Multidimensional data analysis method and device based on holography

Abstract

The invention relates to the technical field of data analysis, and discloses a holographic multi-dimensional data analysis method, which comprises the following steps: acquiring corresponding to-be-analyzed multi-dimensional data, preprocessing the to-be-analyzed multi-dimensional data, inputting the preprocessed to-be-analyzed multi-dimensional data into a multi-dimensional relationship model for analysis, and acquiring a corresponding preliminary analysis result; determining a corresponding target user, displaying a preliminary analysis result to the user in a three-dimensional image through real-time dynamic updating based on historical behaviors, a current context and a holographic projection technology, and updating a visual interface corresponding to the preliminary analysis result when the to-be-analyzed multi-dimensional data changes; and obtaining corresponding feedback data, judging whether an optimization mechanism is triggered or not based on the feedback data, and if yes, performing optimization through a preset optimization mechanism based on the feedback data and the preliminary analysis result to obtain a corresponding target analysis result. According to the method and the device, multi-dimensional data analysis can be efficiently and accurately performed.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and in particular to a method and device for holographic multidimensional data analysis. Background Art

[0002] With the advent of the big data era, the amount of data is growing exponentially. Traditional one-dimensional or two-dimensional data analysis methods can no longer meet the needs of complex scenarios. Especially when it is necessary to consider space, time and multiple attribute factors at the same time, how to efficiently and accurately process these multi-dimensional information has become an urgent problem to be solved. Although some existing solutions try to improve processing capabilities by increasing computing resources, they often ignore the inherent connections between data, resulting in incomplete or inefficient analysis results.

[0003] In addition, in some fields such as medical diagnosis and environmental monitoring, there are extremely high requirements for the authenticity and intuitiveness of data, and traditional display methods are difficult to achieve the expected results. Therefore, it is particularly important to develop a data analysis method that can comprehensively consider multi-dimensional information and present it in a more intuitive form.

[0004] From the above, we can see that how to achieve efficient and accurate analysis of multi-dimensional data remains to be solved. Summary of the invention

[0005] In order to achieve efficient and accurate analysis of multi-dimensional data, the present application provides a holographic multi-dimensional data analysis method and device.

[0006] In the first aspect, the present application provides a holographic multi-dimensional data analysis method, which adopts the following technical solution:

[0007] A method for analyzing multidimensional data based on holography comprises: obtaining corresponding multidimensional data to be analyzed, preprocessing the multidimensional data to be analyzed, retrieving a pre-trained multidimensional relationship model, inputting the preprocessed multidimensional data to be analyzed into the multidimensional relationship model for analysis, and obtaining corresponding preliminary analysis results; wherein the preprocessing comprises adopting an adaptive data cleaning algorithm and a dynamic anomaly detection mechanism, and the multidimensional relationship model integrates a graph neural network, a time series analysis, and a hybrid model; determining a corresponding target user, retrieving the historical behavior and current context corresponding to the target user, and displaying the preliminary analysis results to the user in a three-dimensional stereoscopic image through real-time dynamic updating based on the historical behavior, the current context, and the holographic projection technology; when the multidimensional data to be analyzed changes, updating the visualization interface corresponding to the preliminary analysis results through an incremental update mechanism; obtaining corresponding feedback data during the interaction between the target user and the visualization interface, judging whether an optimization mechanism is triggered based on the feedback data, and if triggered, optimizing through a preset optimization mechanism based on the feedback data and the preliminary analysis results, and obtaining the corresponding target analysis results, wherein the feedback data comprises an evaluation of the visualization interface results, a record of the interaction behavior, and improvement suggestions proposed by the user.

[0008] By adopting the above technical solutions, multidimensional data is preprocessed through adaptive data cleaning and dynamic anomaly detection, and efficient and accurate analysis is performed using a multidimensional relationship model that integrates graph neural networks, time series analysis, and hybrid models. Combining the historical behavior and current context of the target user, the system uses holographic projection technology to dynamically display the preliminary analysis results in three-dimensional stereoscopic images in real time, and incrementally updates them according to data changes. In addition, the system collects feedback data from users' interactions with the visualization interface, triggering optimization mechanisms to further improve the quality of the analysis results, ensuring that the analysis process is both efficient and accurate, while providing a highly personalized user experience.

[0009] Optionally, after retrieving the historical behavior and current context corresponding to the target user, the method also includes: obtaining the target user's preference settings, extracting corresponding key features for the preference settings, the historical behavior and the current context, wherein the key features include the type of data most frequently accessed by the target user, the most frequently used functions, and the preferred time periods; adjusting the display priority of each data in the preliminary analysis results based on the key features; and displaying the preliminary analysis results to the user in a three-dimensional stereoscopic image based on the adjusted display priority.

