Large-sample exploratory experiment conclusion macro-micro interactive verification method and large-sample exploratory experiment conclusion macro-micro interactive verification system

Abstract

The invention relates to a large sample exploratory experiment conclusion macro-micro interaction verification method and system, the method is used for constructing a'macro-interaction-micro 'three-layer linkage analysis framework, and the framework comprises a macro conclusion layer, an interaction analysis layer and a micro instance layer; the interaction analysis layer is used for connecting the macroscopic conclusion layer and the microscopic instance layer; the method comprises the four steps of automatic discovery and presentation of a macroscopic conclusion, interactive deep exploration and verification based on the macroscopic conclusion, macro-micro bidirectional linkage and sample screening, and concrete redisk and causal verification of micro instances. Through the method, interactive verification of a macroscopic conclusion and a microcosmic instance is realized, and the interpretability and credibility of the conclusion are remarkably improved. According to the method, a macroscopic and microcosmic barrier is broken through, a cognitive closed loop from'knowing the course 'to'knowing the course' is constructed, and a more robust basis is provided for decision making.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to a method and system for macro-micro interactive verification of conclusions from large-sample exploratory experiments. Background Technology

[0002] With the development of computer simulation and data analysis technologies, large-sample exploratory experiments have become a key research tool in scientific research and business. These experiments generate massive amounts of high-dimensional data, aiming to discover general conclusions about system operation. Currently, the analysis methods for such experimental data mainly include isolated macro-statistical analysis and micro-recapitulation of single samples. Macro-statistical analysis uses algorithms to calculate the entire sample, deriving global, abstract statistical conclusions (such as a positive correlation between factor A and indicator B); while micro-recapitulation focuses on a single, specific sample that has already occurred, tracing the reasons for its success or failure through process replay, its core being "recapitulating a story." These two methods are usually used separately; analysts first review the macro-statistical report, then manually select samples of interest, and then use independent recapitulation tools for observation.

[0003] However, the aforementioned existing technologies have the following serious shortcomings when processing large-sample exploratory experimental data: First, there is a severe disconnect between macroscopic statistical conclusions and microscopic specific instances. Analysts cannot intuitively correlate abstract statistical conclusions with specific sample evolution processes, resulting in a superficial understanding and insufficient trust in heuristic conclusions. Second, global statistical analysis often obscures a large number of interaction effects and nonlinear conclusions that only hold true under specific conditions, leading to a serious lack of depth in the discovery of heuristic conclusions. Finally, the entire analysis process is linear and passive, making it difficult for analysts to conduct efficient interactive exploration and verification of data based on their own insights and hypotheses, resulting in low analytical efficiency and difficulty in forming a cognitive loop.

[0004] Therefore, how to solve the serious disconnect between macro-statistical conclusions and micro-specific examples, which makes it difficult to deeply understand and reliably verify the conclusions; how to solve the problem that traditional statistical analysis methods are unable to discover complex interaction effects and conditional conclusions hidden under global data, resulting in insufficient depth of conclusion discovery; and how to solve the technical problems that the analysis process is linear and passive, which prevents analysts from conducting hypothesis-driven exploration efficiently, resulting in low analysis efficiency and difficulty in forming a cognitive loop, have become urgent technical problems to be solved. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and system for macro-micro interactive verification of the conclusions of large-sample exploratory experiments to address the above-mentioned technical problems.

