Relation graph analysis method based on time sequence

Abstract

The invention relates to the technical field of knowledge maps, and discloses a time sequence-based relationship map analysis method, which comprises the following steps of: firstly, collecting original data containing entities and relationships, each piece of data having a timestamp, and then extracting the entities, the relationships and the timestamps in the data; secondly, forming time sequence data from the extracted data, performing statistics and aggregation analysis on the time sequence data in each time window, and calculating key indexes; the entities, the relations and time-varying attributes of the entities and the relations are stored in a graph database, and timestamps and key indexes calculated in time windows are stored in a non-relational database; interacting and analyzing the time sequence data, and identifying the change mode and process of the entity and the relationship along with the time; and finally, displaying data generated by interaction and analysis based on dynamic and static visualization tools. According to the method, the dynamic time sequence graph is created, and the time sequence data and the entity relation data are combined, so that the comprehensive analysis of the network structure along with the time change is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a time-series-based relationship graph analysis method. Background Art

[0002] Traditional knowledge graph technologies have demonstrated strong capabilities in describing and analyzing relationships between data. Through graph database engines, they can build complex networks of entities and their relationships, and then use visualization tools to intuitively display these abstract connections. The application of algorithms such as graph search and path planning further enhances the depth of analysis and mining of these networks. However, the main limitation of this methodology is its static nature - that is, data and relationships are regarded as fixed snapshots, and fail to fully consider the dynamic changes of entities and their relationships in the real world. Over time, the attributes of entities and the relationships between entities will continue to evolve. Traditional knowledge graphs have difficulty capturing the details of these changes, so they have shortcomings in reflecting the real-time changes and evolution of network structures.

[0003] On the other hand, time series analysis technology focuses on the macro trends of data development over time, identifying anomalies and trends by establishing baselines and predictive models, but such methods often focus on the attribute changes of a single entity and fail to fully consider the temporal evolution of the relationship between entities. Although the relationship graph shows the complexity of the interaction between entities through visualization, the lack of the time dimension limits its application depth in dynamic data analysis scenarios. Especially when faced with complex data sets containing rich timestamp information, existing technologies often find it difficult to effectively integrate time series features with relationship network structures, and lack the ability to conduct precise mining and analysis at the micro level. Summary of the invention

[0004] In order to solve the problems and deficiencies in the above-mentioned prior art, the present invention specifically proposes a time-series-based relationship graph analysis method. The present invention creates a dynamic time-series knowledge graph. The core of the present invention is to combine time series data with entity relationship data to achieve a comprehensive analysis of network structure changes over time. The present invention can be used to process complex data containing time dimensions, construct time-series knowledge graphs, and perform in-depth analysis and prediction. It is suitable for multiple fields such as network security, financial analysis, and social network research, and has broad application prospects.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is as follows: On the one hand, the present invention proposes a time-series-based relationship graph analysis method, which mainly includes the following steps: Step S1. Collecting raw data, the raw data includes entity information and relationship data between entities, and each piece of data includes a timestamp; Step S2. Extracting entity information, relationship data between entities and timestamp data in the original data; Step S3. Form time series data based on the extracted data, and divide the time series data into time windows; then perform statistical and aggregate analysis on the time series data in each time window, and calculate key indicators; Step S4. Design a graph database to store entity information and relationship data between entities and their attributes that change over time, and a non-relational database to store timestamps and key indicator data calculated within a time window; design a database query strategy and optimize indexes; Step S5. Based on the user's query instructions, interact and analyze the time series data to identify the change patterns and processes of entity information and relationships between entities over time; Step S6. The data generated by the interaction and analysis are used to display the changing patterns and processes of entity information and relationships between entities over time based on dynamic and static visualization tools.

[0006] Preferably, step S2 also includes preprocessing the extracted timestamp data, and the preprocessing includes: first identifying the timestamp field in the data, then using the corresponding parsing function or library to convert the timestamp into an operable date and time object, and finally performing unified time zone, format standardization and timestamp completion operations on the extracted timestamp.

[0007] Preferably, in step S2, for the extracted entity information and the relationship data between entities, if it is unstructured text, natural language processing technology or pattern recognition algorithm is used to extract the entity information and the relationship data between entities; if it is structured text, the fields are directly mapped to entities, and then the entity information and the relationship between entities in the data are derived based on the semantic rules of structured field naming and value content.