[0010] By adopting the above technical solution, key features (such as the most frequently accessed data types, the most frequently used functions, and preferred time periods) are extracted by comprehensively considering the target user's preference settings, historical behaviors, and current context, and the display priority of each data in the preliminary analysis results is dynamically adjusted based on these features, so as to display the analysis results through three-dimensional stereo images in a manner that better meets user needs.

[0011] Optionally, the method also includes: determining different perspective dimensions that need to be analyzed based on the multidimensional data to be analyzed, wherein the different perspective dimensions include time dimension, space dimension, and category dimension; performing multi-perspective analysis on the multidimensional data to be analyzed based on the different perspective dimensions, and obtaining preliminary analysis results corresponding to each perspective, wherein the multi-perspective analysis includes time dimension analysis, space dimension analysis, category dimension analysis, and other dimension analysis; integrating the preliminary analysis results from different perspectives to generate a corresponding multi-perspective analysis report, and presenting the multi-perspective analysis results to the user in the form of a three-dimensional image, wherein the multi-perspective analysis report includes key findings and conclusions for each perspective.

[0012] By adopting the above technical solution, multi-dimensional data is analyzed from multiple perspectives, covering multiple dimensions such as time, space, and category. The preliminary analysis results from different perspectives are integrated to generate a comprehensive multi-perspective analysis report, which is presented to users in the form of three-dimensional images, thereby providing deeper and more comprehensive data insights and helping users understand and interpret data from multiple perspectives.

[0013] Optionally, after obtaining the preliminary analysis results, the method also includes: calling pre-set expert rules, wherein the expert rules include the experience and knowledge of domain experts; evaluating the preliminary analysis results according to the rules defined by the expert rules through an inference engine to obtain corresponding evaluation results; comparing the preliminary analysis results with the evaluation results, and if the comparison results are inconsistent, adjusting the preliminary analysis results.

[0014] By adopting the above technical solution, after obtaining the preliminary analysis results, the results are evaluated using pre-set expert rules and reasoning engines, and adjustments are made when the preliminary analysis results are inconsistent with the evaluation results, thereby integrating the experience and knowledge of experts in the field and improving the accuracy and reliability of the analysis results.

[0015] Optionally, after obtaining the preliminary analysis results, the method also includes: calling a corresponding anomaly detection algorithm, using the anomaly detection algorithm to detect potential outliers in the preliminary analysis results, generating a corresponding anomaly detection report, the anomaly detection report marking all detected outliers; obtaining abnormal feedback from the target user on the anomaly detection report, wherein the abnormal feedback includes the user's confirmation of the outliers, correction suggestions, etc.; combining the anomaly detection report with the abnormal feedback, and re-evaluating the outliers to determine whether there are omissions or misjudgments.

[0016] By adopting the above technical solution, potential outliers are identified through anomaly detection algorithms and reports are generated, and re-evaluation is performed in combination with the feedback from target users on the outliers to ensure the accuracy and completeness of anomaly detection and reduce omissions and misjudgments.

[0017] Optionally, the method also includes: performing uncertainty assessment on each predicted value in the preliminary analysis results to generate a corresponding confidence level; sorting based on all confidence levels, and highlighting the results corresponding to the results with the highest confidence levels when displayed in a visual interface; and generating a corresponding explanation report for the results corresponding to the results with the lowest confidence levels, wherein the explanation report includes possible causes, influencing factors, and recommended further verification methods.

[0018] By adopting the above technical solution, uncertainty assessment is performed on each predicted value in the preliminary analysis results and confidence is generated. High-confidence results are highlighted in the visualization interface according to the confidence ranking, and detailed explanation reports are generated for low-confidence results, thereby helping users better understand the reliability and potential uncertainties of the analysis results.

