An intelligent decision graph construction method based on dynamic timing event data

Abstract

The application is suitable for the technical field of data analysis, and provides a kind of intelligent decision graph construction method based on dynamic time series event data, comprising: reading and cleaning original event data, processing field, unfolding nested information through structured analysis and entity alignment technology, generating cleaned structured data;Calculate the frequency of different field combinations, generate frequency feature columns combined with time stamp and weight factor, and merge multi-dimensional relationships;Screen event pairs based on dynamic frequency threshold, construct directed graph data structure, use event unique identifier as node, establish graph edge and set edge weight, optimize the time series relationship between events through time series matching and dynamic weighting mechanism;Output graph atlas containing node number, edge number and graph edge data based on frequency screening.The application realizes a precise, efficient and intelligent graph construction method, promotes the technical progress in the field of dynamic event data analysis, and can be applied to practical scenarios that require processing dynamic event data and constructing associated graph atlas.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and in particular relates to a method for constructing an intelligent decision graph based on dynamic time-series event data. Background Technology

[0002] With the advent of the big data era, the amount of data in all sectors of society is growing exponentially. Especially in areas such as event analysis, public opinion monitoring, and social network analysis, the processing and mining of massive amounts of data has become an important research direction. Traditional data analysis methods focus on structured data, typically concentrating on a single data source or static data for analysis. However, in practical applications, data is often highly complex, containing not only rich relationships and spatiotemporal characteristics but also often spanning multiple sources and dimensions. This undoubtedly presents a significant challenge to existing technologies.

[0003] Currently, network analysis, text mining, and graph data processing technologies have been widely applied in fields such as public opinion analysis, social network analysis, and event prediction. Graph data processing, in particular, can effectively reveal potential connections and patterns between data items by constructing relational graphs of data. In a graph, nodes represent different data objects, and edges represent the relationships between them. This graph-based construction and analysis method provides a completely new perspective for data mining.

[0004] However, existing technologies still have certain limitations. Regarding graph construction, current research generally employs manual rules or Natural Language Processing (NLP)-based Named Entity Recognition (NER) and relation extraction techniques in the data processing stage to identify event subjects, event content, and their relationships. However, these methods have significant limitations: on the one hand, manual rules rely on the experience of domain experts and are difficult to adapt to different scenarios and complex data; on the other hand, NLP-based methods typically rely on large-scale labeled datasets, which are limited by the quality and coverage of training data, easily introducing noise and errors. Furthermore, due to the diversity of natural language expressions, the same event may be described differently in different texts, leading to difficulties in information matching and affecting the accuracy of temporal analysis. Based on this, the constructed graphs often have many problems. Over-reliance on these flawed methods often leads to the neglect of the temporal sequence and multi-level relationships of events during graph construction. In the graph data analysis stage, existing graph data analysis methods typically use static graphs as the basis for analysis, lacking consideration of dynamic event flows, resulting in potentially lagging or untimely analysis results in rapidly changing social environments. Even though some advanced technologies attempt to combine time series analysis with network graph construction to improve this situation, they still face significant challenges when dealing with large-scale heterogeneous data sources and multidimensional relationships, including extremely high computational complexity and resource consumption. Therefore, improving the automation and intelligence of graph construction while ensuring timeliness has become one of the current technological difficulties. In view of this, this invention proposes an intelligent decision graph construction method based on dynamic time-series event data. Summary of the Invention

[0005] The purpose of this invention is to provide a method for constructing an intelligent decision graph based on dynamic time-series event data, aiming to solve the problems mentioned in the background art.

[0006] The objective of this invention is achieved through the following technical solution: A method for constructing an intelligent decision graph based on dynamic time-series event data includes the following steps: Step S1: Data preprocessing; Read and clean the raw event data, format and normalize the time field, fill and format the text field, and expand the nested information through structured parsing and entity alignment technology to generate cleaned structured data. Step S2: Frequency calculation and merging of multidimensional relationships; Calculate the frequency of different field combinations, generate frequency feature columns by combining timestamps and weight factors, and merge multidimensional relationships to reflect the correlation strength between different field combinations; Step S3: Map construction; Based on the dynamic frequency threshold for filtering event pairs, a directed graph data structure is constructed, with the unique identifier of the event as the node, the edges of the graph are established and the edge weights are set, and the temporal relationship between events is optimized through temporal matching and dynamic weighting mechanisms. Step S4: Output and Application; The output includes the number of nodes, the number of edges, and graph edge data filtered by frequency, supporting visualization and analysis applications.

