Event inference visualization method, device and equipment based on pre-aggregation engine

Abstract

This application relates to an event deduction visualization method, apparatus, and device based on a pre-aggregation engine, belonging to the field of data processing technology. The method includes: efficiently processing massive amounts of data through a pre-aggregation engine, including intelligent preprocessing, basic aggregation, and intelligent pre-aggregation; and combining this with an event deduction model based on an event processing engine for dynamic event deduction prediction and visualization. The deep integration of the pre-aggregation engine and the event deduction model enables dynamic adjustment of the aggregation strategy based on real-time needs and feedback from the deduction process, resulting in real-time optimized aggregated data, which is then fed back for event deduction optimization, forming a two-way feedback architecture of "deduction needs → aggregation strategy → data output → deduction optimization". This method can adapt to the processing needs of massive real-time data, improving the prediction accuracy and dynamic visualization capabilities of event deduction.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to an event deduction visualization method, apparatus and device based on a pre-aggregation engine. Background Technology

[0002] With the rapid development of technologies such as the Internet of Things and 5G, event data is experiencing explosive growth, which places higher demands on event simulation technology.

[0003] However, current conventional event simulation technologies face multiple challenges. At the data processing level, traditional manual data entry and static modeling methods are insufficient to handle the demands of massive real-time data, and the lag in system architecture limits overall performance improvement. While researchers have attempted to improve the prediction accuracy of event simulations by constructing spatiotemporal data matrices and introducing deep learning models such as convolutional neural networks and residual networks, these methods are often computationally inefficient when dealing with large-scale, complex data, failing to meet the real-time requirements of practical applications. In terms of visualization, existing technologies also have significant shortcomings. Most event simulation systems can only provide static results displays, lacking dynamic visualization support for the event simulation process. Although graph-based emergency event simulation methods can construct scenario evolution paths, their real-time data processing and interactive visualization capabilities still need improvement. This lack of visualization directly affects decision-makers' accurate understanding and rapid response to event developments.

[0004] In summary, the shortcomings of existing event simulation technologies in terms of data processing efficiency, prediction accuracy, and visualization effectiveness have become key bottlenecks restricting the development of this field. Therefore, there is an urgent need to develop a new event simulation method that can deeply integrate big data processing technologies and possess high-efficiency computing power, accurate prediction performance, and intuitive visualization effects. Summary of the Invention

[0005] Therefore, it is necessary to provide an event deduction visualization method, apparatus, and device based on a pre-aggregation engine to address the aforementioned technical problems.

[0006] An event inference visualization method based on a pre-aggregation engine, the method comprising:

[0007] Construct a pre-aggregation engine that includes an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and integrate the pre-aggregation engine into a pre-built event inference model based on an event processing engine;

[0008] The event processing engine analyzes in real time whether there is aggregated data in the pre-aggregation engine that meets the data aggregation attribute requirements of the event object. If so, the aggregated data is called directly. If not, the intelligent pre-aggregation module analyzes the data aggregation attribute requirements of the event object and controls the intelligent pre-processing module to collect data related to the event object from multi-source heterogeneous data for preprocessing. It also controls the basic aggregation module to aggregate the preprocessed data, obtain aggregated data, and cache it for the event processing engine to call.

[0009] The event processing engine calls aggregated data to drive event inference and drives the event inference model to visualize the event inference. At the same time, it provides real-time feedback on the inference process and the inference effect to the pre-aggregation engine. Based on the inference feedback, the intelligent pre-aggregation module dynamically adjusts the aggregation strategy and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy, so as to obtain real-time optimized aggregated data and feed it back to the event processing engine for event inference optimization.

[0010] An event deduction visualization device based on a pre-aggregation engine, the device comprising:

[0011] The architecture building module is used to construct a pre-aggregation engine, which includes an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and to connect the pre-aggregation engine to a pre-built event inference model based on an event processing engine.

[0012] The data aggregation module is used to analyze in real time whether the pre-aggregation engine has cached aggregated data that meets the data aggregation attribute requirements of the event object. If so, it directly calls the aggregated data; otherwise, it analyzes the data aggregation attribute requirements of the event object according to the intelligent pre-aggregation module, controls the intelligent preprocessing module to collect data related to the event object from multi-source heterogeneous data for preprocessing according to the data aggregation attribute requirements, and controls the basic aggregation module to aggregate the preprocessed data to obtain aggregated data and cache it for the event processing engine to call.

[0013] The bidirectional feedback optimization module is used by the event processing engine to call aggregated data to drive event inference and drive the event inference model to visualize the event inference. At the same time, it provides real-time feedback on the real-time requirements and inference results of the inference process to the pre-aggregation engine. Based on the inference feedback, it drives the intelligent pre-aggregation module to dynamically adjust the aggregation strategy and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy, so as to obtain real-time optimized aggregated data and feed it back to the event processing engine for event inference optimization.

[0014] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program performing the following steps:

[0015] Construct a pre-aggregation engine that includes an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and integrate the pre-aggregation engine into a pre-built event inference model based on an event processing engine;

[0016] The event processing engine analyzes in real time whether there is aggregated data in the pre-aggregation engine that meets the data aggregation attribute requirements of the event object. If so, the aggregated data is called directly. If not, the intelligent pre-aggregation module analyzes the data aggregation attribute requirements of the event object and controls the intelligent pre-processing module to collect data related to the event object from multi-source heterogeneous data for preprocessing. It also controls the basic aggregation module to aggregate the preprocessed data, obtain aggregated data, and cache it for the event processing engine to call.

[0017] The event processing engine calls aggregated data to drive event inference and drives the event inference model to visualize the event inference. At the same time, it provides real-time feedback on the inference process and the inference effect to the pre-aggregation engine. Based on the inference feedback, the intelligent pre-aggregation module dynamically adjusts the aggregation strategy and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy, so as to obtain real-time optimized aggregated data and feed it back to the event processing engine for event inference optimization.

[0018] The aforementioned event simulation visualization method, apparatus, and equipment based on a pre-aggregation engine efficiently process massive amounts of data through the pre-aggregation engine, including intelligent preprocessing, basic aggregation, and intelligent pre-aggregation. Combined with an event simulation model based on an event processing engine, it performs dynamic event simulation prediction and visualization. The pre-aggregation engine is not an independent front-end module but is deeply integrated with the event simulation model. It can dynamically adjust the aggregation strategy based on real-time needs and simulation effect feedback, obtaining real-time optimized aggregated data and feeding it back to optimize event simulation. This forms a two-way feedback architecture of "simulation needs → aggregation strategy → data output → simulation optimization." This not only accurately adapts to the processing needs of massive real-time data, significantly improving the data processing efficiency and reducing response latency in event simulation, but also strengthens the data's support accuracy for simulation prediction through dynamic linkage optimization between the aggregation strategy and event simulation. It reduces interference from invalid data, making the simulation prediction results more closely match actual scene changes, thus improving the prediction accuracy and dynamic visualization capabilities of event simulation. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an event inference visualization method based on a pre-aggregation engine in one embodiment;

[0020] Figure 2 This is a schematic diagram illustrating the business interface design between the pre-aggregation engine and the event inference model in one embodiment.