[0019] In the second aspect, the present application provides a holographic multi-dimensional data analysis device, which adopts the following technical solution:

[0020] A holographic multidimensional data analysis device, comprising:

[0021] A preliminary analysis result acquisition module acquires the corresponding multidimensional data to be analyzed, preprocesses the multidimensional data to be analyzed, retrieves a pre-trained multidimensional relationship model, and inputs the preprocessed multidimensional data to be analyzed into the multidimensional relationship model for analysis, so as to obtain the corresponding preliminary analysis results; wherein the preprocessing includes adopting an adaptive data cleaning algorithm and a dynamic anomaly detection mechanism, and the multidimensional relationship model integrates a graph neural network, time series analysis, and a hybrid model;

[0022] The preliminary analysis result display module determines the corresponding target user, retrieves the historical behavior and current context corresponding to the target user, and uses the preliminary analysis result to display the three-dimensional image to the user through real-time dynamic update based on the historical behavior, current context and holographic projection technology. When the multi-dimensional data to be analyzed changes, the visualization interface corresponding to the preliminary analysis result is updated through the incremental update mechanism;

[0023] The target analysis result acquisition module obtains the corresponding feedback data during the interaction between the target user and the visualization interface, and determines whether to trigger the optimization mechanism based on the feedback data. If triggered, optimization is performed through a preset optimization mechanism based on the feedback data and the preliminary analysis results to obtain the corresponding target analysis results, wherein the feedback data includes the evaluation of the visualization interface results, the interaction behavior records, and the improvement suggestions proposed by the user.

[0024] In a third aspect, the present application provides a holographic multi-dimensional data analysis method, which adopts the following technical solution:

[0025] A method for analyzing multidimensional data based on holography comprises a processor in which a program of any one of the above-mentioned methods for analyzing multidimensional data based on holography is running.

[0026] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:

[0027] A storage medium stores a program based on any one of the above-mentioned holographic multi-dimensional data analysis methods.

[0028] In summary, the present application includes at least one of the following beneficial technical effects:

[0029] 1. The method based on holographic multidimensional data analysis pre-processes multidimensional data through adaptive data cleaning and dynamic anomaly detection, and uses a multidimensional relationship model that integrates graph neural networks, time series analysis and hybrid models for efficient and accurate analysis. Combined with the historical behavior, current context and preference settings of the target user, the system can dynamically adjust the display priority of each data in the preliminary analysis results, and display the three-dimensional stereoscopic analysis results in real time and dynamically through holographic projection technology. In addition, the system introduces expert rules and reasoning engines to evaluate and adjust the preliminary analysis results to ensure the accuracy and reliability of the analysis results; at the same time, through anomaly detection algorithms and uncertainty quantification technology, the quality and credibility of the analysis results are further improved.

[0030] 2. Through multi-perspective analysis technology, this method can comprehensively analyze data from multiple dimensions such as time, space, and category, generate a comprehensive multi-perspective analysis report, and present it to users in the form of three-dimensional images, providing deeper and more comprehensive data insights. The system also collects feedback data from users' interactions with the visualization interface, triggering optimization mechanisms to further improve the quality of analysis results, ensuring that the analysis process is both efficient and accurate, while providing a highly personalized user experience. These technical means work together to make multi-dimensional data analysis not only efficient and accurate, but also more in line with the actual needs of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The figure is a flow chart of a holographic multi-dimensional data analysis method according to an exemplary embodiment.

[0032] Figure 2 It is a structural block diagram of a holographic multi-dimensional data analysis device according to an exemplary embodiment. DETAILED DESCRIPTION

[0033] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings.

[0034] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0035] The present application embodiment discloses a method for analyzing multidimensional data based on holography, referring to Figure 1 ,include:

[0036] S100, obtaining corresponding multidimensional data to be analyzed, preprocessing the multidimensional data to be analyzed, retrieving a pre-trained multidimensional relationship model, inputting the preprocessed multidimensional data to be analyzed into the multidimensional relationship model for analysis, and obtaining corresponding preliminary analysis results.

[0037] Among them, multidimensional data is collected from various data sources (such as databases, sensors, log files, etc.), which may include time series data, structured data, unstructured data, etc., and data from different sources are integrated to ensure the consistency of data formats, which may involve operations such as data conversion and data alignment, and preliminary checks on data quality to ensure data integrity and consistency. For example, check whether there are missing values, duplicate records, or inconsistent data.

[0038] When preprocessing the multidimensional data to be analyzed, adaptive data cleaning algorithms and dynamic anomaly detection mechanisms are used. Specifically:

[0039] Adaptive data cleaning algorithm:

[0040] 1. Missing value processing: Use interpolation methods (such as linear interpolation, KNN interpolation), statistical methods (such as mean, median filling), or other machine learning methods (such as predictive models) to fill missing values.

[0041] 2. Outlier detection and correction: Use statistical methods (such as Z-score, IQR) or machine learning methods (such as isolation forest, local outlier factor) to detect outliers, and choose to retain, correct or delete these outliers according to the specific situation.

[0042] 3. Data standardization / normalization: Standardize or normalize the data to make it meet the requirements of model input. Common methods include Z-score standardization, Min-Max normalization, etc.

