Search

CN-121980071-A - Visual analysis method and platform for real-time trust monitoring data

CN121980071ACN 121980071 ACN121980071 ACN 121980071ACN-121980071-A

Abstract

The invention discloses a visual analysis method and a visual analysis platform for real-time trust monitoring data, which relate to the technical field of data visualization and real-time monitoring, and are used for carrying out standardized processing on real-time evidence elementary streams and relationship edges, establishing a hierarchical index structure of a time slice and an entity cluster, dynamically screening high-precision data blocks according to user attention sets and thresholds, carrying out propagation boundary cutting, loading and prefetching the data blocks through a self-adaptive scheduling strategy by using real-time monitoring system resources, and dynamically adjusting the visualization granularity based on rendering performance indexes to realize flicker-free high-precision data replacement. The platform includes a plurality of functional modules that are interconnected. The invention solves the problems of low data processing efficiency, visual blocking and insufficient multi-granularity display in the prior art, and realizes the efficient and smooth visual analysis of large-scale real-time trust data.

Inventors

  • DONG XIAOMENG
  • YU JING

Assignees

  • 北京亿家老小科技有限公司

Dates

Publication Date
20260505
Application Date
20260126

Claims (10)

  1. 1. A visual analysis method for real-time trust monitoring data is characterized by comprising the following steps of; Step S1, receiving evidence elementary streams and relation edges, completing standardized processing of edge weights and behavior characteristics, generating sub-graph summaries of time slices and entity clusters according to time slices and entity clusters, merging from bottom to top to form multi-resolution hierarchical indexes, outputting an index table, a block catalog and external memory blocks, and setting version identification and verification summaries; S2, under the constraint of a concerned time window and a concerned entity set, calculating evidence node weight based on source reliability, multi-source consistency and time attenuation, matching index positioning candidate blocks, cutting a visible set according to a propagation radius, and sequencing the blocks to be unfolded; step 3, monitoring bandwidth, bandwidth utilization, determining concurrency and batch size of memory and CPU load in real time, maintaining a current loading queue and a pre-fetching queue, scheduling high-weight blocks according to priority, determining a pre-fetching window based on user interaction prediction and performing self-adaptive scheduling; And S4, replacing summary data with high-precision blocks at a frame switching point, monitoring the ratio of frame interval statistics to loading completion, and when the rollback is triggered, raising the unfolding threshold, reducing the propagation radius, performing version consistency check and performing marking display on a time range and an entity cluster.
  2. 2. The visual analysis method of real-time trust monitoring data according to claim 1, wherein the evidence elementary streams and the relationship edges are normalized, the entities are aggregated based on the relationship strength and the behavior similarity to obtain entity clusters, the evidence is segmented according to time slices and sub-graph summary information is calculated, the multi-level index structure is generated by combining the cluster dimension and the time dimension from bottom to top, and the mapping relationship between the index table and the block catalogue is established.
  3. 3. The visual analysis method of real-time trust monitoring data according to claim 2, wherein the relationship strength is obtained by weighted combination of side weight attributes, the behavior similarity is obtained by a similarity function of behavior characteristics, the entity clusters are controlled to be combined by taking the combination cost and a threshold value as criteria, and the generated time slice x entity cluster subgraphs store abstract characteristics for subsequent retrieval and rendering replacement; The index table, the partition directory and the external memory partition carry version identification and verification abstracts, the memory and the external memory corresponding objects are regularly compared, updating is carried out when inconsistency is found, a correction log is generated, and the corrected time range and the entity cluster are marked for displaying and tracing; Under the constraint of the concerned time window and the concerned entity set, calculating the weight of the evidence node as the product of the source reliability, the multi-source consistency score and the time attenuation factor; and setting a spreading threshold and a lower limit of the time overlapping rate, and distinguishing the blocks needing to be loaded by the abstracts from the blocks needing to be spread.
  4. 4. The visual analysis method of real-time trust monitoring data according to claim 1, wherein the continuous time axis is divided into time slices, and entity clusters are aggregated based on relationship strength and behavior similarity; the time slice length is: , wherein, Is a natural number of the Chinese characters, Is the minimum of event intervals; the time slice number is: Wherein, the method comprises the steps of, Is the time stamp of the event and, Is the length of each time slice and, Representation pair Performing downward rounding; The relationship strength is: Wherein, the method comprises the steps of, As the weight of the edge-weight attribute, Is the similarity of the related behaviors.
  5. 5. The visual analysis method of real-time trust monitoring data according to claim 4, wherein the relationship strength is Obtained by weighting and summing after normalizing the edge attribute vectors, the weight is determined by a analytic hierarchy process and meets the following requirements And 。
  6. 6. The visual analysis method of real-time trust monitoring data according to claim 4, wherein the behavior similarity Cosine similarity is measured for static vectors and mapped To the interval of [0,1], the DTW distance is taken for the time sequence And normalized to Wherein, the method comprises the steps of, Is the average sequence length.
  7. 7. The visual analysis method of real-time trust monitoring data according to claim 1, wherein network bandwidth, memory and CPU load are monitored and a concurrency upper limit and a batch size are calculated to schedule data loading, wherein the concurrency upper limit Batch size Meeting the constraint And Wherein For the bandwidth of the network, In order to be in a free memory, As a single block of the average number of bytes, For a single block of the average memory footprint, Is the safety coefficient of the internal memory, Time is allowed for I/O operations.
  8. 8. The visual analysis method of real-time trust monitoring data according to claim 7, wherein the next window is predicted based on user interaction behavior, and the length of the window is prefetched Wherein The average speed of the timeline is dragged for the user, In order to focus on the frequency of change of the set of entities, And Is a coefficient.
  9. 9. The visual analysis method of real-time trust monitoring data according to claim 1, wherein the method is characterized in that high-precision data is used for replacing a low-precision abstract during rendering, and a cartoon state triggering adjustment strategy is monitored, comprising adjustment of an expansion threshold, a propagation radius and a concurrency degree, wherein the adjustment of the expansion threshold is to increase or decrease the number of time slices or data blocks which need to be expanded, the adjustment of the propagation radius is to reduce or enlarge the range of a propagation boundary, and the adjustment of the concurrency degree is to adjust the concurrency number of data loading according to resource conditions; 95 quantile values for a sequence of frame intervals Exceeding a maximum allowable value or load completion ratio If yes, determining that the device is in a stuck state, wherein, To load the lower limit of the completion ratio.
  10. 10. A real-time trust monitoring data visualization analysis platform for implementing the real-time trust monitoring data visualization analysis method according to any one of claims 1-9, characterized in that; The hierarchical index construction module is used for receiving the evidence elementary stream and the relation edge, finishing the standardized processing of the edge weight and the behavior characteristic, generating a sub-graph abstract of a time slice and an entity cluster according to the time slice and the entity cluster, combining from bottom to top to form a multi-resolution hierarchical index, outputting an index table, a block catalog and an external memory block, and setting a version identification and a verification abstract; Under the constraint of the concerned time window and the concerned entity set, calculating evidence node weight based on source reliability, multi-source consistency and time attenuation, matching index positioning candidate blocks, cutting a visible set according to a propagation radius, and sequencing the blocks to be unfolded; the resource scheduling pre-fetching module monitors bandwidth, bandwidth utilization rate, concurrency and batch size of memory and CPU load determination, maintains a current loading queue and a pre-fetching queue, schedules high-weight blocks according to priority, determines a pre-fetching window based on user interaction prediction and performs self-adaptive scheduling; And the rendering rollback correction module is used for replacing abstract data by high-precision blocks at a frame switching point, monitoring the ratio of frame interval statistics to loading completion, and when the rollback is triggered, the spreading threshold is adjusted upwards, the propagation radius is reduced, the version consistency check is carried out, and the mark display is carried out on the related time range and the entity cluster.