[0021] Figure 3This is a schematic diagram illustrating the event progression in one embodiment;

[0022] Figure 4 This is a schematic diagram of the structure of an event deduction model in one embodiment;

[0023] Figure 5 This is a schematic diagram of the event processing engine in one embodiment;

[0024] Figure 6 This is a schematic diagram of a dynamic visualization process in one embodiment;

[0025] Figure 7 This is a schematic diagram illustrating the implementation mechanism of the move event object in one embodiment;

[0026] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0028] In one embodiment, such as Figure 1 As shown, an event inference visualization method based on a pre-aggregation engine is provided, including the following steps:

[0029] Step 1: Construct a pre-aggregation engine that includes an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and then connect the pre-aggregation engine to a pre-built event inference model based on an event processing engine.

[0030] Among them, the business interface design of the pre-aggregation engine and the event inference model is as follows: Figure 2As shown in the diagram. The dynamic data access interface enables real-time access and standardized conversion of multi-source heterogeneous data. This interface supports multiple protocols (HTTP / HTTPS, etc.) and has automatic data format recognition and metadata extraction capabilities, serving as the data input channel for the entire pre-aggregation engine. The feature dimension interaction interface establishes a bidirectional feature information exchange channel between the intelligent preprocessing module and the event inference model. This interface supports the transmission of key feature dimensions of event objects, enabling the intelligent preprocessing module to dynamically adjust cleaning rules and feature extraction logic according to inference requirements, achieving inference-oriented data preparation. The pre-computation scheduling interface sends task scheduling instructions from the basic aggregation module to the intelligent preprocessing module / intelligent pre-aggregation module, specifying the source, type, priority, and expected output format of the data to be processed, ensuring high coordination between the preprocessing process and subsequent aggregation tasks. The business direct connection interface establishes a direct communication channel between the event object and the intelligent pre-aggregation module, enabling rapid acquisition of real-time status change information of the event object, avoiding intermediate delays, and ensuring the real-time nature of the inference response. The dynamic parameter neural interface transmits the dynamic aggregation strategy parameters generated by the intelligent pre-aggregation module to the event inference model in a neural network-parseable form, achieving deep coupling between the aggregation strategy and the inference logic. The strategy optimization interface serves as a strategy feedback channel between the basic aggregation module and the event inference model. Its main functions include: receiving four-fold inference feedback from the event inference model (historical call feedback, real-time parameter feedback, cost feedback, and effect feedback); converting feedback information into strategy adjustment instructions, driving the intelligent pre-aggregation module to re-evaluate the aggregation strategy, outputting new aggregation strategy suggestions, indirectly influencing the behavior of the basic aggregation module through the "predicted computation scheduling interface," and supporting reinforcement learning-based iterative strategy optimization, forming a closed-loop self-evolution mechanism. The scheduling logic native embedding interface enables native integration between the event inference model and the event processing engine. This interface directly embeds the event processing engine's scheduling logic into the event inference model, giving the inference process strong real-time performance and high controllability, avoiding the interface call overhead and latency of traditional architectures. The results evaluation interface is used by the event processing engine to evaluate the quality of the simulation results, outputting key indicators such as accuracy, error rate, and response latency, and sending the evaluation results back to the simulation feedback as an important basis for strategy optimization.

[0031] It is important to note that although the event processing engine and the event deduction model are closely related in function, they have fundamental differences in architectural positioning and division of responsibilities: (1) In terms of core functions, the event processing engine tends to perform real-time event detection, filtering, correlation, and rule matching, while the event deduction model tends to predict future trends and simulate scenarios based on historical and current data; (2) In terms of pattern processing, the event processing engine tends to process streaming data, while the event deduction model tends to perform a hybrid processing of batch and streaming data; (3) In terms of time characteristics, the event processing engine tends to focus on current and recent events, while the event deduction model tends to focus on the past, present, and future; (4) In terms of output format, the event processing engine tends to focus on current and recent events, while the event deduction model tends to focus on the past, present, and future. Therefore, in actual system deployment, the event processing engine is usually used as the underlying infrastructure to provide event perception and scheduling capabilities; while the event deduction model is built on top of it, using the event stream and service interface it provides for modeling and deduction. The two are deeply integrated through the "scheduling logic natively embedded interface", but remain logically independent.

[0032] Depend on Figure 2 As can be seen, the output of the pre-aggregation engine is not stored independently, but is directly converted into scheduling logic instructions (such as memory-level data pointers and calculation operator parameters) of the event processing engine. This allows the event processing engine to call the engine without secondary parsing, reducing response latency to the millisecond level. For example, in real-time risk control scenarios, the intelligent pre-aggregation module can directly embed "the user's transaction aggregation results in the past hour" into the decision tree node of the risk control engine, skipping the traditional data reading-parsing process.

[0033] Step 2: The event processing engine analyzes in real time whether the pre-aggregation engine has cached aggregated data that meets the data aggregation attribute requirements of the event object. If so, the aggregated data is called directly. If not, the intelligent pre-aggregation module analyzes the data aggregation attribute requirements of the event object, controls the intelligent preprocessing module to collect data related to the event object from multi-source heterogeneous data for preprocessing, and controls the basic aggregation module to aggregate the preprocessed data to obtain aggregated data and cache it for the event processing engine to call.

[0034] Step 3: The event processing engine calls the aggregated data to drive event inference and drives the event inference model to visualize the event inference. At the same time, it provides real-time feedback on the real-time requirements and inference effects of the inference process to the pre-aggregation engine. Based on the inference feedback, the intelligent pre-aggregation module dynamically adjusts the aggregation strategy and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy, so as to obtain real-time optimized aggregated data and feed it back to the event processing engine for event inference optimization.