[0006] This invention provides a method for macro-micro interactive verification of conclusions from large-sample exploratory experiments. The method is used to construct a three-layer linkage analysis framework of "macro-interaction-micro", which includes a macro conclusion layer, an interactive analysis layer, and a micro instance layer. The interactive analysis layer is used to connect the macro conclusion layer and the micro instance layer. The method includes four steps: automatic discovery and presentation of macro conclusions, interactive in-depth exploration and verification based on macro conclusions, macro-micro bidirectional linkage and sample selection, and concrete review and causal verification of micro instances. In the automatic discovery and presentation of the macro-level conclusions, the macro-level conclusion layer includes a causal discovery and association mining algorithm module. This module receives a large sample experimental raw dataset, performs calculations on the raw dataset, automatically identifies the most significant and universal causal paths or association rules, and automatically filters out typical samples strongly correlated with each conclusion. Based on these typical samples, a typical sample library for the conclusion set is constructed. The dataset contains at least multi-dimensional sample data, statistical analysis data, causal data, and debriefing report data. In the interactive in-depth exploration and verification based on macroscopic conclusions, the interactive analysis layer is triggered by the user selecting a target conclusion of interest from the conclusion set through the front-end interactive interface at the macro level. The multi-dimensional interactive visualization module of the interactive analysis layer calls the original dataset and preprocessed feature data, filters out a subset of samples that meet the attributes of the target conclusion through a data association algorithm, and highlights all sample data that meet the conclusion. Based on the highlighted sample data, the user "swipes" at least one sample data, and the front-end interactive interface automatically displays the associated microscopic corroborating data. The user discovers interactive effects and nonlinear relationships through a "hypothesis-verification" cycle of human-computer interaction. In the macro-micro bidirectional linkage and sample screening, in the interactive exploration based on the macro-based conclusions, the user selects a group of samples with specific behaviors, and the user triggers a linkage command to seamlessly transmit the ID of the characteristic sample to the micro instance layer. In the visualization review and causal verification of the micro-instance, the sample subset of the interactive analysis layer is received, and the micro-instance layer starts the correlation review module to replay the process of the sample. During the replay, the causal chain highlighting and tracing module is activated. Based on the conclusions verified by the sample, the causal chain highlighting and tracing module dynamically highlights the changes in data values ​​in relevant intermediate tables or the occurrence of relevant events at key nodes of the review timeline.

[0007] Optionally, the multi-dimensional sample data includes at least equipment operating parameters, environmental monitoring indicators, and user operation logs collected from the target system; the statistical analysis data includes at least a failure rate statistics table, performance indicator time series graph, and correlation analysis report generated based on the sample data; the causal data includes the correspondence between recorded events and their root causes; and the review report data includes historical records and effect evaluation data.

[0008] Optionally, the causal path or association rule is presented to the user through concise text or visual graphics.

[0009] Optionally, the visualization is a causal chain diagram.

[0010] Optionally, the multidimensional interactive visualization module is a parallel coordinate graph used to intuitively present high-dimensional data; the user performs an interactive "scanning" operation on the coordinate axis, and after the "scanning" operation, the conclusion is displayed in the multidimensional interactive visualization module.

[0011] Optionally, the interactive in-depth exploration and verification of the macroscopic conclusions also includes the user solidifying the heuristic conclusion screening conditions to trigger the conditional conclusion hypothesis and verification module. The conditional conclusion hypothesis and verification module is used to generate conclusions for the selected sample subset based on a preset structured template, and to feed back the "conditional conclusions" that are true only on the subset to the user.

[0012] Optionally, the samples of the specific behavior include samples that satisfy the conditional conclusion and perform optimally, or abnormal samples that do not meet the conclusion.

[0013] Optionally, the macro-micro bidirectional linkage and sample selection also include interactive exploration based on the interactive in-depth exploration and verification based on the macro conclusions. When the user reviews a sample at the micro level, a reverse linkage command is triggered, and the sample will be uniquely highlighted in the global view of the interactive analysis layer.

[0014] Optionally, the confirmed conclusions may include global conclusions or conditional conclusions.

[0015] Optionally, the visualization and causal verification of the micro-examples also include a single-sample conclusion presentation module that synchronously displays the transmission relationship of each variable in the causal chain over time using a time series graph.

