such as entity alignment, context association, and semantic association to obtain complete and accurate entity information, thereby better improving the system's reasoning and decision-making capabilities. Attached Figure Description

[0043] Figure 1 A flowchart for adaptive dynamic knowledge updates that incorporate spatiotemporal data.

[0044] Figure 2 System architecture diagram for adaptive dynamic knowledge update that incorporates spatiotemporal data. Detailed Implementation

[0045] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0046] Example 1

[0047] Please see Figures 1 to 2 As shown, this invention provides an adaptive dynamic knowledge update method combining spatiotemporal data, comprising:

[0048] S1. Based on the timeliness requirements of data, knowledge data is divided into static knowledge data and real-time knowledge data, and the corresponding data storage mechanisms are used to store the knowledge data and the relationship between knowledge data of the same type.

[0049] S2. For real-time knowledge data that is updated frequently, a stream-batch integrated architecture is used to create knowledge update tasks by receiving data change notifications from information sources; for static knowledge data that is updated every minute or hour, a timed knowledge update mechanism is used.

[0050] S3. Adaptively select a knowledge update mechanism based on the timeliness requirements of the data, the importance of the data, and the status of environmental resources;

[0051] S4. Implement full-process management of knowledge update tasks, monitor the execution status of knowledge update tasks, and initiate compensation measures as needed for tasks that fail to execute.

[0052] S5. Using entity alignment, context association, and semantic association, obtain the relationship between static knowledge data and real-time knowledge data, and then integrate and apply the real-time knowledge data with the static knowledge data to obtain complete and accurate entity information.

[0053] In this embodiment,

[0054] In S1, for static knowledge data, a graph database is used to store the static knowledge data and the relationships between static knowledge data; for real-time knowledge data, such as spatiotemporal data where time and location are constantly changing, a large database platform or time series database is used to store the real-time knowledge data and the relationships between real-time knowledge data; to improve retrieval capabilities, indexing platforms such as Elasticsearch and Solr can be used to store snapshot data.

[0055] The creation of the knowledge update task described in S2 includes: using Kafka as a message middleware, subscribing to the message notification topic of the data source for high-frequency updates; when the high-frequency update data source changes, sending a data change message to Kafka; the data format of the change message is shown in Table 1.

[0056] Table 1 Change Message Data Format

[0057]

[0058] Upon receiving a data change message, the message is parsed to obtain the operation type, topic name, and data content. The primary key information of the data to be updated is extracted from the data content. A knowledge update task is created based on the operation type, topic name, and primary key attribute information, and the knowledge update task is placed in the knowledge update thread pool queue.

[0059] The scheduled knowledge update mechanism includes: setting a scheduled knowledge update mechanism for minute- or hourly updated data, using fixed time points or fixed time intervals for updates; the scheduled knowledge update mechanism includes non-incremental updates and incremental updates; the non-incremental update refers to clearing the original data and reloading the updated data each time a scheduled knowledge update task is executed; the incremental update refers to comparing the differences between the existing data in the knowledge graph and the source data in the update task each time an update task is executed, and performing add, delete, and modify operations based on the differences; for example, if there are 100 records of aircraft type data in the knowledge graph, and these 100 records are extracted from the domain database table zzll_jx at a certain moment, when the number of records in the zzll_jx table becomes 110, the incremental update first compares the differences between the 100 existing aircraft type data in the graph and the 110 records in the corresponding source domain database table zzll_jx, and keeps the aircraft type data in the knowledge graph consistent with the data in the domain database table zzll_jx based on the differences, thus avoiding the data in the knowledge graph becoming outdated. The differences between existing data and source data are compared using hash value verification methods, including MD5 and SHA-256.

[0060] In S3, based on the data characteristics of the original information source and the user's timeliness requirements for knowledge data, an initial knowledge update mechanism is manually set. This includes: for knowledge data where the original information source has already sent change notification messages via Kafka and the user's timeliness requirement is at the second level, the update mechanism is set to a message notification-based knowledge update mechanism; for knowledge data where the user's timeliness requirement is at the minute or hour level, the update mechanism is set to a scheduled knowledge update mechanism; for real-time knowledge data, a message notification-based knowledge update mechanism or a scheduled knowledge update mechanism can be used; for static knowledge data, a no-update mechanism or a scheduled update mechanism can be used.

[0061] The daily access frequency of different types of data in the knowledge graph is statistically analyzed. The importance of different types of data in the knowledge graph is set according to the daily access frequency threshold over a continuous period of time. The importance is divided into important, average, and unused. Priority weights are set for knowledge update tasks of different types of data in the knowledge graph according to their importance.