[0008] Preferably, step S3 specifically includes the following contents: Sort the data by timestamp to form time series data; Divide the time series data into set time windows, perform statistical and aggregate analysis on the time series data within the windows, and calculate key indicators based on the divided time windows; Integrate timestamps, entity information and relationships between entities, time windows, key indicators within windows, and other related data to form a timeline dataset.

[0009] Preferably, the designed graph database has a time attribute index.

[0010] Preferably, the aggregate analysis may be performed based on data in a single continuous time window, or based on data in multiple non-continuous time windows.

[0011] Preferably, in step S3, the aggregate analysis includes calculating the average, median, maximum, minimum and sum of the key indicators.

[0012] Preferably, in step S5, interacting and analyzing time series data includes: based on the user's query instructions, filtering the timeline data of a specific time range, applying a timing analysis algorithm to analyze the timeline data to identify the changing patterns and processes of entity information and relationships between entities over time.

[0013] Preferably, in step S5, the interaction and analysis of the time series data includes performing indicator operations and visualization operations on the data.

[0014] Preferably, dynamic and static visualization tools are used to display the change patterns and processes of entity information and relationships between entities over time, including: Macro data rendering: When the time window is greater than or equal to the preset window, a specific data rendering method is used to show the trend of the overall data within the time window; Micro data rendering: When the time window is smaller than the preset window, a specific data rendering method is used to display the status of specific entity information or the relationship between entities in a specific time window or a specific time point.

[0015] On the other hand, the present invention also proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable in the processor, wherein when the processor executes the computer program, the above-mentioned relationship graph analysis method is implemented.

[0016] On the other hand, the present invention further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed in a computer processor, the above-mentioned relationship graph analysis method is implemented.

[0017] Beneficial effects of the present invention: 1. The present invention combines time series data with entity relationship data to create a dynamic time series knowledge graph to achieve a comprehensive analysis of network structure changes over time. The present invention can be used to process complex data containing time dimensions, construct a time series knowledge graph, and perform in-depth analysis and prediction. It is suitable for network security, financial analysis, social network research and other fields, and has broad application prospects.

[0018] 2. The present invention provides rendering strategies for different time windows to meet the needs of visualization analysis in different dimensions and time scales. Users are allowed to select different time windows as needed, making the analysis more flexible and adaptable to different analysis needs and scenarios.

[0019] 3. The present invention adopts two data storage strategies, storing statistical data of different time windows as non-relational data, and storing graph data as graph data, to meet the real-time or periodic update of entity attributes and relationships, and ensure that the knowledge graph reflects the latest data. Non-relational databases can process high-throughput data streams in real time, provide low-latency responses, and support horizontal expansion. They can seamlessly expand storage and processing capabilities as the amount of data grows, allowing various types of data to be stored without a strict data table structure, simplifying data storage and query. Graph databases can more intuitively represent and analyze complex relationships between entities. On the other hand, the query language of graph databases can be used to perform complex time series analysis to meet the needs of finding all relationships involving specific entities, or discovering abnormal behavior patterns within a certain time window. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The foregoing and following detailed description of the present invention will become more apparent when read in conjunction with the following drawings, in which: Figure 1 is a flow chart of the method of the present invention; Figure 2 This is a time granularity visualization interactive effect diagram of the present invention; Figure 3 This is a visualization effect diagram of the macroscopic thermal map of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions for achieving the purpose of the present invention will be further described below through several specific embodiments. It should be noted that the technical solutions claimed for protection in the present invention include but are not limited to the following embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0022] At present, the existing relationship graph analysis methods usually have the following defects and shortcomings: (1) Limitations of static graphs: The traditional knowledge graph update mechanism is inflexible and it is difficult to reflect changes in entity attributes and relationships in real time.

[0023] (2) Neglect of the time dimension: Time series analysis and relationship graphs fail to effectively integrate time series features with the dynamic changes in relationships between entities. The strength and nature of relationships between entities change over time, and this change is crucial for understanding patterns and detecting anomalies.

[0024] (3) Insufficient micro-analysis capabilities: Existing technologies are relatively mature in macro-trend analysis, but at the micro level, that is, when conducting in-depth analysis of the time series of specific events or behaviors, there is a lack of effective tools and methods.

[0025] Based on this, an embodiment of the present invention proposes a time-series-based relationship graph analysis method. The present invention creates a dynamic time-series knowledge graph, combines time series data with entity relationship data, to achieve a comprehensive analysis of how the network structure changes over time, and provides rendering strategies for different time windows to meet the needs of visual analysis in different dimensions and time scales.