[0016] Compared with the prior art, this application has the following main advantages: (1) A three-layer linkage analysis framework of "macro-interaction-micro" was constructed. This invention innovatively designed a three-layer coupled architecture consisting of a macro conclusion layer, an interactive analysis layer, and a micro instance layer. This framework no longer regards macro analysis and micro review as isolated steps, but integrates them into an organic whole. Through the interactive analysis layer as the core hub, it realizes the seamless flow of data between the three layers and the free switching of analytical perspectives, fundamentally solving the problem of macro-micro disconnect. (2) An interactive analysis layer with multidimensional interactive visualization as its core was designed for in-depth conclusion exploration and verification. As the core of this invention, the interactive analysis layer provides an innovative interactive analysis method. It uses visualization techniques such as parallel coordinate graphs to present high-dimensional data intuitively and allows analysts to filter data subsets in real time and dynamically through a "scanning" operation. More importantly, it can instantly recalculate and present "conditional conclusions" that are only true on the selected data subset, thereby constructing a rapid iterative closed loop of "hypothesis formulation-interactive verification", effectively revealing interactive effects and nonlinear relationships that traditional methods cannot discover; (3) A two-way linkage and seamless traceability mechanism for macro and micro data has been realized. This invention designs a data linkage mechanism that runs through a three-layer framework. This mechanism supports one-click drilling down of conclusions discovered at the macro level, highlighting the corresponding sample groups at the interaction layer for in-depth analysis, and seamlessly sending specific sample subsets selected at the interaction layer to the micro level for concrete review and verification. At the same time, it also supports one-click floating up of individual typical cases reviewed at the micro level, locating them in the global view of the interaction layer, thereby understanding the position of individual cases in the overall sample space. This two-way traceability capability completely breaks down the barriers between the macro and micro levels; (4) A concrete method for highlighting and verifying causal chains is proposed. At the micro-level, this invention proposes a mechanism that automatically highlights changes in relevant intermediate variables at key nodes on the timeline based on causal conclusions derived from the macro-level or interaction level during sample review. This method "deduces" the abstract causal path diagram into dynamic and visual process evidence, providing the most direct and robust verification of the conclusion's authenticity, and solving the technical problem that causal relationships are difficult to intuitively understand and trust. Attached Figure Description

[0017] Figure 1 This is a system framework diagram of one embodiment. Detailed Implementation

[0018] 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.

[0019] Existing large-sample analysis techniques suffer from shortcomings such as a disconnect between macro-statistics and micro-level examples, and an inability to uncover deep interactive conclusions. These shortcomings lead to technical problems such as a shallow understanding of conclusions, difficulties in verification, and low analytical efficiency. The core solution to these problems lies in the technical solution proposed in this invention. This solution constructs a three-layered, interconnected analytical framework of "macro-interactive-micro," using an innovative interactive analysis layer as the hub. This enables bidirectional traceability of macro and micro data and in-depth verification of conclusions. These innovations significantly improve the depth of conclusion discovery, analytical efficiency, and the credibility of the conclusions.

[0020] The core technical idea of ​​this invention is to construct a three-layer linkage analysis framework consisting of a macroscopic conclusion layer, an interactive analysis layer, and a microscopic instance layer, with the interactive analysis layer as the connecting hub, to realize a complete analysis loop from the automatic discovery of global conclusions to interactive in-depth exploration, and then to the concrete verification of specific instances, thereby solving the problems of macro-micro disconnection, superficial conclusion discovery, and low analysis efficiency in the existing technology.

[0021] The technical problem of this invention is solved by the following technical solution: Step 1: Automatic discovery and presentation of macro-level conclusions.

[0022] The system first receives a large-sample experimental dataset as input, containing multi-dimensional factors, intermediate variables, and indicators. The causal discovery and association mining algorithm module within the macro-level conclusion layer performs calculations on the entire dataset, automatically identifying the statistically most significant and universal causal paths or association rules. These conclusions are presented to the user in concise text and visual graphics (such as causal chain diagrams). Simultaneously, the system automatically selects typical samples strongly correlated with each conclusion, constructing a typical sample library.