[0062] The server monitors the resource status and issues an alarm when memory usage exceeds 90%. It then adjusts the knowledge update mechanism settings according to the degradation rules, which stipulate that if the importance of the knowledge data is considered average, the time interval for its scheduled updates is doubled.

[0063] S4 periodically outputs the number of active worker threads, the number of completed tasks, the historical maximum thread count, and the number of tasks waiting in the queue in the knowledge update task thread pool; it also outputs the execution status of each task; based on log monitoring results, it promptly grasps the status of knowledge update tasks in the system, adaptively adjusts the update mechanism, or resets failed update tasks; when the number of tasks waiting in the queue exceeds 300, it adjusts the settings of the knowledge update mechanism in the knowledge graph, doubling the time interval for data resources with general knowledge importance, or increasing the resource status of the deployment environment; for tasks that fail to execute, it allows manual selection of whether to re-add them to the task resource pool.

[0064] In S5, entity alignment is used to obtain associations as follows: Based on attribute information (such as target ID, target model, time, location, region, object name, object type, object identifier, etc.) in real-time knowledge data within spatiotemporal information, the inherent attributes corresponding to these attributes are identified. The corresponding entities in the knowledge graph are found based on these inherent attributes, and the spatiotemporal information is aligned with the entities in the knowledge graph to obtain the association between the static knowledge data of the entities in the knowledge graph and the real-time knowledge data in the spatiotemporal information. The entity alignment uses the following formula to calculate the similarity score. When the score exceeds a threshold, the spatiotemporal information is used as a dynamic attribute and fused with the entities in the knowledge graph. For example, the formula for calculating the similarity score between the target model and the entity name is:

[0065] ;

[0066] Where A represents the target model attribute value, and B={B1,B2,...Bn} represents the entity name attribute value and its set of synonyms;

[0067] Taking an aircraft (Boeing 737) entity as an example, the aircraft model and manufacturer are static knowledge data, while the aircraft's location, remaining fuel, flight speed, and local weather conditions are real-time knowledge data. The fusion of static and real-time knowledge data refers to identifying the aircraft as a Boeing 737 based on its flight characteristics, and then merging it with the Boeing 737 in the static knowledge into a fused entity that includes both the inherent attributes and real-time attribute information of the Boeing 737.

[0068] The method of obtaining associations using contextual association is as follows: Within a fixed range of static knowledge data of an entity, spatiotemporal data within a certain time and space range are statistically analyzed. The statistical analysis results are then fused with the static knowledge data of the entity in the knowledge graph to obtain the association between static knowledge data and real-time knowledge data. For example, given a spatial region entity (AMC-315), whose center point coordinates and radius are known, this is static knowledge data. The number of aircraft in this spatial region is counted as real-time knowledge and fused with the static knowledge of this region to generate the spatial association between the current entity and similar targets within the region, thus obtaining complete information about the region (AMC-315).

[0069] Obtaining associations using semantic association involves calculating the association between the semantic information of entities in the knowledge graph and spatiotemporal data based on preset state conditions and mapping conditions between object entities, thus obtaining the association between static knowledge data and real-time knowledge data. For example, if the preset aircraft type A has an altitude range of [500-1000km] corresponding to an event entity (low-altitude reconnaissance), when an airborne facility that meets this condition is found in real-time knowledge, it is associated with the event entity (low-altitude reconnaissance).

[0070] Example 2

[0071] Please see Figures 1 to 2 As shown in Embodiment 1, this invention provides an adaptive dynamic knowledge update system that combines spatiotemporal data, comprising:

[0072] The storage module is used to divide knowledge data into static knowledge data and real-time knowledge data according to the timeliness requirements of the data, and to use the corresponding data storage mechanism to store the knowledge data and the relationship between knowledge data of the same type.

[0073] The update module is used for real-time knowledge data that is updated frequently. It uses a stream-batch integrated architecture and creates knowledge update tasks by receiving data change message notifications from information sources. For static knowledge data that is updated every minute or hour, a timed knowledge update mechanism is used.

[0074] The adaptive update mechanism module is used to adaptively select the knowledge update mechanism based on the timeliness requirements of the data, the importance of the data, and the status of environmental resources.

[0075] The monitoring and compensation module is used to manage the entire process of knowledge update tasks, monitor the execution status of knowledge update tasks, and initiate compensation measures as needed for tasks that fail to execute.

[0076] The fusion application module is used to obtain the relationship between static knowledge data and real-time knowledge data by using entity alignment, context association, and semantic association. It then associates and fuses the real-time knowledge data with the static knowledge data to obtain complete and accurate entity information.