[0043] 4. Data denoising: Use smoothing techniques (such as moving average, exponential smoothing) to reduce noise in the data and improve data quality.

[0044] Dynamic anomaly detection mechanism:

[0045] 1. Real-time monitoring: Monitor new data points in the data stream in real time and detect outliers.

[0046] 2. Dynamic update: Continuously update the anomaly detection model based on new data to ensure that it can adapt to changes in data.

[0047] 3. Feedback mechanism: Combine user feedback and historical data to adjust anomaly detection thresholds and rules to improve detection accuracy.

[0048] Then, by selecting a multidimensional relationship model suitable for the current task, the multidimensional relationship model integrates graph neural network (GNN), time series analysis (such as LSTM, ARIMA) and hybrid models (such as ensemble learning methods). Load the pre-trained multidimensional relationship model from the storage medium. The multidimensional relationship model has been trained with a large amount of data and can capture the complex dependencies and time series characteristics between data. It is also necessary to set the hyperparameters and other configurations of the multidimensional relationship model to ensure that the multidimensional relationship model can run efficiently.

[0049] Convert the preprocessed data into the input format required by the model, which may include operations such as feature engineering and data segmentation, and input the data into the multidimensional relationship model, which analyzes the data and generates preliminary analysis results. Specific analysis tasks may include prediction, classification, clustering, etc. The preliminary analysis results include predicted values, classification labels, clustering results, etc.

[0050] S110, determine the corresponding target user, retrieve the historical behavior and current context corresponding to the target user, and display the preliminary analysis results to the user in the form of three-dimensional images through real-time dynamic updates based on the historical behavior, current context and holographic projection technology. When the multidimensional data to be analyzed changes, the visualization interface corresponding to the preliminary analysis results is updated through the incremental update mechanism.

[0051] The user currently using the system is determined through user login information, device identifier or other identity authentication mechanisms, and then the relevant information of the target user is retrieved from the user database or user management system, including basic information (such as name, role), historical behavior data (such as past query records, operation records) and preference settings (such as interface layout preferences, color preferences).

[0052] Collect historical behavior data of target users from data sources such as user activity logs and operation records. These data may include target users' past query records, click behaviors, browsing records, etc. Then clean and organize the collected historical behavior data to extract key features, such as the most frequently accessed data types, the most frequently used functions, and preferred time periods.

[0053] At the same time, the user's current context information is obtained, such as geographic location, device type, time, environmental conditions, etc. The context information is combined with historical behavior data to generate comprehensive user context information. By combining the user's historical behavior and current context, the analysis results are closer to the user's actual needs and situations, improving the personalization and accuracy of the analysis.

[0054] Holographic projection technology converts the preliminary analysis results into a format suitable for holographic projection display, such as three-dimensional point cloud, mesh model, etc., and it is necessary to design the layout and style of the three-dimensional stereo image to ensure the clear presentation of information, for example, using different colors, shapes and sizes to represent different types of data.

[0055] Holographic projection technology can also be used to render three-dimensional images in real time and dynamically adjust the displayed content based on the user's behavior and context. In addition, user interaction is also possible, including:

[0056] 1. Gesture control: allows users to interact with the view by rotating, zooming and panning it through gestures;

[0057] 2. Voice control: supports voice commands, users can operate through voice commands, such as "enlarge", "rotate", etc.;

[0058] 3. Touch operation: On devices that support touch screens, users can interact by touching the screen.

[0059] When the multidimensional data to be analyzed changes, it is necessary to continuously monitor the changes in the data source, detect the inflow of new data, and detect whether the data has changed through data version control or hash verification. It should be pointed out here that only the changed data is preprocessed and analyzed. If the data changes greatly, it may be necessary to retrain or fine-tune the model to adapt to the new data model. At the same time, based on the changed data, the analysis model is re-run to generate new analysis results, and the new analysis results are updated to the visualization interface in real time to ensure that users see the latest data and analysis results.

[0060] Through these steps, not only the accuracy and efficiency of data analysis are improved, but also a highly personalized user experience is provided, ensuring that users can understand complex data analysis results in a timely and intuitive manner.

[0061] S120, obtaining corresponding feedback data during the interaction between the target user and the visualization interface, and determining whether to trigger the optimization mechanism based on the feedback data. If triggered, optimizing is performed through a preset optimization mechanism based on the feedback data and the preliminary analysis results to obtain the corresponding target analysis results.