[0023] In one embodiment of the present invention, this step can quickly provide users with a global understanding of the system's operational conclusions, effectively reducing the cognitive load when facing massive amounts of data, and providing a high-quality analytical starting point for subsequent in-depth exploration.

[0024] In one embodiment of the present invention, the dataset used in this step includes multi-dimensional sample data, statistical analysis data, causal data, and debriefing report data. Specifically, the multi-dimensional sample data may be equipment operating parameters, environmental monitoring indicators, and user operation logs collected from the target system; the statistical analysis data may be a failure rate statistics table, performance indicator time series graph, and correlation analysis report generated based on the sample data; the causal data may be records of the correspondence between events and their root causes; and the debriefing report data may be data containing historical records and effect evaluations.

[0025] Step 2: Interactive in-depth exploration and verification based on macroscopic conclusions.

[0026] After a user selects a conclusion of interest at the macro level, the system triggers the interactive analysis layer. All sample data that match this conclusion are highlighted in the multi-dimensional interactive visualization module, for example, through a parallel coordinate graph. Users can interactively "swipe" across any dimension to define a specific data range. For example, users can swipe along the x-axis and / or y-axis of a two-dimensional coordinate system to select intervals where "Factor B" is "low" to observe whether the initial conclusion still holds true under this premise. After the "swipe" operation, the sample data that match the conclusion are displayed through the multi-dimensional interactive visualization module. Simultaneously, users can solidify heuristic conclusion selection criteria, triggering the conditional conclusion hypothesis and verification module. This module generates conclusions for only the selected sample subset based on a preset structured template and provides feedback to the user on "conditional conclusions" that hold true only on that subset.

[0027] In one embodiment of this invention, this step is the core of the invention. Through the "hypothesis-verification" cycle of human-computer interaction, interactive effects and nonlinear relationships that traditional global statistics cannot reveal can be efficiently discovered, greatly improving the depth of conclusion discovery. For example, it can be discovered that "factor A can only significantly improve index C when factor B is low."

[0028] Step 3: Macro-micro bidirectional linkage and sample screening.

[0029] In the second step of interactive exploration, the user selects a set of samples with specific behaviors, such as "samples that meet the conditions and perform optimally" or "abnormal samples that do not meet the conclusions." The user can trigger a linkage command, and the system seamlessly transmits the IDs of this specific set of samples to the micro-instance layer. Conversely, when the user reviews a sample at the micro-level, they can also trigger a reverse linkage command, and that sample will be uniquely highlighted in the global view of the interactive analysis layer.

[0030] In one embodiment of the present invention, this step completely breaks down the barrier between macroscopic analysis and microscopic verification through an innovative two-way linkage mechanism, allowing the analytical perspective to be switched freely and seamlessly between different granularities, thus forming a smooth analytical closed loop.

[0031] Step 4: Concretization and causal verification of micro-level examples.

[0032] Upon receiving a subset of samples from the interactive analysis layer, the micro-instance layer initiates the correlational debriefing and replay module to replay the process of the samples. During the replay, the causal chain highlighting and tracing module is activated. Based on the conclusions confirmed by the sample, which can be global or conditional, it dynamically highlights the numerical changes of relevant intermediate variables or the occurrence of relevant events at key nodes on the debriefing timeline. Simultaneously, the single-sample conclusion presentation module synchronously displays the transmission relationships of each variable in the causal chain over time in the form of time series graphs, etc.

[0033] In one embodiment of the present invention, this step "translates" abstract and potentially difficult-to-understand causal conclusions into dynamic and observable concrete processes, providing the most robust and intuitive evidence for the authenticity of the conclusions, and greatly enhancing the interpretability and credibility of the conclusions.

[0034] The innovative points of this invention can be summarized as follows: 1. A three-tiered, interconnected analytical framework of "macro-interaction-micro" was constructed. This invention innovatively designs a three-tiered coupled architecture consisting of a macro-level conclusion layer, an interactive analysis layer, and a micro-level instance layer. This framework no longer treats macro-level analysis and micro-level review as isolated steps, but integrates them into an organic whole. Through the interactive analysis layer, which serves as the core hub, seamless data flow and free switching of analytical perspectives are achieved between the three levels, fundamentally solving the problem of the disconnect between macro and micro perspectives.