[0026] First of all, the following concepts need to be explained.

[0027] Entity Recognition (ER): Also known as Named Entity Recognition (NER), it is an important task in the field of natural language processing (NLP). It usually relies on machine learning algorithms and deep learning models. It learns the characteristics and patterns of entities by training a large number of labeled data sets. It can identify and classify entities with specific meanings from text, such as names of people, places, names of organizations, time, currency values, etc.

[0028] Relation Extraction (RE): It is a key task in the field of natural language processing (NLP). Its core purpose is to automatically identify and extract semantic relationships between entities from text. These relationships can be relationships between people and organizations (such as "works at", "affiliated with"), relationships between two organizations (such as "merger", "acquisition"), relationships between people (such as "spouse", "parents"), etc. Relation extraction is an important part of building knowledge graphs, information retrieval, question-answering systems and other applications.

[0029] Materialized view: It is a special type of data object in the database, which stores a data set generated by the query results of one or more tables. Unlike ordinary views, which do not actually store data but dynamically query and calculate results from base tables at runtime, materialized views pre-calculate and store query results when they are created or regularly updated, so that they can be used directly when needed without having to execute complex queries again, thereby significantly improving query performance. Materialized views are usually used in the following scenarios: Acceleration of complex queries: When there are complex queries in the database, such as connections involving multiple tables, aggregate functions, etc., materialized views can pre-calculate these results to avoid repeated calculations for each query. Frequently accessed aggregate data: If an application needs to frequently access some aggregate data, such as statistical data, summary information, etc., materialized views can store these aggregate results to improve query efficiency. Data warehouse: In a data warehouse environment, materialized views are used to store pre-calculated summary data to support fast report generation and data analysis.

[0030] Column storage: It is a data storage method, which is opposite to the traditional row-based storage. In row-based storage, data is stored in the form of rows, that is, each row corresponds to a record, and the storage is performed in the order of the records. In column storage, data is stored in the form of columns, that is, the data in each column is stored and managed as a whole. Column storage is widely used in data warehouses, big data analysis platforms and OLAP (online analytical processing) systems, such as Apache Parquet, ORC, Google BigQuery, etc.

[0031] Relationship graph: also commonly known as knowledge graph, is a data model that represents entities and their relationships in the form of a graph structure. In the relationship graph, entities are represented as nodes, and the relationships between entities are represented by edges.

[0032] As the most basic implementation method of the present invention, this embodiment discloses a time-series-based relationship graph analysis method. Figure 1 , the method mainly comprises the following steps: Step S1. Collect raw data from multiple data sources. The data sources provide rich entity information and relationship data between entities, and each piece of data contains a timestamp, which can be used for subsequent time series data analysis.

[0033] In the embodiments described in the present invention, data collection and extraction covers all log data containing timestamps. Generally speaking, data sources are divided into structured data and heterogeneous unstructured data. In some embodiments, structured data includes database data, json data, and csv data. Heterogeneous unstructured data includes network logs, transaction records, social media interactions, etc.

[0034] Furthermore, network logs usually record all activities of network devices and systems, including user access records, system status, error messages, etc. Transaction records involve data on financial transactions, including information such as transaction time, amount, and transaction parties. Social media interactions include various activities of users on social media platforms, such as posting, commenting, and liking.

[0035] It is understandable that in a relationship graph, entity information refers to nodes in the graph, which can be users, devices, accounts, etc. In some embodiments, such as in the field of network security, entities can also be IP addresses, domain names, user accounts, etc. Relationship data refers to the connection between entities, i.e., edges, which describe a certain association or influence between entities, such as "a user visited a website", "a transfer was made between accounts", etc.

[0036] Step S2: preprocessing the collected raw data, including extracting the timestamp of the raw data and unifying the format, and extracting the entity information and the relationship between entities in the data.

[0037] In the embodiments described in the present invention, timestamp data usually exists in the data in different formats. The steps to extract these timestamp data are as follows: (1) Identify the timestamp field: First, it is necessary to identify the timestamp field in the data, which requires familiarity with the structure and format of the data source; (2) Parse the timestamp: Timestamps may exist in multiple formats, such as Unix timestamps, ISO8601 format, custom format, etc. Use the corresponding parsing function or library to convert the timestamp into an operable date and time object; (3) Handle the time zone issue: Data spans different time zones, and time zones need to be processed and converted to ensure that all timestamps are compared in the same time zone. In some special cases, the timestamp format may also be standardized and the timestamp may be completed.