[0007] Furthermore, the data preprocessing steps specifically include: Read and clean: Read raw event data, clean null values, remove invalid data, and perform outlier detection operations; Time field processing: Natural language processing technology is used to format and transform the time field in the data, and event data that meets the set conditions is filtered out according to the set time range; Text field processing: Fill empty values ​​in text fields with "unknown", convert them to string format, process text fields and event descriptions using deep learning-based text parsing methods, and use the safe_literal_eval method to parse dictionary or list type data in the fields; Data expansion: Perform an "expand" operation on each column of data, split the field containing multiple values, and use a structured parsing strategy to break down nested information into independent data units. Then, use entity alignment technology to normalize the same entities in different data sources, and finally obtain cleaned structured data.

[0008] Furthermore, the frequency calculation and multidimensional relationship merging steps specifically include: Frequency calculation: Calculate the frequency of different field combinations in the data. During the calculation process, the event timestamp information is combined to introduce weight factors, calculate the associations at different levels, and generate a new frequency feature column for each combination. Multidimensional relationship merging: The generated frequency feature columns are merged to obtain frequency-enhanced data, which reflects the correlation strength between different field combinations.

[0009] Furthermore, the map construction steps specifically include: Filtering event pairs: Based on the frequency-enhanced data, set a dynamic frequency threshold and filter event pairs that meet the preset frequency threshold; Establishing a graph structure: In a directed graph data structure, the unique identifier of the event data is used as a node, and the edges of the graph are established based on the frequency between the filtered event pairs; the prerequisite for establishing an associated edge in the graph is that the frequency value of the event pair is greater than a preset threshold. Temporal matching and relationship optimization: Temporal matching of event relationships in the graph is performed, graph edges that conform to the temporal order are constructed based on event timestamps, and a dynamic weighting mechanism is used to dynamically assign weights to event relationships; Edge weight settings: The weight value of an edge is determined by frequency information, representing the strength of the relationship between events.

[0010] Furthermore, the graph learns the correlation patterns between events through graph neural networks and supports real-time updates of causal relationships.

[0011] A computer device for implementing the above-described intelligent decision graph construction method based on dynamic time-series event data, comprising: The central processing unit is used to perform data preprocessing and event analysis tasks; Graphics processors are used to accelerate the computation of graph neural networks during the event graph construction process; Memory serves as temporary data storage space; Storage units are used to store raw data, intermediate calculation results, and the final constructed event graph; The network communication module is used to enable communication between computer devices and remote servers or cloud computing resources, thereby achieving data sharing and distributed computing. Input devices are used to provide an operating interface; Display devices are used to visualize event graphs and present data analysis results; Data interface, used to connect external storage devices or sensors; Operating system and software environment, used to provide the runtime environment.

[0012] Furthermore, the central processing unit (CPU) reads the Excel data file from the storage unit, performs data preprocessing steps, and temporarily stores the cleaned data in memory; the CPU performs frequency calculation and multidimensional relationship merging steps, stores the calculation results in the storage unit, and displays the preliminary analysis results through a display device; the graphics processing unit (GPU) processes large-scale data, performs graph neural network calculations, and executes the graph construction step, during which the network communication module supports interaction between the computer device and the remote server; finally, the constructed event graph is stored in the storage unit, and the display device visualizes the graph results using Matplotlib and NetworkX.

[0013] Compared with the prior art, the beneficial effects of the present invention are: Enhanced temporal graph construction: Unlike traditional static data graph construction methods, this invention is based on temporal event data, which can effectively capture the temporal dependencies and dynamic changes between events, ensuring that the nodes and edges in the graph conform to the actual propagation order of events. This improves the accuracy and practicality of the graph, effectively overcomes the static nature and limitations of traditional methods, and provides a more accurate and intelligent data analysis framework.

[0014] Frequency analysis of multidimensional relationships: This invention improves the expressive power of the graph and the depth of data mining by calculating the frequency and dynamically merging multidimensional data relationships (such as the relationship between the main person and the event type, the relationship between the main person and the economic impact, etc.).

[0015] Automated and intelligent processing: This method reduces manual intervention by automatically filtering event pairs that meet frequency thresholds, thereby improving the efficiency and intelligence of data processing and meeting the needs of large-scale data processing.