Description

Visual analysis method and platform for real-time trust monitoring data Technical Field The invention relates to the technical field of data visualization and real-time monitoring, in particular to a real-time trust monitoring data visualization analysis method and a platform. Background With rapid development of information technology, especially in the fields of distributed systems, social networks, internet of things, financial wind control and the like, real-time monitoring and evaluation of trust relationships among entities are becoming increasingly important. Trust data is typically generated in a dynamic, multi-source, high-dimensional form, covering multiple dimensions of user behavior, interaction records, source reliability, and the like. How to perform efficient and visual analysis on these data to support real-time decision making and risk identification has become a hotspot for current research and application. In the prior art, some methods have attempted to visually display trust relationship data, such as network visualization based on graph structures, time-series trust trend graphs, and the like. However, these methods still suffer from the following deficiencies in processing large-scale real-time data: The traditional method usually adopts a full loading or static aggregation strategy, cannot adapt to the high throughput and low delay requirements of a real-time data stream, so that the visual response is slow, the user experience is influenced, a dynamic scheduling mechanism for data importance, timeliness and visual load is lacked, interface jamming or information loss caused by data overload often occurs, a hierarchical index structure of time and entity dimension is not established in most of the traditional system, seamless switching from macroscopic trend to microscopic detail of a user is difficult to support, and when the data is updated in real time, the visual content is inconsistent with the real state and analysis and judgment are influenced easily due to the lack of an effective version management and correction mechanism. The present invention proposes a solution to the above-mentioned problems. Disclosure of Invention In order to overcome the above-mentioned drawbacks of the prior art, embodiments of the present invention provide a method and a platform for visual analysis of real-time trust monitoring data, so as to solve the problems set forth in the background art. In order to achieve the above purpose, the present invention provides the following technical solutions: a visual analysis method for real-time trust monitoring data comprises the following steps; Step S1, receiving evidence elementary streams and relation edges, completing standardized processing of edge weights and behavior characteristics, generating sub-graph summaries of time slices and entity clusters according to time slices and entity clusters, merging from bottom to top to form multi-resolution hierarchical indexes, outputting an index table, a block catalog and external memory blocks, and setting version identification and verification summaries; S2, under the constraint of a concerned time window and a concerned entity set, calculating evidence node weight based on source reliability, multi-source consistency and time attenuation, matching index positioning candidate blocks, cutting a visible set according to a propagation radius, and sequencing the blocks to be unfolded; step 3, monitoring bandwidth, bandwidth utilization, determining concurrency and batch size of memory and CPU load in real time, maintaining a current loading queue and a pre-fetching queue, scheduling high-weight blocks according to priority, determining a pre-fetching window based on user interaction prediction and performing self-adaptive scheduling; And S4, replacing summary data with high-precision blocks at a frame switching point, monitoring the ratio of frame interval statistics to loading completion, and when the rollback is triggered, raising the unfolding threshold, reducing the propagation radius, performing version consistency check and performing marking display on a time range and an entity cluster. In a preferred embodiment, step S1 comprises the following: normalizing the evidence elementary streams and the relationship edges, aggregating the entities based on the relationship strength and the behavior similarity to obtain entity clusters, segmenting the evidence according to time slices, calculating sub-graph summary information, merging from bottom to top according to cluster dimensions and time dimensions to generate a multi-level index structure, and establishing a mapping relation between an index table and a block directory; the relation strength is obtained by weighted combination of edge weights, the behavior similarity is obtained by a similarity function of behavior characteristics, entity cluster merging is controlled by taking merging cost and a threshold value as criteria, and the generated time slic