[0077] In this embodiment,

[0078] In the storage module, for static knowledge data, a graph database is used to store the static knowledge data and the relationships between static knowledge data; for real-time knowledge data, such as spatiotemporal data that changes constantly in time and location, a big data platform or time-series database is used to store the real-time knowledge data and the relationships between real-time knowledge data; to improve retrieval capabilities, indexing platforms such as Elasticsearch and Solr can be used to store snapshot data.

[0079] The creation of knowledge update tasks described in the update module includes: using Kafka as a message middleware, subscribing to the message notification topic of a frequently updated data source; sending a data change message to Kafka when the frequently updated data source changes; the data format of the change message is shown in Table 1.

[0080] Upon receiving a data change message, the message is parsed to obtain the operation type, topic name, and data content. The primary key information of the data to be updated is extracted from the data content. A knowledge update task is created based on the operation type, topic name, and primary key attribute information, and the knowledge update task is placed in the knowledge update thread pool queue.

[0081] The scheduled knowledge update mechanism includes: setting a scheduled knowledge update mechanism for minute- or hourly updated data, using fixed time points or fixed time intervals for updates; the scheduled knowledge update mechanism includes non-incremental updates and incremental updates; the non-incremental update refers to clearing the original data and reloading the updated data each time a scheduled knowledge update task is executed; the incremental update refers to comparing the differences between the existing data in the knowledge graph and the source data in the update task each time an update task is executed, and performing add, delete, and modify operations based on the differences; for example, if there are 100 records of aircraft type data in the knowledge graph, and these 100 records are extracted from the domain database table zzll_jx at a certain moment, when the number of records in the zzll_jx table becomes 110, the incremental update first compares the differences between the 100 existing aircraft type data in the graph and the 110 records in the corresponding source domain database table zzll_jx, and keeps the aircraft type data in the knowledge graph consistent with the data in the domain database table zzll_jx based on the differences, thus avoiding the data in the knowledge graph becoming outdated. The differences between existing data and source data are compared using hash value verification methods, including MD5 and SHA-256.

[0082] The adaptive update mechanism module combines the data characteristics of the original information source with the user's timeliness requirements for knowledge data. It manually sets the initial knowledge update mechanism, including: for knowledge data where the original information source has already sent change notification messages via Kafka and the user's timeliness requirement is at the second level, the update mechanism is set to a message notification-based knowledge update mechanism; for knowledge data where the user's timeliness requirement is at the minute or hour level, the update mechanism is set to a scheduled knowledge update mechanism; for real-time knowledge data, a message notification-based knowledge update mechanism or a scheduled knowledge update mechanism can be used; for static knowledge data, a no-update mechanism or a scheduled update mechanism can be used.

[0083] The daily access frequency of different types of data in the knowledge graph is statistically analyzed. The importance of different types of data in the knowledge graph is set according to the daily access frequency threshold over a continuous period of time. The importance is divided into important, average, and unused. Priority weights are set for knowledge update tasks of different types of data in the knowledge graph according to their importance.

[0084] The server monitors the resource status and issues an alarm when memory usage exceeds 90%. It then adjusts the knowledge update mechanism settings according to the degradation rules, which stipulate that if the importance of the knowledge data is considered average, the time interval for its scheduled updates is doubled.

[0085] The monitoring and compensation module periodically outputs the number of active worker threads, the number of completed tasks, the historical maximum thread count, and the number of tasks waiting in the queue in the knowledge update task thread pool; it also outputs the execution status of each task; based on the log monitoring results, it promptly grasps the status of knowledge update tasks in the system and adaptively adjusts the update mechanism or resets failed update tasks; when the number of tasks waiting in the queue exceeds 300, it adjusts the settings of the knowledge update mechanism in the knowledge graph, doubling the time interval for data resources with general knowledge importance, or increasing the resource status of the deployment environment; for tasks that fail to execute, it allows manual selection of whether to re-add them to the task resource pool.