[0016] In summary, this invention, by combining temporal analysis, multi-dimensional analysis, and automated processing, achieves a more accurate, efficient, and intelligent graph construction method, thus advancing technological progress in the field of dynamic event data analysis. It can be widely applied in various practical scenarios requiring the processing of dynamic event data and the construction of relational graphs, such as social network analysis, crisis prediction, and public opinion monitoring. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the computer device structure upon which the method of this invention depends.

[0018] Figure 2 This is a flowchart of the method of the present invention.

[0019] Figure 3 This is a schematic diagram of the data preprocessing steps.

[0020] Figure 4 This is a schematic diagram illustrating the steps of frequency calculation and merging multidimensional relationships.

[0021] Figure 5 This is a schematic diagram of the map construction steps.

[0022] Figure 6 This is a schematic diagram of the final constructed event graph.

[0023] Figure 7 This is a schematic diagram of the event chain mined by the method of the present invention.

[0024] In the diagram: Central Processing Unit 1001, Graphics Processing Unit 1002, Memory 1003, Storage Unit 1004, Network Communication Module 1005, Input Device 1006, Display Device 1007, Data Interface 1008, Operating System and Software Environment 1009. Detailed Implementation

[0025] In order to provide a clearer understanding of the technical features, objectives and beneficial effects of the present invention, the technical solution of the present invention will now be described in detail below, but it should not be construed as limiting the scope of implementation of the present invention.

[0026] To address the problems of existing technologies, this invention incorporates innovative designs in several aspects. First, it employs automated data cleaning and formatting methods to reduce human intervention, and combines this with domain-adaptive parsing techniques to improve data usability and consistency. In data cleaning, unstructured data is standardized to ensure data integrity and consistency. Regarding data formatting, the standardization of event-related fields enhances data operability, making it suitable for subsequent event frequency calculations and time-series analysis. Furthermore, compared to traditional NLP-based methods, this invention utilizes a multi-level data parsing strategy, combined with keyword matching, pattern recognition, and statistical analysis techniques, to achieve efficient processing of event data. For example, during event parsing, the system automatically identifies entities in the text and matches them to a standard event library to ensure the uniqueness and traceability of events. Simultaneously, a frequency-based dynamic data filtering mechanism filters out low-frequency events and noisy data, improving data reliability. Through these optimizations, this invention significantly improves efficiency and accuracy compared to traditional methods, laying a solid foundation for subsequent time-series analysis and providing data support for more complex event reasoning and prediction.

[0027] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0028] One embodiment of the present invention provides a method for constructing an intelligent decision graph based on dynamic time-series event data, the flowchart of which is shown below. Figure 2 As shown, the method includes the following steps: Step S1: Data preprocessing; specific steps are as follows (see...) Figure 3 ): Read and clean: Read the raw event data, clean up null values, remove invalid data, and perform outlier detection to filter out potentially erroneous data.

[0029] Time field processing: Natural Language Processing (NLP) techniques are used to format and transform the time field in the data, accurately parsing and normalizing the time information. This includes conversions between relative times (e.g., "yesterday," "last month") and absolute times (e.g., "2022-06-15"), ensuring a consistent format for time information within the data. Then, based on a defined time range, event data within that range is filtered out.

[0030] Text field processing: Due to the diversity of data sources, there may be missing fields or format mismatches. Therefore, we first fill in empty values ​​for text fields (such as "Main Person" and "Event Type") by replacing them with "Unknown," and then uniformly convert them to string format. During this process, we use methods such as synonym replacement to process text fields and event descriptions to reduce information loss caused by different expressions (such as synonyms and abbreviations). Simultaneously, we use the `safe_literal_eval` method to parse dictionary or list type data in the fields.

[0031] Data expansion: Perform an "expand" operation on each column of the data to break down fields containing multiple values. For complex nested information that may be contained in the event description, such as an event being associated with multiple related personnel or multiple influencing factors, a structured parsing strategy is used to break down the nested information into independent data units. Then, through entity alignment technology, the same entities in different data sources are normalized, and finally, cleaned structured data is obtained.