[0035] In one embodiment, the intelligent preprocessing module is controlled to collect data related to the event object from multi-source heterogeneous data for preprocessing according to data aggregation attribute requirements, including:

[0036] First, adaptive data cleaning is performed on the collected data, including data deduplication, missing value imputation, and outlier handling. Data deduplication employs a combined approach of "distributed fingerprinting + self-supervised learning"—first, Hive / Spark tools are used to generate data fingerprints for initial deduplication, then a self-supervised model is used to identify implicit repetition patterns (such as synonymous heterogeneity in text data or coordinate offset repetition in geographic data), reducing the false negative rate by 15%-30% compared to traditional methods. Missing value imputation involves dynamic imputation decisions based on missing value patterns. For random distributed missing values, improved KNN (K-nearest neighbor) interpolation with local data density weights is used for imputation; for systematic missing values ​​(such as sensor malfunctions), missing value attribution analysis is triggered, and imputation values ​​with confidence are generated based on the domain rules of the event object (rather than simple deletion or mean imputation). Specifically, the local data density weights can be dynamically calculated in Python using the Pandas extension library. Outlier handling includes: integrating statistical methods with business boundaries—in addition to statistical methods such as Z-score (standard score) and IQR (interquartile range), a "business rule verification layer" is introduced. For example, for financial transaction data, outliers not only need to meet the statistical deviation, but also need to be verified a second time through business rules such as transaction frequency and amount thresholds to avoid mistakenly deleting legitimate but rare business data.

[0037] Secondly, the cleaned data undergoes stratified standardization, including Min-Max or Z-score standardization for structured data to ensure numerical comparability; and dynamic window alignment for time series data, with the window size adaptively adjusted according to the frequency of data fluctuations.

[0038] Then, knowledge enhancement feature selection is performed on the standardized data, including introducing the domain knowledge graph to which the event object belongs, and strengthening the association between data features and business based on the domain knowledge graph. Specifically, for structured data, principal component analysis is used to reduce dimensionality and the semantic correlation between data features and target variables in the domain knowledge graph is combined to assign business weights to the principal components, thereby avoiding the loss of key business features by pure statistical dimensionality reduction. For text data, entity linking technology is used to associate with the domain knowledge graph to extract composite data features including keywords and entity relationships, which reduces dimensionality while retaining business semantics.

[0039] It should be understood that the pre-aggregation engine supports adaptive preprocessing capabilities such as distributed fingerprint deduplication, improved KNN interpolation missing value handling, and business rule validation outlier filtering. It can achieve real-time fusion of TB-level structured data, unstructured data, and time-series data, increasing data processing throughput to over 100GB / s, meeting the core needs of massive data processing. Furthermore, the core objective of data preprocessing is to improve data quality, but traditional methods are less adaptable to data distribution. This application introduces a knowledge-enhanced feature selection mechanism, allowing preprocessing strategies to dynamically match data characteristics, improving preprocessing adaptability while ensuring data quality.

[0040] In one embodiment, controlling the basic aggregation module to aggregate the preprocessed data includes: controlling the basic aggregation module to aggregate the preprocessed data according to three independent aggregation strategies, including adaptive grouping aggregation, time-aware aggregation, and spatial aggregation.

[0041] Adaptive Grouping and Aggregation: Building upon the traditional grouping query statement `groupBy()`, this feature adds a "self-optimizing grouping dimension" mechanism. Specifically, it's used to group and aggregate data based on the data aggregation attribute requirements of the event object. For example, in retail data, if the aggregation result of "region + product category" is queried three times more frequently than "region + brand," then the former is automatically set as the pre-calculated priority dimension. In Spark, this can be dynamically updated using a custom `agg()` function.

[0042] Time-aware aggregation automatically adjusts the time window size for data aggregation based on the time density of event objects (such as peak-hour event density in traffic flow data). This allows for a second-level sliding window during periods of high event density and a switch to an hour-level window during periods of low event density, ensuring analytical accuracy during critical periods while reducing computational overhead during non-critical periods.

[0043] Spatial aggregation employs a two-tiered aggregation strategy based on the spatial range of an event object, building both a spatial index and hierarchical aggregation. The bottom layer uses R-trees or quadtrees for efficient spatial queries, while the top layer utilizes a regional hierarchical linkage mechanism. This means that when aggregating data from a specific region, it automatically combines the aggregated data from its subordinate regions (rather than recalculating), and when cross-regional comparisons are needed, it automatically merges the aggregated data from adjacent regions. This avoids full data recalculation and improves response speed by over 40%.

[0044] In one embodiment, the intelligent pre-aggregation module dynamically adjusts the aggregation strategy based on the inference feedback and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy. This includes: the intelligent pre-aggregation module obtaining four-fold inference feedback from the event inference model output by the event processing engine. This four-fold inference feedback constitutes a four-dimensional adaptive optimization system of "value, accuracy, efficiency, and effect," including reflecting the business value density and query frequency distribution of the aggregation results, and innovatively introducing a business value decay function. The historical aggregated data is used to dynamically evaluate the value of the aggregation results. Indicates the initial business value. The system provides four key feedback mechanisms: 1) business value at time t; 2) real-time model parameter feedback to capture spatiotemporal focus changes in the event extrapolation model and achieve millisecond-level adaptive adjustment of aggregation granularity; 3) cost feedback to quantify computational resource consumption and time overhead and establish a Pareto-optimal accuracy-efficiency model; and 4) extrapolation effect feedback to monitor changes in prediction error and drive continuous evolution of the aggregation strategy. These four feedback mechanisms are interconnected and dynamically balanced, forming a self-optimizing reinforcement loop for the aggregation strategy.

[0045] Specifically, based on historical aggregated data call feedback, the query frequency and business value of historical aggregated data are obtained. This information is then used to control the basic aggregation module to dynamically adjust the grouping priority of adaptive grouping aggregation, automatically discarding low-value aggregation results and adding high-value aggregation dimensions. The business value assessment employs an exponential decay model, dynamically adjusting the decay coefficient for different business scenarios. This ensures that high-value aggregation dimensions are prioritized.

[0046] Based on real-time model parameter feedback, the time window, spatial range, and accuracy requirements of the current event simulation are analyzed. This allows the basic aggregation module to incorporate a spatiotemporal attention heatmap model, dynamically adjusting the granularity of time-series and spatial aggregation. Fine-grained aggregation is applied to the focused area, while the aggregation granularity is automatically downgraded for non-focused areas, resulting in a 40% improvement in computational resource allocation efficiency. For example, in disaster simulations, when focusing on "rainfall in a county within one hour," a two-dimensional aggregation of "spatial (county) + temporal (1-hour sliding window)" is prioritized.

[0047] Based on the cost feedback analysis, the computing resources (CPU / memory) and time overhead of the current event simulation are obtained, and then the basic aggregation module is controlled to build a multi-level resource buffer pool. When the system load exceeds the threshold, the aggregation granularity degradation sequence is automatically triggered, and the aggregation granularity of the three types of aggregation strategies is dynamically adjusted to achieve a balance between accuracy and efficiency. When resources are tight, the "accuracy-efficiency balance mechanism" can be automatically started - such as temporarily reducing the granularity of spatial aggregation (from street level to district level) and ensuring that the simulation process is not interrupted.