[0086] The entity alignment method used in the fusion application module to obtain associations is as follows: Based on attribute information (such as target ID, target model, time, location, region, object name, object type, object identifier, etc.) in real-time knowledge data within the spatiotemporal information, the inherent attributes corresponding to these attributes are identified. The corresponding entities in the knowledge graph are found based on these inherent attributes, and the spatiotemporal information is aligned with the entities in the knowledge graph to obtain the association between the static knowledge data of the entities in the knowledge graph and the real-time knowledge data in the spatiotemporal information. The entity alignment uses the following formula to calculate the similarity score. When the score exceeds a threshold, the spatiotemporal information is used as a dynamic attribute and fused with the entities in the knowledge graph. For example, the formula for calculating the similarity score between the target model and the entity name is:

[0087] ;

[0088] Where A represents the target model attribute value, and B={B1,B2,...Bn} represents the entity name attribute value and its set of synonyms;

[0089] Taking an aircraft (Boeing 737) entity as an example, the aircraft model and manufacturer are static knowledge data, while the aircraft's location, remaining fuel, flight speed, and local weather conditions are real-time knowledge data. The fusion of static and real-time knowledge data refers to identifying the aircraft as a Boeing 737 based on its flight characteristics, and then merging it with the Boeing 737 in the static knowledge into a fused entity that includes both the inherent attributes and real-time attribute information of the Boeing 737.

[0090] The method of obtaining associations using contextual association is as follows: Within a fixed range of static knowledge data of an entity, spatiotemporal data within a certain time and space range are statistically analyzed. The statistical analysis results are then fused with the static knowledge data of the entity in the knowledge graph to obtain the association between static knowledge data and real-time knowledge data. For example, given a spatial region entity (AMC-315), whose center point coordinates and radius are known, this is static knowledge data. The number of aircraft in this spatial region is counted as real-time knowledge and fused with the static knowledge of this region to generate the spatial association between the current entity and similar targets within the region, thus obtaining complete information about the region (AMC-315).

[0091] Obtaining associations using semantic association involves calculating the association between the semantic information of entities in the knowledge graph and spatiotemporal data based on preset state conditions and mapping conditions between object entities, thus obtaining the association between static knowledge data and real-time knowledge data. For example, if the preset aircraft type A has an altitude range of [500-1000km] corresponding to an event entity (low-altitude reconnaissance), when an airborne facility that meets this condition is found in real-time knowledge, it is associated with the event entity (low-altitude reconnaissance).

Claims

1. An adaptive dynamic knowledge updating method for integrating spatio-temporal data, characterized in that, Comprise: S1, according to the data timeliness requirement, the knowledge data is divided into static knowledge data, real-time knowledge data, and the corresponding data storage mechanism is used to store the association between knowledge data and the same type of knowledge data; S2, for high-frequency updating of real-time knowledge data, using a stream batch integrated architecture, through the data change message notification of the receiving source, create knowledge update task;For minutes or hours of static knowledge data, use the timing knowledge update mechanism; S3, for the original source has sent change notification message through kafka, and the user's knowledge data timeliness requirement is second-level error, set the update mechanism to knowledge update mechanism based on message notification;For the user's knowledge data timeliness requirement is minute or hour level error, set the update mechanism to timing knowledge update mechanism;According to the timeliness requirement of data, the importance of data, the state of environment resources, adaptively select the knowledge update mechanism; S4, the knowledge update task is managed in the whole process, the execution state of the knowledge update task is monitored, and the compensation measures are started as needed for the failed task; S5, using entity alignment, context association, semantic association method, the association between static knowledge data and real-time knowledge data is obtained, the real-time knowledge data and static knowledge data are associated and fused, and complete and accurate entity information is obtained.

2. The adaptive dynamic knowledge updating method of spatio-temporal data according to claim 1, characterized in that, The data storage mechanism in S1 comprises: For static knowledge data, use graph database to store static knowledge data and the association between static knowledge data; For real-time knowledge data, use large database platform or time series database to store real-time knowledge data and the association between real-time knowledge data.

3. The adaptive dynamic knowledge updating method of spatio-temporal data according to claim 2, characterized in that, The creation of knowledge update task in S2 comprises: Use kafka as a message middleware, for high-frequency updating data source, subscribe to the data source message notification theme; When the high-frequency updating data source changes, send data change message to kafka; After receiving the data change message, parse the change message, get the operation type, theme name and data content, extract the data primary key information to be updated in the data content;According to the operation type, theme name and primary key attribute information, create a knowledge update task, and put the knowledge update task into the knowledge update thread pool queue.

4. The adaptive dynamic knowledge updating method of spatio-temporal data according to claim 3, characterized in that, The timing knowledge update mechanism in S2 comprises: For minute or hour level updating data, set the timing knowledge update mechanism, adopt fixed time point or fixed time interval update;The timing knowledge update mechanism comprises any one of non-incremental update and incremental update; The non-incremental update refers to that each time the timing knowledge update task is executed, the original data is cleared, and the updated data is reloaded; The incremental update refers to that each time the update task is executed, the difference between the existing data in the knowledge graph and the corresponding source data is compared, and the difference is combined to add, delete or modify;The difference between the existing data and the source data is compared using hash value checking method, including any one of MD5 and SHA-256.