[0032] The data preprocessing steps of this invention integrate multiple advanced technologies such as automatic parsing, format conversion, and data cleaning, overcoming many drawbacks of traditional rule-based methods (such as regular expression matching, stop word removal, and keyword extraction). While traditional methods can be effective in processing structured data, they are easily affected by grammatical variations, spelling errors, and ambiguities when dealing with unstructured text data. Furthermore, when processing multilingual data or specialized terminology, they rely excessively on manual annotation and dictionary matching, resulting in poor scalability. In contrast, this method is highly automated, significantly reducing reliance on manual annotation and improving data processing efficiency. It is applicable not only to unstructured text such as social media data and news reports but also to the analysis of structured data such as government reports and research papers. Due to its strong generalization ability, it plays a crucial role in time-series event analysis tasks across various fields, including economic dynamics analysis, international political event tracking, and corporate crisis management.

[0033] Step S2: Frequency calculation and multidimensional relationship merging; specific steps are as follows (see...). Figure 4 ): Frequency Calculation: Based on cleaned structured data, the frequency of different field combinations (such as the combination of "key figure" and "event type," "key figure" and "economic impact," and "event type" and "economic impact") is calculated. During the calculation, event timestamp information is incorporated, and weighting factors are introduced to calculate the association at different levels, such as "key figure-event type" and "event type-economic impact." A new frequency feature column is generated for each combination to measure the event's influence across different dimensions. For example, calculating the frequency of events involving a specific key figure measures the figure's activity level and influence on a particular type of event; calculating the co-occurrence frequency of an event with different economic influencing factors detects the event's potential direct or indirect effects on the economic system.

[0034] Multidimensional relationship merging: The generated frequency feature columns are merged to obtain frequency-enhanced data, reflecting the correlation strength between different field combinations. For example, if A is the main person and B is the event type, the merged result is frequency 1; if C is the main person and D is the economic impact, the merged result is frequency 2; frequency 1 and frequency 2 together constitute the frequency.

[0035] Compared to traditional frequency analysis methods that rely on keyword statistics and TF-IDF calculations, the multidimensional frequency calculation method proposed in this invention offers significant advantages. Traditional methods often have considerable limitations when processing unstructured data, especially when the data volume is large and contains significant noise. Statistical results are easily affected by extreme values ​​and data imbalances, leading to large biases in the analysis results. Furthermore, TF-IDF calculations are primarily based on word frequency and lack the ability to model the temporal relationships between events, limiting their effectiveness in time-sensitive event analysis tasks. This method, however, can more comprehensively reveal the complex relationships within an event system, accurately capture key event relationships, adapt to different data scales and application scenarios, and has high versatility. Simultaneously, frequency analysis combined with time information can further reveal the evolutionary trends of events, providing support for temporal relationship modeling. For example, in policy analysis and public event tracing, it can help identify high-impact events, providing data support for decision-making, and has great application potential in social science research, financial analysis, and public opinion monitoring.

[0036] Step S3: Map construction; specific steps are as follows (see...) Figure 5 ): Event pair filtering: Based on the frequency-enhanced data, a dynamic frequency threshold is set, and data filtering criteria are automatically determined based on the overall distribution of events. Low-frequency and invalid associations are automatically filtered out, statistically significant outliers are removed, and the accuracy and reliability of the data are improved. Based on this, event pairs that meet the preset frequency threshold are selected.

[0037] Establishing a graph structure: In a directed graph data structure, unique identifiers of event data are used as nodes. Edges are created based on the frequency of filtered event pairs to represent the relationships between events. Only event pairs with a frequency value greater than a preset threshold will have an edge established in the graph.

[0038] Temporal matching: Event relationships in the graph are matched temporally, constructing graph edges that conform to the chronological order based on event timestamps to ensure that the relationships between events are chronologically consistent and avoid temporal discrepancies. Simultaneously, a dynamic weighting mechanism is employed, dynamically assigning weights to event relationships based on multiple dimensions such as time interval, event influence, and historical co-occurrence patterns, to more accurately measure the strength of association between events. Specifically, first, the time interval between any two events is calculated; the closer the times, the higher the weight, decreasing as the time difference increases. Second, the average influence of the event pair is calculated based on the importance indicators of the events themselves (such as attention level, rating, etc.); the greater the influence, the higher the weight. Next, the frequency of the event pair appearing together or adjacently in historical data is statistically analyzed; the higher the frequency, the stronger the correlation trend, and the higher the weight is assigned. In the synthesis phase, these three types of indicators are standardized, and the weight coefficients of the three are dynamically adjusted according to the current data distribution characteristics to ensure that the weight allocation remains sensitive and adaptable to data characteristics.