[0048] Based on the feedback from the simulation results, the improvement in the accuracy of the current event simulation (or the reduction in prediction error) is analyzed. This information is then used to control the basic aggregation module to iteratively optimize the selection logic of the aggregation granularity through reinforcement learning. A Q-learning algorithm is employed to construct the aggregation strategy value function, allowing the aggregation strategy to continuously evolve according to the business scenario in which the event object exists. This results in a week-on-week improvement in simulation accuracy of 5% to 8%.

[0049] In one embodiment, event objects are described based on a dynamic graphical object model (DGM). The DGM is a unified abstraction of spatial entities and their motion processes, and a significant extension of the static spatial data model. This model uses "events" as the fundamental driving force of spatial change, explicitly defining events to describe the movement and alteration of spatial entities. The DGM defines all dynamic graphical objects as containing three basic attributes: object identifier, attribute set, and operation set. The object identifier serves as a globally unique identifier, providing a basis for the rapid location and retrieval of corresponding dynamic graphical objects in the database. The attribute set defines the attributes of the dynamic graphical object, including spatial domain, temporal domain, thematic domain, and data aggregation attributes. The spatial domain, temporal domain, and thematic domain describe the spatial location, existence period, and application-related thematic information of the dynamic graphical object, respectively. The data aggregation attributes define the association rules between the dynamic graphical object and multi-source dynamic data, providing fundamental support for real-time data processing and event triggering. The operation set defines the functions of the dynamic graphical object. Different types of dynamic graphics objects have different operations: spatial entities and marker objects typically support shape modification, position movement, combination and splitting, etc.; event objects include operations such as execution, prohibition and triggering, as well as conditional judgment operations based on data aggregation results.

[0050] Dynamic graphic objects are divided into two basic types: spatial entities and marker objects. Spatial entities are graphic objects that depict actual geographic features and objectively exist in space and time (such as houses, rivers, bridges, and vehicles). Their data aggregation attributes include pre-aggregation parameters, specifically aggregation granularity (such as 100-meter grid, 5-minute time window), real-time data sources (such as IoT sensor streams, mobile terminal GPS data), and data aggregation rules (such as weight allocation, outlier filtering), used to efficiently integrate multi-source dynamic data and support real-time updates and analysis of entity status. Marker objects are descriptive graphic symbols displayed on electronic maps that do not represent actual geographic features. They are used to express the specific intentions of the user or creator (such as warning signs, temporary annotations). Their status and display attributes adjust with spatial or temporal changes, but they do not contain data aggregation attributes.

[0051] The change process of a spatial entity in spacetime is defined as an event object. This can be defined based on the entity's state at a specific moment or by recording the entity's behavior over a period of time. Event objects are defined through event identifiers, spatial attributes, temporal attributes, event operation sets, and data aggregation attributes. The event identifier is a unique code for each event object, used for identification and reference. Spatial attributes include the event object's scope of influence (indicating which spatial entities or marked objects the event object triggers changes) and change parameters (specifying the specific change methods of the event object). Temporal attributes define the event object's lifecycle, including the start and end times. The event operation set defines the event object's function. The event object's data aggregation attributes define the data processing rules required for event triggering and execution, including pre-aggregation parameters such as data update thresholds (e.g., triggering aggregation when the location offset exceeds 50 meters), aggregation strategies (e.g., mean, maximum, density clustering), and cache validity (e.g., 30 seconds). This ensures that aggregated data conforming to the rules can be directly called during event processing, reducing the computational overhead of the original data.

[0052] Furthermore, in addition to their own attributes, event objects are closely related to event triggers in the event processing engine. Event triggers define the conditions for event triggering (such as time nodes, spatial entities entering a specific area, and aggregated data reaching a threshold), and also abstractly describe the derivative relationships between events (such as "Event B is executed within 30 seconds after Event A is triggered"). Based on the inherent laws of spatial processes and changes, event objects can be classified according to their triggering source into data-driven (dependent on aggregated data), spatiotemporal-driven (dependent on location or time), and interaction-driven (dependent on user operation). Moreover, multiple dynamic graphical objects (spatial entities, marker objects, and event objects) constitute composite objects. Composite objects are complex structures combined according to specific logical rules, used to express temporal relationships (such as "Event B is triggered after Event A ends"), spatial relationships (such as "Marker E is activated when Entity C is within the range of Entity D"), or semantic relationships (such as "The subordinate relationship between traffic events and associated road segments"). Their attributes are inherited from the core characteristics of the constituent objects.

[0053] In one embodiment, event-driven process simulation is based on a dynamic graphical object model, with an optimized event processing engine at its core. This engine, in collaboration with a pre-aggregation engine, enables efficient spatial process simulation. Spatiotemporal events drive the simulation, spatial entities form the main body of the process, and the spatial changes are ultimately presented by dynamic map display controls. The event processing engine invokes aggregated data to drive event deduction, such as... Figure 3 As shown, it includes:

[0054] After creating and generating event objects based on the inference concept and the real-time status of spatial entities, the event objects are injected into the event queue maintained by the event processing engine for invocation. When the event processing engine invokes and parses the event object, it prioritizes calling the real-time aggregated data cache in the pre-aggregation engine that meets the data aggregation attribute requirements of the event object (such as satisfying aggregation algorithms and validity conditions) to drive the corresponding spatial entity status changes, triggering the generation of new event objects to realize event inference, and significantly reducing the overhead of repeated queries and calculations from the original data source. The aggregated data cache is dynamically updated based on the pre-aggregation parameters of the spatial entities and event objects (such as periodic refreshes at aggregation granularity and updates triggered by data update thresholds). When real-time aggregated data cache that meets the data aggregation attribute requirements is missing or expires, the event processing engine will trigger the pre-aggregation engine to perform real-time data aggregation and synchronously update the cache, ensuring the efficiency of subsequent event processing.

[0055] The event processing engine drives changes in the state of corresponding spatial entities, specifically manifested as positional movement, shape alteration, or display style adjustment. These changes in entity state may trigger new events (e.g., when a moving target enters a danger zone, its aggregated position data exceeds a threshold, triggering an alarm flashing event). Through the event processing engine's step-by-step processing of event sequences, the state of each spatial entity at any given time can be calculated. Finally, using map display controls, the state changes of all spatial entities within the visible range are continuously presented in chronological order, achieving efficient and accurate simulation of spatiotemporal dynamic processes. By introducing data aggregation attributes to spatial entities and event objects, they are endowed with dynamic data perception and preprocessing capabilities. Combined with the collaborative optimization of the event processing engine and the pre-aggregation engine, a closed loop of "data aggregation - event triggering - entity response" is constructed, solving the problems of high real-time data processing latency and redundant raw data querying in traditional models, significantly improving the efficiency and real-time performance of spatial dynamic process simulation.