5. The adaptive dynamic knowledge updating method of spatio-temporal data according to claim 4, characterized in that, The adaptive selection of knowledge update mechanism in S3 comprises: According to the data characteristics of the original source and the timeliness requirement of the user for the knowledge data, an initialized knowledge updating mechanism is set, including: for real-time knowledge data, any one of a knowledge updating mechanism based on message notification and a timing knowledge updating mechanism; for static knowledge data, any one of a non-updating mechanism and a timing updating mechanism; According to the importance of the data in the knowledge graph, the automatic adjustment of the updating mechanism and the priority setting of the environment resource state, including: counting the daily access frequency of different types of data in the knowledge graph, setting the importance of different types of data in the knowledge graph according to the daily access frequency threshold in a continuous period of time, the importance is divided into important, general and not used; setting the priority weight of the knowledge updating task of different types of data in the knowledge graph according to the importance; monitoring the resource state of the server, when the memory usage exceeds the preset memory threshold, the server alarms, and adjusts the setting of the knowledge updating mechanism according to the degradation rule, the degradation rule is: if the importance of the knowledge data is general, double the time interval of the timing updating.

6. The adaptive dynamic knowledge updating method of spatio-temporal data according to claim 5, characterized in that, The full-process management of the knowledge updating task in S4 includes: Timing output of the number of active working threads, the number of completed tasks, the historical maximum thread data, and the number of waiting tasks in the queue in the knowledge updating task thread pool; Output the execution state of each task; According to the log monitoring result, the knowledge updating task situation in the system is grasped in time, and the updating mechanism is adjusted or the failed updating task is reset; when the number of waiting tasks in the queue exceeds the task threshold, the setting of the knowledge updating mechanism in the knowledge graph is adjusted, the timing updating time interval of the data resource with general importance is doubled, or the deployment environment resource state is increased; for the failed task, whether to rejoin the task resource pool is manually selected.

7. The adaptive dynamic knowledge updating method of spatio-temporal data according to claim 6, characterized in that, The association relationship between the static knowledge data and the real-time knowledge data is obtained by using the entity alignment, context association and semantic association methods in S5, including: The association relationship obtained by using the entity alignment method is: identifying the inherent attribute corresponding to the attribute information in the real-time knowledge data in the spatio-temporal information according to the attribute information, finding the corresponding entity in the knowledge graph according to the inherent attribute, and performing entity alignment on the spatio-temporal information and the entity in the knowledge graph to obtain the association relationship between the static knowledge data of the entity in the knowledge graph and the real-time knowledge data in the spatio-temporal information; The association relationship obtained by using the context association method is: within a certain fixed entity static knowledge data range, the spatio-temporal data within a certain time and space range is counted, and the statistical analysis result is fused with the static knowledge data of the entity in the knowledge graph to obtain the association relationship between the static knowledge data and the real-time knowledge data; The association relationship obtained by using the semantic association method is: the semantic information of the entity in the knowledge graph and the spatio-temporal data are calculated according to the preset state condition and object entity mapping condition to obtain the association relationship between the static knowledge data and the real-time knowledge data.

8. An adaptive dynamic knowledge updating system incorporating spatio-temporal data, characterized by, including: The storage module is used for dividing the knowledge data into static knowledge data and real-time knowledge data according to the data timeliness requirement, and storing the knowledge data and the association relationship between the knowledge data of the same type by using a corresponding data storage mechanism; The updating module is used for creating a knowledge updating task by receiving a data change message notification of a source for the real-time knowledge data with high frequency updating, and using a timing knowledge updating mechanism for the static knowledge data with minute or hour level updating; The adaptive updating mechanism module is used for setting an updating mechanism as a knowledge updating mechanism based on a message notification for the knowledge data with a second level error of the timeliness requirement of the user, and setting the updating mechanism as a timing knowledge updating mechanism for the knowledge data with a minute or hour level error of the timeliness requirement of the user; and adaptively selecting a knowledge updating mechanism according to the timeliness requirement, the data importance and the environment resource state of the data; The monitoring and compensation module is used for performing full-process management on the knowledge updating task, monitoring an execution state of the knowledge updating task, and starting a compensation measure as required for a task with a failed execution; The fusion application module is used for acquiring an association relationship between the static knowledge data and the real-time knowledge data by using an entity alignment, context association and semantic association manner, associating and fusing the real-time knowledge data with the static knowledge data, and acquiring complete and accurate entity information.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 7.

Images