[0035] 2. An interactive analysis layer centered on multidimensional interactive visualization was designed for in-depth conclusion exploration and verification. As the core of this invention, the interactive analysis layer provides an innovative interactive analysis method. It employs visualization techniques such as parallel coordinate graphs to intuitively present high-dimensional data and allows analysts to dynamically filter data subsets in real time through a "scanning" operation. More importantly, based on the filtered data subset, it can instantly recalculate and present "conditional conclusions" that hold true only on that subset, thereby constructing a rapid iterative closed loop of "hypothesis formulation - interactive verification," effectively revealing interactive effects and nonlinear relationships that traditional methods cannot discover.

[0036] 3. A two-way linkage and seamless traceability mechanism for macro and micro data has been achieved. This invention designs a data linkage mechanism that runs through a three-layer framework. This mechanism supports one-click drill-down of conclusions discovered at the macro level, highlighting the corresponding sample groups for in-depth analysis at the interaction layer, and seamlessly sending specific sample subsets selected at the interaction layer to the micro level for concrete review and verification. Simultaneously, it also supports one-click uploading of individual typical cases reviewed at the micro level, locating them in the global view of the interaction layer, thereby understanding the position of individual cases within the overall sample space. This two-way traceability capability completely breaks down the barriers between macro and micro levels.

[0037] 4. A concrete method for highlighting and verifying causal chains is proposed. At the micro-level, this invention proposes a mechanism that automatically highlights changes in relevant intermediate variables at key nodes on the timeline based on causal conclusions derived from upper-level analysis (macro or interactive layers) during sample review. This method "deduces" the abstract causal path diagram into dynamic, visualized process evidence, providing the most direct and robust verification of the conclusion's authenticity, and solving the technical challenge of making causal relationships difficult to intuitively understand and trust.

[0038] In summary, the present invention achieves the following beneficial effects: 1. This invention achieves interactive verification between macroscopic conclusions and microscopic examples, significantly improving the interpretability and credibility of the conclusions. It breaks down the barriers between the macro and micro levels, constructing a cognitive loop from "knowing what" to "knowing why."

[0039] 2. Significantly enhances the ability to discover hidden conclusions and greatly increases the depth of analysis. Through the interactive analysis layer, it can effectively reveal conditional conclusions and interactive effects that are difficult to discover using traditional methods, making the analysis more accurate and in-depth.

[0040] 3. By transforming the analysis process from passive reception to active exploration, the efficiency and quality of analysis are significantly improved. An efficient hypothesis-validation cycle empowers analysts with unprecedented data insights, enabling them to rapidly iterate and validate ideas.

[0041] 4. It provides a more robust basis for decision-making. Through concrete causal tracing, decision-makers can not only see the statistical results, but also understand the complete mechanism behind the results, thereby making more reliable and robust decisions.