[0039] Edge weight settings: The weight value of an edge is determined by frequency information, representing the strength of the relationship between events.

[0040] In past research, the construction of event graphs has largely relied on text-based event relation extraction, such as dependency parsing and knowledge triple construction. These methods typically extract entities and their relationships from text data based on pre-trained models or rule templates, and then construct knowledge graphs through relation links. However, they have significant limitations in handling temporal information. On the one hand, traditional event relation extraction methods lack accurate modeling of time factors. Event time information in natural language text is diverse in format, such as absolute time ("March 15, 2022"), relative time ("last month"), and time span ("the past two years"), making it difficult for knowledge graphs extracted directly from text to form accurate temporal relationships. Furthermore, existing methods often focus on static relationships, ignoring the time constraints in the event's development, failing to accurately characterize causal relationships, and limiting the practical application value of the graph. On the other hand, traditional methods for constructing event graphs rely on pre-defined relation templates or manually labeled datasets. While this may be effective in specific domains, when extended to cross-domain analysis scenarios, the predefined event relation types are insufficient to cover all possible associations, resulting in poor performance of knowledge graphs in complex event analysis. Moreover, manual annotation is costly and the training dataset is limited in size. Machine learning-based event relationship extraction methods are difficult to guarantee high accuracy when dealing with massive amounts of unstructured text data.

[0041] Given the numerous drawbacks of traditional methods, this approach proposes a novel construction strategy. First, core events and their relationships are identified through event frequency calculation, and a temporal relationship network is constructed based on time information. Statistical patterns from historical data ensure the reliability of relationships between events. By introducing time attributes, causal sequences are established in the event graph, enabling connections between events to not only reflect "correlation" but also express "sequence" and potential causal relationships. This temporal processing effectively enhances the application value of knowledge graphs in predictive analytics, such as predicting the chain reactions or long-term impacts of certain events. During graph construction, a data filtering mechanism is crucial. Actual event data contains significant noise, such as abnormal data caused by input errors, information redundancy, and duplicate records. Low-frequency events, if not filtered, can interfere with analysis results and obscure key event relationships. Automatically filtering data by setting dynamic frequency thresholds provides high-quality data input for graph construction. Furthermore, edge weights are determined by frequency information, intuitively reflecting the strength of relationships between events and providing a quantitative basis for analyzing event associations. Simultaneously, a dynamic weighting mechanism further optimizes the modeling of temporal relationships between events. In traditional event graphs, edge weights are often the same or based on simple metrics. This method, however, dynamically assigns weights to event relationships based on multiple dimensions. If event A frequently triggers event B under specific conditions and is observed multiple times, the association weight will increase, helping the model to more accurately identify high-confidence causal relationships during prediction.

[0042] Step S4: Output and Application; Output the final constructed event graph (see Figure 6 The event graph contains the number of nodes, edges, and frequency-filtered edge data, which can support subsequent graph analysis, event prediction, or network analysis applications. Using network structure analysis and visualization techniques, the networkx library is used to visually display the connections between event nodes. By highlighting key events with different colors and node sizes, the constructed time-series event graph becomes more interpretable. Visually displaying node connections and highlighting key events helps analysts quickly grasp the complex relationships between events, assisting decision-makers in formulating targeted strategies. Compared to traditional event extraction methods, the time-series event graph constructed in this study can more accurately reflect the development of events and has broad application prospects in policy research, public opinion analysis, and economic forecasting.

[0043] Another embodiment of the present invention provides a computer device (see Figure 1 ), used to implement the intelligent decision graph construction method based on dynamic time-series event data described above, including: Central Processing Unit (CPU) 1001: Primarily performs data preprocessing and event analysis tasks, including data cleaning, text parsing, outlier detection, and time format conversion.

[0044] Graphics Processing Unit (GPU) 1002: Used to accelerate large-scale data processing, especially to accelerate network operations during event graph construction, such as graph neural network (GNN) computation.

[0045] RAM 1003: Serves as temporary data storage space, supporting batch data loading and calculation to improve processing efficiency.

[0046] Storage unit (SSD / HDD) 1004: Used to store raw data (such as Excel files), intermediate calculation results, and the final constructed event graph.

[0047] Network communication module 1005: Enables communication between computer devices and remote servers or cloud computing resources, realizing data sharing and distributed computing.

[0048] Input device (keyboard / mouse) 1006: Provides an operation interface for users, allowing them to set parameters, query analysis results, or adjust calculation tasks.