[0038] Furthermore, for extracting entity information and relationships between entities, the present invention uses entity recognition (ER) technology to identify entities in the data. These entities can be people, organizations, places, time, numbers, etc. In the field of network security, entities may also include IP addresses, domain names, user IDs, device IDs, etc. Entity recognition is usually implemented in the following ways: (1) Rule matching: using predefined rules or regular expressions to identify entities; (2) Machine learning: training models to identify entities, such as using named entity recognition (NER) models; (3) Natural language processing (NLP): for text data, using NLP technology to identify entities.

[0039] The process of identifying relationships between entities using relation extraction (RE) technology. In some embodiments, relationships may include "access", "transfer", "control", etc. Relationship extraction can use the following methods: (1) Pattern matching: using predefined patterns or templates to identify relationships between entities; (2) Machine learning: training models to predict relationships between entities, such as using relation extraction models; (3) Semantic analysis: for text data, identifying relationships by analyzing sentence structure and semantics.

[0040] In general, in order to improve the recognition accuracy of entities and entity relationships, training can be performed on industry-specific corpora to obtain an industry-customized NER model.

[0041] Step S3: Construct timeline data, analyze and count the time series data in each time window, and calculate key indicators.

[0042] In the implementation described in the present invention, the extracted data is first sorted according to timestamps to form time series data, and then the time series data is divided into preset time windows, such as hours, days, weeks, etc., for subsequent statistical and aggregate analysis.

[0043] Time window partitioning refers to dividing continuous time series data into a series of non-overlapping or partially overlapping time periods, which are called time windows. The size and position of the time window can be determined according to the analysis objectives and data characteristics. In general, time windows can be divided by year, month, week, day, hour, minute, second, millisecond, or based on a sliding window, that is, the window slides on the time series, sliding one time unit each time, and the time unit is less than or equal to the current time window scale. In some embodiments, network security monitoring can use time windows of different lengths such as hours, days, and weeks to analyze network traffic, login attempts, transaction activities, etc., while business analysis may use time windows of days, weeks, and months to analyze sales data, user behavior, etc.

[0044] After dividing the time series data based on time windows, statistical and aggregate analysis is performed on the time series data. In each time window, real-time statistical calculations are performed on entity information and relationship data between entities, and materialized views are created to extract meaningful information, simplify data, and facilitate subsequent analysis and modeling.

[0045] In each time window, the entity information and the relationship data between entities are pre-statisticed, that is, in each time window, various statistical summary operations are performed based on the data of entities and their relationships, and then the data is aggregated and analyzed to calculate key indicator data. The content and method of statistics can be determined according to specific business needs. It can be statistics that calculate the number of citations or usages, or statistics based on higher-level indicators, such as the number of active users, transaction volume, communication frequency, and abnormal events.

[0046] The number of active users is the number of users who have performed at least one activity in a given time window. Transaction volume: the total amount of transactions that occurred in a given time window. Communication frequency: the number of communications between entities in a given time window. Abnormal event number: the number of abnormal or suspicious events detected in a given time window.

[0047] Furthermore, abnormal events usually include: Sudden change in behavior frequency: The number of activities of an entity suddenly increases or decreases abnormally within a certain time window. For example: the communication frequency of a certain IP address is abnormally high (which may be a sign of an attack). A user account initiates a large number of operations in a short period of time.

[0048] The relationship network of an entity changes abnormally in time sequence. For example, a user account is suddenly associated with multiple new devices. A node (IP) in the network traffic graph connects to a large number of new nodes in a short period of time.

[0049] Furthermore, the data is aggregated and analyzed to calculate key indicator data, such as the average, median, maximum, minimum, and sum of key indicators. Aggregation can be based on data in a single continuous time window or multiple non-continuous time windows. The key indicators can be the number of active users, transaction volume, etc., but the specific calculations are defined by the user, that is, the data that the user is concerned about is the key indicator. In general, statistical indicators are measures of quantitative data features and are used to describe the behavior of entities.