[0056] In one embodiment, driving the event deduction model to visualize event deduction includes:

[0057] like Figure 4As shown, the event inference model adopts a four-fold closed-loop architecture of "perception, decision-making, presentation, and feedback." This includes building a B / S application system architecture, providing a foundational support layer for a distributed computing environment and geospatial data base; using a unified object model to abstract the spatiotemporal characteristics of spatial entities, event objects, and marked objects, supporting dynamic graphical object components with millisecond-level object state updates; implementing a three-level collaborative architecture of event queue management, TR+ tree spatiotemporal indexing, and event rule processors, with an event processing engine processing latency of less than 50ms; a graphics rendering engine that adaptively matches visualization attributes based on object semantic features and the current rendering environment, supporting 2D / 3D hybrid rendering; and a dynamic graphics service engine that provides real-time monitoring information and dynamic solution information services in the form of web services, with data synchronization errors controlled within 50ms. Through the native embedded interface established between the pre-aggregation engine and the event inference model, the aggregation results are directly converted into scheduling logic instructions for the event processing engine, further reducing response latency to the millisecond level.

[0058] The foundational support layer is used to build a B / S (Browser / Server) application system architecture. Within this architecture, GeoVis geographic data visualization technology and a Web-GIS (Internet Geographic Information System) engine are used to access and display dynamic information in real time, and a web browser is provided to display event simulations in real time. The dynamic graphics object component, event handling engine, graphics rendering engine, and dynamic graphics service engine are essentially extensions of the Web-GIS component's functionality. They collaborate to ensure the efficient processing and presentation of dynamic information.

[0059] Dynamic Graphics Objects (DGOCs) components are used to create spatial entities, event objects, and marker objects with specific features. It's important to note that this component is not a single, fixed module, but rather a set of DGOC interface standards—any component conforming to this standard can be included in the DGOC category and become part of the system. This interface-based design gives the model strong extensibility: when application requirements change, there's no need for large-scale adjustments to the overall model; only targeted upgrades or modifications to the DGOCs are needed to adapt to new requirements, significantly enhancing the model's flexibility and adaptability, and reducing maintenance and iteration costs.

[0060] The event processing engine plays a core scheduling role in the dynamic visualization process, continuously maintaining the event queue and executing multiple core operations sequentially. First, it loads created event objects into the event queue, ensuring effective import of event data. Second, based on event triggering conditions and priorities, it calls aggregated data and dynamically processes and responds in real-time to event objects in the event queue. During dynamic visualization operation, it also dynamically generates new events according to scenario requirements, meeting the changing needs of complex scenarios. Finally, after the dynamic simulation ends, it automatically cleans up the event queue and related runtime environment, releasing occupied resources and avoiding data redundancy or residue, laying a foundation for smooth operation of subsequent simulation tasks.

[0061] The graphics rendering engine is the core module for visualization. It triggers rendering requests based on the dynamic processing results of events from the event handling engine. By invoking the symbol system and combining the current rendering environment with the semantic features inherent in the graphics, it accurately matches and extracts the corresponding visualization attributes of the graphics. Finally, it displays the graphics using the symbol system and, after drawing with either a 2D or 3D graphics rendering engine, pushes the rendering results to the dynamic graphics service engine. Due to the diversity of symbol systems, different symbol systems can generate graphics rendering engines with varying functions and characteristics—even the same spatial entity can present drastically different visual effects through different graphics rendering engines. Furthermore, provided the data model is compatible, 2D or 3D graphics rendering engines can be flexibly selected for drawing, enabling both concise and intuitive 2D view displays and realistic 3D effect presentations, meeting diverse visualization needs.

[0062] The dynamic graphics service engine is derived from the basic GIS service engine. In essence, it is a GIS service engine used to provide dynamic information services to clients in the form of web services. It covers real-time monitoring information services and dynamic solution information services, and realizes real-time dynamic display of event inference through the web browser in the basic support layer.

[0063] In one embodiment, such as Figure 5 As shown, the event processing engine includes an event queue, an event index, and an event rule processor. It's important to note that the objects in this engine themselves do not have the ability to directly identify specific event objects; however, by setting up virtual event interfaces, access to various event objects can be achieved, ensuring the integrity and scalability of the event processing flow. Figure 5 The diamond symbol represents a combination relationship, where "1" indicates that the quantity correspondence between the two components is one. "This indicates that there are multiple corresponding relationships between two components, such as the relationship between an event rule handler and an event trigger." "" means that one event rule processor can correspond to multiple event triggers.

[0064] The event queue consists of multiple event objects, used to maintain a list of event objects to be processed during event derivation, ensuring that all event objects to be processed are managed in an orderly manner. The event index is used to build a TR+ tree index based on the time characteristics of the event queue. This TR+ tree (ternary search tree) index enables fast lookup of event objects, improving event retrieval efficiency. The event rule processor contains multiple event triggers, used during the invocation of an event object to determine whether the execution of the current event object will trigger a new event object; if it is determined that a new event object will be generated, it is responsible for generating the new event object and adding it to the event queue, ensuring the continuity of the event processing flow. Event triggers are used to abstract the derivation relationships between event objects. By defining formal judgment rules, the conditions that must be met for the generation of an event object are clarified, providing a basis for the orderly derivation of event objects.

[0065] It should be understood that event-driven dynamic visualization, using an event simulation model, is an event-driven dynamic visualization process. Event-driven dynamic visualization mainly comprises three stages: dynamic data acquisition, dynamic data processing, and dynamic display. The dynamic data source primarily provides dynamic information for dynamic visualization, specifically encompassing real-time information, replay information, and simulation information. This data source consists of two parts: a real-time information acquisition component and a dynamic spatial database. Dynamic data processing is implemented through an event processing engine, while dynamic display is achieved using map display controls. The event-driven dynamic visualization process is as follows: Figure 6 As shown.

[0066] Furthermore, in the dynamic visualization process, the state evolution of spatial entities mainly relies on the execution of specific events, and event objects, as the core component for realizing dynamic visualization, play a crucial role. The event object system includes various types, specifically covering movement event objects, deformation event objects, thematic attribute change event objects, generation event objects, disappearance event objects, merging event objects, segmentation event objects, blinking event objects, and map interaction events. These objects support the dynamic presentation of spatial entity changes by defining different event logics. To facilitate understanding of the implementation mechanism of event objects, we will now use the movement event object, which has the most widespread application scenarios, as an example. This approach of analyzing the core mechanism through typical cases can clearly present the working principle of event objects in dynamic visualization without increasing the complexity of understanding. The implementation mechanism of the movement event object is as follows: Figure 7 As shown. This application achieves real-time response and state update of mobile event objects by constructing a high-concurrency, low-latency event-driven architecture. This mechanism aims to solve the performance bottleneck and system stability problems in the processing of massive mobile event data, ensuring the continuity and accuracy of the event inference process. The specific business implementation process is as follows:

[0067] (1) Event Generation Phase: Upon receiving raw information, the event generator transforms the raw information into standardized event objects. This process includes timestamp injection, spatial coordinate resolution, and attribute normalization to ensure that the events have a unified data structure and semantic integrity. For movement events, the event generator will also determine whether they are valid movement trajectory points according to preset rules and mark their event type.