[0042] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for macro- and micro-level cross-verification of conclusions from large-sample exploratory experiments, characterized in that, The method described above is used to construct a three-layer linkage analysis framework of "macro-interaction-micro". The three-layer linkage analysis framework includes a macro conclusion layer, an interactive analysis layer, and a micro instance layer. The interactive analysis layer is used to connect the macro conclusion layer and the micro instance layer. The method includes four steps: automatic discovery and presentation of macro conclusions, interactive in-depth exploration and verification based on macro conclusions, macro-micro bidirectional linkage and sample screening, and concrete review and causal verification of micro instances. In the automatic discovery and presentation of the macro-level conclusions, the macro-level conclusion layer includes a causal discovery and association mining algorithm module. This module receives a large sample experimental raw dataset, performs calculations on the raw dataset, automatically identifies the most significant and universal causal paths or association rules, and automatically filters out typical samples strongly correlated with each conclusion. Based on these typical samples, a typical sample library for the conclusion set is constructed. The dataset contains at least multi-dimensional sample data, statistical analysis data, causal data, and debriefing report data. In the interactive in-depth exploration and verification based on macroscopic conclusions, the interactive analysis layer is triggered by a target conclusion selected by the user from the conclusion set through the front-end interactive interface at the macro level. The multi-dimensional interactive visualization module of the interactive analysis layer calls the original dataset and preprocessed feature data, filters out a subset of samples that meet the attributes of the target conclusion through a data association algorithm, and highlights all sample data that meet the conclusion. Based on the highlighted sample data, the user "swipes" at least one sample data, and the front-end interactive interface automatically displays the associated microscopic corroborating data. The user discovers interactive effects and nonlinear relationships through a "hypothesis-verification" cycle of human-computer interaction. In the macro-micro bidirectional linkage and sample screening, in the interactive exploration based on the macro-based conclusions, the user selects a group of samples with specific behaviors, and the user triggers a linkage command to seamlessly transmit the ID of the feature sample to the micro instance layer. In the visualization review and causal verification of the micro-instance, the sample subset of the interactive analysis layer is received, and the micro-instance layer starts the correlation review module to replay the process of the sample. During the replay, the causal chain highlighting and tracing module is activated. Based on the conclusions verified by the sample, the causal chain highlighting and tracing module dynamically highlights the changes in data values ​​in relevant intermediate tables or the occurrence of relevant events at key nodes of the review timeline.

2. The macro-micro interactive verification method according to claim 1, characterized in that, The multi-dimensional sample data includes at least the equipment operating parameters, environmental monitoring indicators, and user operation logs collected from the target system; the statistical analysis data includes at least the failure rate statistics table, performance indicator time series diagram, and correlation analysis report generated based on the sample data; the causal data includes the correspondence between recorded events and their root causes; and the review report data includes historical records and effect evaluation data.

3. The macro-micro interactive verification method according to claim 1, characterized in that, The causal paths or association rules are presented to the user through concise text or visual graphics.

4. The macro-micro interactive verification method according to claim 1 or 3, characterized in that, The visualization is a causal chain diagram.

5. The macro-micro interactive verification method according to claim 1, characterized in that, The multidimensional interactive visualization module is a parallel coordinate graph used to intuitively present high-dimensional data; the user performs an interactive "scanning" operation on the coordinate axis, and after the "scanning" operation, the conclusion is displayed in the multidimensional interactive visualization module.

6. The macro-micro interactive verification method according to claim 1 or 5, characterized in that, The interactive in-depth exploration and verification of the macroscopic conclusions also includes the user solidifying the heuristic conclusion screening conditions to trigger the conditional conclusion hypothesis and verification module. The conditional conclusion hypothesis and verification module is used to generate conclusions for the selected sample subset based on a preset structured template, and to feed back the "conditional conclusions" that are true only on the subset to the user.

7. The macro-micro interactive verification method according to claim 1, characterized in that, The samples of the specific behavior include samples that satisfy the conditional conclusion and perform optimally, or abnormal samples that do not meet the conclusion.

8. The macro-micro interactive verification method according to claim 1 or 7, characterized in that, The macro-micro bidirectional linkage and sample selection also include interactive exploration based on the interactive in-depth exploration and verification of the macro conclusions. When the user reviews a sample at the micro level, a reverse linkage command is triggered, and the sample will be uniquely highlighted in the global view of the interactive analysis layer.

9. The macro-micro interactive verification method according to claim 1, characterized in that, The conclusions confirmed include global conclusions or conditional conclusions.

10. The macro-micro interactive verification method according to claim 1 or 9, characterized in that, The visualization and causal verification of the micro-examples also include a single-sample conclusion presentation module that uses time series graphs to synchronously display the transmission relationship of each variable in the causal chain over time.

Images