[0050] Finally, the time window, key indicator data calculated within the window, and other related data (including entity static attributes, dynamic attributes, entity categories, relationship attributes, relationship weights, relationship strength, etc.) are integrated into a data set and stored. The data can be stored in a table format, where each row represents a time window and the columns contain the start and end time of the time window, indicator values, etc. By integrating entity information, relationships between entities, time, and indicator data, through the above steps, a timeline data set organized by time and containing key statistical indicators can be constructed.

[0051] Step S4. Data storage In the embodiment described in the present invention, data storage adopts two data storage strategies, namely, using a non-relational database to store statistical data and timestamps within a time window, and using a graph database to store time series entity information and relationship data between entities and their attribute data that changes over time.

[0052] Design a non-relational database to accommodate the needs of time window statistics. This may include defining a document structure or key-value pairs to store statistics within a specific time window, with each time window's statistics associated with one or more timestamps to identify the creation time of the data and the start and end time of the window. Create appropriate indexes, especially indexes on timestamp fields, to increase query speed. Choose a data storage structure (e.g., columnar storage) to optimize the processing of time series data. The advantages of non-relational data are: (1) Non-relational databases can process large throughput data streams in real time and provide low-latency responses.

[0053] (2) Supports horizontal expansion and can seamlessly expand storage and processing capabilities as the amount of data grows.

[0054] (3) It allows the storage of various types of data without the need for a strict data table structure, thus simplifying data storage and query.

[0055] In some embodiments, one or more key-value pairs may be defined for each hourly, daily, or weekly window, including one or more statistical indicators within the time window.

[0056] When designing a graph database, you need to consider how entities, relationships, and time attributes are stored to ensure that the data structure can support efficient time series analysis. Graph databases typically include nodes (entities), edges (relationships), and attributes, where: Nodes: represent entities in the graph. Edges: represent relationships between entities. Attributes: can be static or time series attributes, the latter used to store data that changes over time. In some embodiments, nodes can represent users, devices, accounts, IP addresses, and edges can represent access, transfers, and controls. Attributes can be timestamps, transaction amounts, number of communications, and alarm names. In order to improve query performance, time attributes need to be indexed so that data at a specific point in time or time window can be quickly located. This may include creating indexes for time attributes and optimizing query strategies (for example, using time windows and range queries).

[0057] Time information can be expressed through explicit time nodes or timestamps embedded in attributes, and supports query and analysis of multi-granularity time windows, such as Figure 2 shown.

[0058] The present invention supports queries of multi-granularity time windows according to analysis requirements, among which fine-grained queries (micro level): second or minute level, suitable for scenarios with rapid response to events; coarse-grained queries (macro level): aggregated by day, week or month, suitable for long-term trend analysis.

[0059] When storing time-series entity relationship data, graph databases can, on the one hand, more intuitively represent and analyze complex relationships between entities. On the other hand, they can use the query language of graph databases (such as Cypher) to perform complex time series analysis to find all relationships involving specific entities or discover abnormal behavior patterns within a time window.

[0060] Step S5: Based on the user's query instructions, interact and analyze the time data, and then identify the change process and pattern of entities and relationships over time.

[0061] Data updates, access to real-time data, can be integrated using a stream processing framework, and real-time updates of non-relational data time series statistical structures and graph database time series relational structures. This ensures that the entities, relationships, and time attributes in the graph are synchronized with the latest statistical indicator data. The data interaction provided by the embodiment of the present invention can meet time operations, that is, the interaction and data operations of time series data in the database, that is, allowing users to perform various operations and indicator operations on the data, that is, involving the management and analysis of key indicators, and visualization operations, that is, providing intuitive data display and actions. In general embodiments, time operations include time range selection and time granularity adjustment. Data operations include querying, filtering, and aggregation. Indicator operations include baseline setting and threshold setting. Visualization operations include dragging the timeline, zooming and scrolling, dynamic refresh of views, and interactive charts.

[0062] The present invention provides a time filtering function, which enables users to query data based on a specific time range, and apply time series analysis algorithms, such as time series clustering, anomaly detection, trend prediction, etc., to identify the change patterns and processes of entity information and inter-entity relationships over time. For example, specific time series analysis techniques, combined with analysis scenarios such as baselines and forecasts, identify abnormal behaviors and events.

[0063] Step S6. The data generated by the interaction and analysis are used to display the changing patterns and processes of entities and relationships over time based on dynamic and static visualization tools.