[0068] (2) Event encapsulation and enqueueing: The generated event object is encapsulated as an "event containing original information" and written to a circular buffer. This buffer adopts a lock-free circular queue design, supporting high-throughput writing and reading, and avoiding performance bottlenecks caused by memory allocation. The buffer capacity can be dynamically adjusted according to the system load to ensure stability under sudden traffic surges.

[0069] (3) Multi-threaded event processing: To meet the real-time processing needs of massive mobile events, an event processing thread pool was introduced. This thread pool is dynamically created and managed by the event processing thread generator, and automatically expands or shrinks the number of threads according to the current event queue length and CPU utilization to achieve elastic resource scheduling. Each thread executes the processing logic in the event processing class, and is responsible for retrieving events from the circular buffer and calling the corresponding business rules to perform operations such as status judgment, trajectory calculation, and correlation analysis. The event processing class adopts an object-oriented design and supports plug-in extension, which facilitates the configuration of different processing strategies according to different types of mobile events.

[0070] (4) Task Flow Collaborative Processing: Intermediate results or derivative tasks generated during the processing will be submitted to the task flow processing class to form an asynchronous task chain. This class supports task priority sorting, dependency management and timeout control to ensure the orderly execution of complex deduction logic.

[0071] (5) Exception handling mechanism: If data format errors, network interruptions, or calculation failures occur during event handling, the system will automatically trigger the exception handling class. This class is responsible for recording exception logs, retry mechanisms (such as exponential backoff), alarm notifications, and degradation processing (such as skipping the current event and using historical data to complete the event), ensuring that the entire system has robustness and fault tolerance.

[0072] (6) Status update and visualization feedback: After the above processing, the status of the moving event object is updated in real time to the dynamic graphic object model, driving the event inference model to perform visualization presentation, realizing a closed loop from data input to dynamic inference.

[0073] A movement event object is primarily defined by a set of movement trajectory points and a trajectory type. The set of movement trajectory points consists of a group of movement trajectory points. A trajectory point is a point with a time parameter: in two-dimensional space, it is represented as (X, Y, T); in three-dimensional space, it is represented as (X, Y, T), where X and Y represent spatial coordinates, and T represents the time at the corresponding location. The trajectory type defines the movement mode of the spatial entity; common trajectory types include polylines and spline curves. During the simulation, the trajectory type directly determines the position of the sampling points. The execution of a movement event refers to the target moving along the path defined by the set of trajectory points. The specific algorithm is as follows:

[0074] In the set of moving trajectory points, find the trajectory point that is closest to time t (i.e., has the smallest time difference) from all trajectory points whose time is earlier than the current time t. Find in the set of movement trajectory points The next trajectory point, i.e. The coordinates of the sampling point corresponding to time t are calculated using a sampling algorithm. For a polyline trajectory, the uniform sampling algorithm is as follows:

[0075] ;

[0076] .

[0077] In summary, the event inference visualization method based on a pre-aggregation engine provided in this application efficiently processes massive amounts of data through a pre-aggregation engine, including intelligent preprocessing, basic aggregation, and intelligent pre-aggregation. It then combines this with an event inference model based on an event processing engine for dynamic event inference prediction and visualization. The pre-aggregation engine is not an independent pre-module but is deeply integrated with the event inference model. It can dynamically adjust the aggregation strategy based on the real-time needs and feedback of the inference process, obtaining real-time optimized aggregated data and feeding it back to optimize the event inference. This forms a two-way feedback architecture of "inference needs → aggregation strategy → data output → inference optimization." This not only accurately adapts to the processing needs of massive real-time data, significantly improving the data processing efficiency and reducing response latency of event inference, but also strengthens the data's support accuracy for inference prediction and reduces invalid data interference through the dynamic linkage optimization of the aggregation strategy and event inference. This makes the inference prediction results more closely reflect actual scene changes, improving the prediction accuracy and dynamic visualization capabilities of event inference.

[0078] Furthermore, this application reconstructs the visualization logic of event deduction, achieving a technological upgrade from "static display" to "dynamic interactive deduction." Specifically, it achieves unified spatiotemporal expression through a dynamic graphical object model: abstracting spatial entities, events, markers, and other elements into unified dynamic graphical objects, and solving the fragmented expression of spatial process changes in traditional GIS through a four-dimensional attribute definition of "spatial domain + temporal domain + thematic domain + data aggregation attributes," supporting second-level drill-down analysis from macro-level trends to micro-level events. Low-latency visualization fusion of multi-source heterogeneous data: relying on the dynamic caching of the event processing engine and the TR+ tree spatiotemporal index, real-time synchronous rendering of multimodal data is achieved, with the dynamic rendering frame rate increased to 60fps@4K resolution, and the multi-source data synchronization visualization error controlled within 50ms, completely solving the problem of "key information being buried." Dynamic reconstruction of the entire event lifecycle: based on an event-driven simulation mechanism, combined with the collaborative scheduling of multiple types of event objects such as movement, deformation, and merging, the entire lifecycle of an event from "occurrence → evolution → related influences → result" can be completely reconstructed, solving the problem of decision-making information gaps caused by traditional "fragmented display." Furthermore, through multi-scale spatial aggregation and event sequence aggregation by the pre-aggregation engine, it automatically associates multi-dimensional information such as the spatiotemporal attributes and scope of influence of events to construct a complete event data map, which is conducive to providing decision-makers with a "holographic" event view and improving the completeness of decision-making basis by 70%.

[0079] In one embodiment, an event deduction visualization device based on a pre-aggregation engine is provided, comprising:

[0080] The architecture building module is used to construct a pre-aggregation engine, which includes an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and to connect the pre-aggregation engine to a pre-built event inference model based on an event processing engine.

[0081] The data aggregation module is used to analyze in real time whether the pre-aggregation engine has cached aggregated data that meets the data aggregation attribute requirements of the event object. If so, it directly calls the aggregated data; otherwise, it analyzes the data aggregation attribute requirements of the event object according to the intelligent pre-aggregation module, controls the intelligent preprocessing module to collect data related to the event object from multi-source heterogeneous data for preprocessing according to the data aggregation attribute requirements, and controls the basic aggregation module to aggregate the preprocessed data to obtain aggregated data and cache it for the event processing engine to call.

[0082] The bidirectional feedback optimization module is used by the event processing engine to call aggregated data to drive event inference and drive the event inference model to visualize the event inference. At the same time, it provides real-time feedback on the real-time requirements and inference results of the inference process to the pre-aggregation engine. Based on the inference feedback, it drives the intelligent pre-aggregation module to dynamically adjust the aggregation strategy and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy, so as to obtain real-time optimized aggregated data and feed it back to the event processing engine for event inference optimization.