[0049] Display device (monitor) 1007: Visualizes the event graph and presents the data analysis results.

[0050] Data interface (USB / PCIe) 1008: Used to connect external storage devices or sensors to obtain more data sources.

[0051] Operating System and Software Environment 1009: Provides a runtime environment that supports the operation of software libraries such as Python, Pandas, and NetworkX to achieve data processing, frequency calculation, and map construction.

[0052] The central processing unit 1001 reads the Excel data file from the storage unit 1004, performs data preprocessing, and temporarily stores the cleaned data in memory 1003 to avoid repeated hard drive reads and improve processing efficiency. The central processing unit 1001 performs frequency calculation and multidimensional relationship merging steps, stores the calculation results in the storage unit 1004, and displays the preliminary analysis results through the display device 1007. The graphics processing unit 1002 processes large-scale data, performs graph neural network (GNN) calculations, and executes efficient graph construction steps. During this process, the network communication module 1005 plays a role, supporting interaction between the computer device and a remote server to obtain larger datasets or perform distributed computing as needed. Finally, the constructed event graph is stored in the storage unit 1004, supporting subsequent query analysis. The display device 1007 visualizes the graph results using Matplotlib and NetworkX to help users understand data relationships.

[0053] like Figure 7 As shown, the method of this invention successfully uncovers hidden high-frequency, long event chains. In the diagram, each node's label represents a specific event ID. By combining statistical and graph network analysis, not only can frequently occurring event sequences be identified, but the robustness of the connections between these events is also enhanced.

[0054] In summary, the temporal event graph constructed by this invention has many significant advantages and can effectively overcome the limitations of traditional static knowledge graphs: Through multi-dimensional frequency calculation in step S2, combined with timestamps and weighting factors, and the directed graph structure constructed in step S3 (where edge weights are determined by frequency and optimized through time-series matching), the graph not only clearly presents the causal path of events but also allows for trend prediction based on historical data, providing support for practical applications such as policy decision-making and economic analysis. Step S1's processing of the time field lays the foundation for time feature modeling, while the time-series matching operation in step S3 automatically identifies the time-series dependencies between events, thereby constructing an event chain based on time progression and revealing the logical evolution trajectory of events. Compared to traditional static knowledge graphs that only focus on entity relationships while ignoring the time dimension, this time-series event graph incorporates time factors. Through time feature modeling and time-series matching, it more accurately conveys event information, especially regarding time sequence and causal relationships. Simultaneously, it combines historical data for trend prediction, significantly enhancing trend analysis capabilities and providing a new perspective for time-series event research. The dynamic weighting mechanism adopted in step S3 dynamically allocates weights based on multi-dimensional factors such as time interval, event influence, and historical co-occurrence patterns, covering event frequency and correlation strength, thus more accurately reflecting the true connections between events. Because the construction process considers time factors and event correlations, when a new event occurs, based on timestamps and existing event relationship networks, combined with a dynamic weighting mechanism, the system can automatically adjust the event chain and update causal relationships, ensuring the graph remains real-time and effective. Due to its dynamic adjustment characteristics, the graph can promptly reflect changes in the social environment and adapt to rapidly changing situations, greatly enhancing its value in practical applications. Furthermore, by combining advanced machine learning methods such as graph neural networks (GNNs), the graph model proposed in this invention can automatically learn complex correlation patterns between events, uncover potential hidden relationships from the data, and improve prediction accuracy.

[0055] In summary, time-series event graphs provide an efficient, accurate, and interpretable solution for social event analysis, policy evaluation, and risk prediction. By integrating time factors, this method not only enhances the applicability of knowledge graphs in dynamic environments but also strengthens the logical reasoning ability between events, providing strong support for research and applications in multiple fields such as social sciences, economic research, and artificial intelligence. In the future, with the broadening of data acquisition channels and the continuous advancement of algorithm optimization, the application scenarios of time-series event graphs will become more extensive, playing a greater role in intelligent decision-making, risk warning, and public governance. Combining the above advantages, time-series event graphs can display causal paths, predict trends, combine external data sources, and dynamically adjust, providing efficient and accurate analysis for social event analysis, policy evaluation, and risk prediction. Furthermore, the construction process and relationship establishment are based on explicit steps, making it interpretable. The multi-step construction of the graph considers time factors, making it better adaptable to dynamic environmental changes. By establishing temporal dependencies and causal relationships, it enhances logical reasoning ability, providing strong support for research and applications in social sciences, economic research, and artificial intelligence. In the future, with the expansion of data acquisition channels and the continuous advancement of algorithm optimization, the application scenarios of time-series event graphs will become more extensive, and they will play a greater role in intelligent decision-making, risk warning, and public governance.