[0064] In the embodiment described in the present invention, for the data generated by data interaction and analysis, the present invention provides a data rendering and visualization strategy, which can be rendered and visualized at different time window scales, namely: Macro data rendering: When the time window is greater than or equal to the preset window, a specific data rendering method is used to show the trend of the overall data in the time window. In some embodiments, it can be set that when the time window is greater than or equal to the weekly degree, the heat map method is used for rendering, and when it is less than the weekly degree but greater than the daily degree, the Gantt chart method is used for rendering, such as Figure 3 shown.

[0065] Micro data rendering: When the time window is smaller than the preset window, a specific data rendering method is used to display in detail the state of a specific entity or relationship in a specific time window or at a specific time point. In some embodiments, when the time window is smaller than the day, point-line rendering can be used.

[0066] Dynamic data rendering: Display data changes over time in an animated form, such as dynamic graphs and time series animations.

[0067] Static data rendering: a snapshot of data at a specific point in time, such as a static relationship graph.

[0068] The above description is only a preferred embodiment of the present invention and does not constitute any form of hindrance to the present invention. Any simple modification or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A time-series-based relationship graph analysis method, characterized in that: The method comprises the following steps: Step S1. Collecting raw data, the raw data includes entity information and relationship data between entities, and each piece of data includes a timestamp; Step S2. Extracting entity information, relationship data between entities and timestamp data in the original data; Step S3. Form time series data based on the extracted data, and divide the time series data into time windows; then perform statistical and aggregate analysis on the time series data in each time window, and calculate key indicators; Step S4. Design a graph database to store entity information and relationship data between entities and their attributes that change over time, and a non-relational database to store timestamps and key indicator data calculated within a time window; design a database query strategy and optimize indexes; Step S5. Based on the user's query instructions, interact and analyze the time series data to identify the change patterns and processes of entity information and relationships between entities over time; Step S6. The data generated by the interaction and analysis are used to display the changing patterns and processes of entity information and relationships between entities over time based on dynamic and static visualization tools.

2. The time-series-based relationship graph analysis method according to claim 1, characterized in that: Step S2 also includes preprocessing the extracted timestamp data, and the preprocessing includes: first identifying the timestamp field in the data, then using the corresponding parsing function or library to convert the timestamp into an operable date and time object, and finally performing unified time zone, format standardization and timestamp completion operations on the extracted timestamp.

3. The time-series-based relationship graph analysis method according to claim 1, characterized in that: In step S2, for the unstructured text in the extracted data, natural language processing technology or pattern recognition algorithm is used to extract entity information and entity relationship data; for the structured text in the data, the fields are directly mapped to entities, and then the entity information and the relationship between entities in the data are derived based on the semantic rules of the structured field naming and value content.

4. The time-series-based relationship graph analysis method according to claim 1, characterized in that: Step S3 specifically includes the following contents: Sort the data by timestamp to form time series data; Divide the time series data into set time windows, perform statistical and aggregate analysis on the time series data within the windows, and calculate key indicators based on the divided time windows; Integrate entity information and relationships between entities, time, time window, key indicators within the window, and other related data to form a timeline dataset.

5. The time-series-based relationship graph analysis method according to claim 1, characterized in that: The graph database has a time attribute index.

6. The time-series-based relationship graph analysis method according to claim 4, characterized in that: Aggregate analysis can be performed based on data in a single continuous time window or based on data in multiple non-continuous time windows.

7. The time-series-based relationship graph analysis method according to claim 1, characterized in that: In step S3, the aggregate analysis includes calculating the average, median, maximum, minimum and sum of the key indicators.

8. The time-series-based relationship graph analysis method according to claim 1, characterized in that: In step S5, the interaction and analysis of time series data includes: based on the user's query instructions, filtering the timeline data of a specific time range, applying a time series analysis algorithm to analyze the timeline data to identify the change patterns and processes of entity information and relationships between entities over time.

9. The method for analyzing a time series relationship graph according to claim 1, characterized in that: In step S5, the interaction and analysis of the time series data includes index operations and visualization operations on the data.

10. The time-series-based relationship graph analysis method according to claim 1, characterized in that: In step S6, dynamic and static visualization tools are used to display the change patterns and processes of entity information and relationships between entities over time, including: Macro data rendering: When the time window is greater than or equal to the preset window, a specific data rendering method is used to show the trend of the overall data within the time window; Micro data rendering: When the time window is smaller than the preset window, a specific data rendering method is used to display the status of specific entity information or the relationship between entities in a specific time window or a specific time point.

Images