[0083] Specific limitations regarding the event deduction visualization device based on the pre-aggregation engine can be found in the limitations of the event deduction visualization method based on the pre-aggregation engine mentioned above, and will not be repeated here. Each module in the aforementioned event deduction visualization device based on the pre-aggregation engine can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0084] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements an event deduction visualization method based on a pre-aggregation engine. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0085] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0086] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to perform the following steps:

[0087] Construct a pre-aggregation engine that includes an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and integrate the pre-aggregation engine into a pre-built event inference model based on an event processing engine;

[0088] The event processing engine analyzes in real time whether there is aggregated data in the pre-aggregation engine that meets the data aggregation attribute requirements of the event object. If so, the aggregated data is called directly. If not, the intelligent pre-aggregation module analyzes the data aggregation attribute requirements of the event object and controls the intelligent pre-processing module to collect data related to the event object from multi-source heterogeneous data for preprocessing. It also controls the basic aggregation module to aggregate the preprocessed data, obtain aggregated data, and cache it for the event processing engine to call.

[0089] The event processing engine calls aggregated data to drive event inference and drives the event inference model to visualize the event inference. At the same time, it provides real-time feedback on the inference process and the inference effect to the pre-aggregation engine. Based on the inference feedback, the intelligent pre-aggregation module dynamically adjusts the aggregation strategy and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy, so as to obtain real-time optimized aggregated data and feed it back to the event processing engine for event inference optimization.

[0090] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0091] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application.

Claims

1. A method for pre-aggregation engine based event deduction visualization, the method comprising: The method comprises: constructing a pre-aggregation engine including an intelligent preprocessing module, a basic aggregation module, and an intelligent pre-aggregation module, and connecting the pre-aggregation engine to an event deduction model of a pre-constructed event processing engine; According to the event processing engine, it is judged whether the data aggregation data meeting the data aggregation attribute requirements of the event object is cached in the pre-aggregation engine in real time. If yes, the aggregation data is directly called. If not, the data aggregation attribute requirements of the event object are analyzed according to the intelligent pre-aggregation module, the data related to the event object is collected from the multi-source heterogeneous data for preprocessing according to the data aggregation attribute requirements, and the preprocessed data is aggregated by the basic aggregation module to obtain the aggregation data and cache it for calling by the event processing engine; The event deduction is driven by the event processing engine calling the aggregation data, and the event deduction model is driven to visualize the event deduction, and the real-time requirements and deduction effects of the deduction process are fed back to the pre-aggregation engine in real time. The intelligent pre-aggregation module dynamically adjusts the aggregation strategy according to the deduction feedback, and controls the basic aggregation module to aggregate data according to the adjusted aggregation strategy to obtain real-time optimized aggregation data and feed back to the event processing engine for event deduction optimization.

2. The pre-aggregated engine based event deducing visualization method of claim 1, wherein, According to the data aggregation attribute requirements, the intelligent preprocessing module collects data related to the event object from the multi-source heterogeneous data for preprocessing, which comprises: Adaptive data cleaning is performed on the collected data, including data deduplication, missing value filling, and outlier processing. The missing value filling includes dynamic filling decision according to the missing mode. If it is randomly distributed missing, improved KNN interpolation with local data density weight is used for filling. If it is systematic missing, missing attribution analysis is triggered, and combined with the domain rule to which the event object belongs, a filling value with confidence is generated; The cleaned data is subjected to hierarchical standardization processing, including Min-Max or Z-score standardization for structured data, and dynamic window alignment for time series data, with the window size being self-adaptively adjusted according to the data fluctuation frequency; Knowledge enhanced feature selection is performed on the standardized data, including introducing the domain knowledge graph to which the event object belongs, and strengthening the association between data features and business according to the domain knowledge graph. For structured data, principal component analysis is used to reduce dimension and combine the semantic correlation between data features and target variables in the domain knowledge graph to give business weight to principal components. For text data, entity linking technology is used to associate the domain knowledge graph to extract complex data features including keywords and entity relationships.

3. The pre-aggregated engine based event deducing visualization method of claim 2, wherein, The preprocessed data is aggregated by the basic aggregation module, which comprises: The preprocessed data is aggregated by the basic aggregation module according to three independent aggregation strategies, including adaptive grouping aggregation, time sequence perception aggregation, and spatial aggregation; The adaptive grouping is used for data grouping aggregation according to data aggregation attribute requirements of the event object; the time sequence awareness aggregation is used for automatically adjusting a time window size to perform data aggregation according to a time density of occurrence of the event object; and the spatial aggregation is used for constructing a double-layer aggregation strategy of spatial indexing and hierarchical aggregation according to a spatial range in which the event object is located, a bottom layer of the double-layer aggregation strategy uses an R-tree or a quad-tree to perform efficient spatial query, and a top layer of the double-layer aggregation strategy uses a regional hierarchical linkage mechanism, that is, when data of a region is aggregated, aggregated data of regions under jurisdiction of the region is automatically combined, and when comparison across regions is required, aggregated data of adjacent regions is automatically combined.

4. The pre-aggregated engine based event deducing visualization method of claim 3, wherein, The intelligent pre-aggregation module obtains fourfold inference feedback output by an event inference model of the event processing engine, including historical aggregated data calling feedback, real-time model parameter feedback, cost feedback and inference effect feedback; According to the real-time model parameter feedback, a time window, a spatial range and precision requirements of current event inference focus are obtained, and then the basic aggregation module is controlled to introduce a time and space attention heat map model, dynamically adjust granularities of the time sequence awareness aggregation and the spatial aggregation, implement fine-grained aggregation on an inference focus region, and automatically degrade an aggregation granularity in a non-focus region; According to the historical aggregation data call feedback analysis, the query frequency and business value of the historical aggregation data are obtained, and then the group priority of the adaptive grouping aggregation of the basic aggregation module is dynamically adjusted, the low-value aggregation result is automatically eliminated, and the high-value aggregation dimension is added; wherein the business value evaluation adopts an exponential decay model, and the decay coefficient is dynamically adjusted for different business scenarios , ensuring that high-value aggregation dimensions are given priority According to the cost feedback, a calculation resource and time overhead of current event inference are obtained, and then the basic aggregation module is controlled to construct a multi-level resource buffer pool, when system load exceeds a threshold value, an aggregation granularity degradation sequence is automatically triggered, granularities of the three types of aggregation strategies are dynamically adjusted to realize precision and efficiency balance, and the inference process is ensured not to be interrupted; According to the inference effect feedback, an accuracy improvement amplitude of current event inference is obtained, and then the basic aggregation module is controlled to iteratively optimize selection logic of the aggregation granularity through reinforcement learning, a Q-learning algorithm is used to construct an aggregation strategy value function, and the aggregation strategy is continuously evolved with a business scenario in which the event object is located. The event object is described based on a dynamic graphic object model, the dynamic graphic object model defines that all dynamic graphic objects include three types of basic attributes, including an object identifier, an attribute set and an operation set; the object identifier is used as a globally unique identifier, and provides a basis for fast positioning and retrieval of a corresponding dynamic graphic object in a database; the attribute set is used to define attributes of the dynamic graphic object, the attributes include a spatial domain, a time domain, a thematic domain and a data aggregation attribute, the spatial domain, the time domain and the thematic domain are respectively used to describe a spatial position, an existence period and application field related thematic information of the dynamic graphic object, the data aggregation attribute is used to define an association rule of the dynamic graphic object and multi-source dynamic data, and provides a basis support for real-time data processing and event triggering; and the operation set is used to define a function of the dynamic graphic object.