[0056] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A method for constructing an intelligent decision graph based on dynamic time-series event data, characterized in that, Includes the following steps: Step S1: Data preprocessing, specifically including: Read and clean: Read raw event data, clean null values, remove invalid data, and perform outlier detection operations; Time field processing: Natural language processing technology is used to format and transform the time field in the data, and event data that meets the set conditions is filtered out according to the set time range; Text field processing: Fill empty values ​​in text fields with "unknown", convert them to string format, process text fields and event descriptions using deep learning-based text parsing methods, and use the safe_literal_eval method to parse dictionary or list type data in the fields; Data expansion: Perform an "expand" operation on each column of data, split the field containing multiple values, and use a structured parsing strategy to decompose nested information into independent data units. Use entity alignment technology to normalize the same entities in different data sources, and finally obtain cleaned structured data. Step S2: Frequency calculation and multidimensional relationship merging, including: Frequency calculation: Calculate the frequency of combinations of different text fields in the data. The text fields correspond to event information. During the calculation process, the event timestamp information is combined to introduce a weight factor. The association at the level of different text field combinations is calculated separately, and a new frequency feature column is generated for each combination. Multidimensional relationship merging: The generated frequency feature columns are merged to obtain frequency-enhanced data, which reflects the correlation strength between different field combinations; Step S3: Map construction; Based on the dynamic frequency threshold for filtering event pairs, a directed graph data structure is constructed, with the unique identifier of the event as the node, the edges of the graph are established and the edge weights are set, and the temporal relationship between events is optimized through temporal matching and dynamic weighting mechanisms. Step S4: Output and Application; The output includes the number of nodes, the number of edges, and graph edge data filtered by frequency, supporting visualization and analysis applications.

2. The intelligent decision graph construction method based on dynamic time-series event data according to claim 1, characterized in that, The map construction steps specifically include: Filtering event pairs: Based on the frequency-enhanced data, set a dynamic frequency threshold and filter event pairs that meet the preset frequency threshold; Establishing a graph structure: In a directed graph data structure, the unique identifier of the event data is used as a node, and the edges of the graph are established based on the frequency between the filtered event pairs; the prerequisite for establishing an associated edge in the graph is that the frequency value of the event pair is greater than a preset threshold. Temporal matching and relationship optimization: Temporal matching of event relationships in the graph is performed, graph edges that conform to the temporal order are constructed based on event timestamps, and a dynamic weighting mechanism is used to dynamically assign weights to event relationships; Edge weight settings: The weight value of an edge is determined by frequency information, representing the strength of the relationship between events.

3. The intelligent decision graph construction method based on dynamic time-series event data according to claim 1, characterized in that, The graph learns the correlation patterns between events through graph neural networks and supports real-time updates of causal relationships.

4. A computer device for implementing the intelligent decision graph construction method based on dynamic time-series event data as described in any one of claims 1-3, characterized in that, include: The central processing unit is used to perform data preprocessing and event analysis tasks; Graphics processors are used to accelerate the computation of graph neural networks during the event graph construction process; Memory serves as temporary data storage space; Storage units are used to store raw data, intermediate calculation results, and the final constructed event graph; The network communication module is used to enable communication between computer devices and remote servers or cloud computing resources, thereby achieving data sharing and distributed computing. Input devices are used to provide an operating interface; Display devices are used to visualize event graphs and present data analysis results; Data interface, used to connect external storage devices or sensors; Operating system and software environment, used to provide the runtime environment.

5. The computer device according to claim 4, characterized in that, The central processing unit reads the Excel data file from the storage unit, performs data preprocessing steps, and temporarily stores the cleaned data in memory; The central processing unit performs frequency calculation and multidimensional relationship merging steps, stores the calculation results in the storage unit, and displays the preliminary analysis results through the display device; the graphics processing unit processes large-scale data, performs graph neural network calculations, and executes the graph construction step, during which the network communication module supports interaction between the computer device and the remote server; finally, the constructed event graph is stored in the storage unit, and the display device visualizes the graph results through Matplotlib and NetworkX.

Images