5. The pre-aggregated engine based event deducing visualization method according to any one of claims 1 to 4, characterized in that, ​ The dynamic graphic object is divided into two basic types of spatial entities and marked objects; wherein the spatial entity is a graphic object depicting an actual geographic element, and the data aggregation attribute includes a pre-aggregation parameter, specifically an aggregation granularity, a real-time data source, and a data aggregation rule; the marked object is an explanatory graphic symbol displayed in an electronic map, which does not represent an actual geographic feature, and is used to express a specific intention of a user or a maker, and the state and display attribute thereof are adjusted according to space or time changes, but does not include a data aggregation attribute; The change process of the spatial entity in space-time is defined as an event object, and the event object is defined by an event identifier, a spatial attribute, a time attribute, an event operation set, and a data aggregation attribute; the event identifier is a unique code of each event object, used for identification and reference; the spatial attribute includes an influence range and a change parameter of the event object; the time attribute is used to define the life cycle of the event object, including a start time and an end time; the event operation set is used to define the function of the event object; and the data aggregation attribute of the event object is used to define the processing rule of the required data when the event is triggered and executed, including a pre-aggregation parameter, specifically a data update threshold, an aggregation strategy, and a cache validity period.

6. The pre-aggregated engine based event deducing visualization method of claim 5, wherein, The aggregation data driven is called by the event processing engine for event deduction, including: After the event object is generated based on the deduction concept and the real-time situation of the spatial entity, the event object is injected into an event queue maintained by the event processing engine for calling, when the event processing engine calls and parses the event object, the real-time aggregation data cache in the pre-aggregation engine that meets the data aggregation attribute requirements of the event object is preferentially called to drive the state change of the corresponding spatial entity, and a new event object is triggered to generate to realize event deduction; wherein the aggregation data cache is dynamically updated based on the pre-aggregation parameters of the spatial entity and the event object, when the real-time aggregation data cache that meets the data aggregation attribute requirements is missing or expired, the event processing engine will trigger the pre-aggregation engine to perform real-time data aggregation and update the cache.

7. The pre-aggregated engine based event deducing visualization method of claim 6, wherein, The event deduction model is driven for event deduction visualization, including: The event deduction model includes a basic support layer, a dynamic graphic object component, an event processing engine, a graphic rendering engine, and a dynamic graphic service engine; The basic support layer is used to build a B / S application system architecture, under the B / S application system architecture, dynamic information is accessed and displayed in real time by means of GeoVis geographic data visualization technology and a Web-GIS engine, and a Web browser is provided to realize real-time display of event deduction; The dynamic graphic object component is used to create spatial entities, event objects, and marked objects of feature dimensions; The event processing engine is used to load the created event objects into an event queue, call aggregation data, and dynamically process and respond to the event objects in the event queue according to event triggering conditions and priorities; in dynamic visualization running, new events are also dynamically generated according to scene requirements; and after the dynamic simulation is completed, the event queue and related running environment are automatically cleaned up, and the occupied resources are released. The graphic rendering engine is used for triggering a rendering request according to an event dynamic processing result of the event processing engine, and combining a current rendering environment and semantic features of the graphic by calling a symbol system, so as to accurately match and extract visual properties of the graphic, finally display the graphic by means of the symbol system, and push a rendering result to the dynamic graphic service engine after drawing by selecting a two-dimensional or three-dimensional graphic rendering engine. The dynamic graphic service engine is used for providing dynamic information service to a client in a Web service form, covering real-time monitoring information service and dynamic scheme information service, and realizing real-time dynamic display of event deduction through a Web browser in the basic support layer.

8. The pre-aggregated engine based event deduction visualization method of claim 1 or 4 or 6 or 7, wherein, The event processing engine comprises an event queue, an event index and an event rule processor. The event queue is composed of a plurality of event objects and is used for maintaining a list of event objects to be processed in an event deduction process. The event index is used for constructing a TR+ tree index according to time characteristics of the event queue, and performing fast searching of the event objects through the TR+ tree index. The event rule processor comprises a plurality of event triggers and is used for judging whether a new event object will be triggered after execution of a current event object in a calling process of the event object; if it is judged that a new event object will be generated, generating the new event object and adding the new event object to the event queue. The event trigger is used for abstractly processing derivative relationships between the event objects, formulating a formalized judgment rule, and clearly defining conditions required for generation of the event objects, so as to provide a basis for ordered derivation of the event objects.

9. A pre-aggregation engine based event deduction visualization apparatus, characterized in that, The apparatus comprises: An architecture building module, configured to build a pre-aggregation engine comprising an intelligent preprocessing module, a basic aggregation module and an intelligent pre-aggregation module, and connect the pre-aggregation engine to a pre-built event deduction model based on an event processing engine; An aggregation data calling module, configured to analyze whether there is aggregation data meeting data aggregation attribute requirements of an event object in the pre-aggregation engine in real time according to the event processing engine, if yes, directly call the aggregation data, and if no, analyze the data aggregation attribute requirements of the event object according to the intelligent pre-aggregation module, control the intelligent preprocessing module to collect data related to the event object from multi-source heterogeneous data for preprocessing according to the data aggregation attribute requirements, and control the basic aggregation module to aggregate the preprocessed data to obtain aggregation data and cache the aggregation data for calling by the event processing engine; A bidirectional feedback optimization module, configured to drive event deduction by calling the aggregation data by the event processing engine, drive the event deduction model to perform event deduction visualization, simultaneously feed back real-time requirements and deduction effects of the deduction process to the pre-aggregation engine in real time, dynamically adjust aggregation strategies of the intelligent pre-aggregation module according to the deduction feedback, and control the basic aggregation module to perform data aggregation according to the adjusted aggregation strategies to obtain real-time optimized aggregation data and feed back the aggregation data to the event processing engine for event deduction optimization. 10.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-9. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 8.

Images