[0062] Among them, collect users' evaluations of the visualization interface results, such as satisfaction scores, comments and suggestions, and record users' interactive behaviors, such as clicks, drags, zooms, etc. These behaviors can reflect the user's interest in specific data or views. It is also necessary to collect improvement suggestions put forward by users, including suggestions on analysis results, display methods or functions. The above feedback data is classified into different types, such as positive evaluations, negative evaluations, specific suggestions, etc.

[0063] In addition, natural language processing (NLP) technology is used to conduct sentiment analysis on user reviews to determine users' emotional tendencies (such as satisfaction, dissatisfaction); analyze users' interactive behavior records to identify their behavior patterns and preferences; and extract key information from improvement suggestions put forward by users, such as specific improvement points and suggested content.

[0064] By setting the threshold for triggering the optimization mechanism, such as the proportion of negative reviews exceeding a certain threshold, users frequently making similar suggestions, etc., the preset rules are used to determine whether the conditions for triggering the optimization mechanism are met. For example, if multiple users make the same improvement suggestions, or the proportion of negative reviews exceeds the set threshold, the optimization mechanism is triggered.

[0065] If triggered, the optimization mechanism is selected first: 1. Parameter adjustment: adjust the model's hyperparameters, such as learning rate, regularization coefficient, etc., based on user feedback; 2. Model retraining: If user feedback indicates that the model has large deviations or deficiencies, retraining the model may require more data or a more complex model structure; 3. Rule update: update business rules or logic based on the user's specific suggestions to better meet user needs; 4. Interface optimization: adjust the interface layout, color, font, etc. based on user feedback on the interface to improve user experience.

[0066] Based on user feedback and preliminary analysis results, targeted optimization is performed through the preset optimization mechanism. For example, the optimized model can be used to re-analyze the data to generate new analysis results, which are then compared with the previous preliminary analysis results to ensure consistency. The optimized target analysis results are then output for visualization.

[0067] Finally, the optimized target analysis results are displayed to users in the form of three-dimensional images through holographic projection technology, and users can be informed through system notifications or emails that the analysis results have been updated, and are invited to view and provide feedback.

[0068] Based on the execution steps from S100 to S120, the data is deeply analyzed using advanced machine learning models (such as neural networks, time series analysis, and hybrid models), complex patterns and trends in the data are extracted, and preliminary analysis results are generated; by combining the user's historical behavior and current context, the analysis results are closer to the user's actual needs; the analysis results are displayed through intuitive three-dimensional images to improve the user's understanding and interactive experience. The display content is updated in real time and the display mode is adjusted according to the user's behavior and context to enhance the user experience; at the same time, the user's satisfaction and needs for the analysis results are understood to provide a basis for subsequent optimization, and the system can respond to user needs in a timely manner by collecting user evaluations, interactive behavior records, and improvement suggestions during the interaction process; the analysis results are adjusted in a timely manner according to user feedback to improve user experience and satisfaction. If the user feedback contains important improvement suggestions or negative comments, the optimization mechanism is triggered to ensure that the system can be continuously improved and perfected; the quality and accuracy of the analysis results are improved through the preset optimization mechanism (such as parameter adjustment, model retraining, rule update, etc.) to better meet user needs.

[0069] After retrieving the historical behavior and current context corresponding to the target user, the method further includes:

[0070] S111, obtaining the target user's preference settings, and extracting corresponding key features of the preference settings, historical behaviors, and current context.

[0071] Among them, the target user's preference settings are obtained from the user database or user profile. These preference settings may include interface layout, color preference, font size, display mode, etc.; then the preference setting data is parsed to extract specific configuration information. For example, the user may prefer data display in the form of a chart, or like a specific color theme.

[0072] Key features include the data types most frequently accessed by target users, the functions most frequently used, and the preferred time periods. Key features extracted from preference settings, historical behaviors, and current contexts are integrated to form a comprehensive user feature set. Specifically:

[0073] 1. Preference features: Extract key features from user preferences, such as the user's preferred data display method (chart, table), color theme, font size, etc.

[0074] 2. Historical behavior features: Extract key features from the user's historical behavior data, such as the most frequently accessed data types (such as sales data, inventory data), the most frequently used functions (such as query, export), and preferred time periods (such as weekdays, weekends).

[0075] 3. Current context features: Extract key features from the user’s current context, such as geographic location, device type (mobile phone, computer), time (daytime, nighttime), etc.

[0076] By extracting and integrating key features, we can better understand user behavior patterns and preferences, providing a basis for subsequent adjustment of the display priority of analysis results.

[0077] S112, adjusting the display priority of each data in the preliminary analysis result based on the key features.

[0078] It should be noted here that in the embodiment of the present application, the priority rule is defined as:

[0079] Based on preference: Determine which data types or functions should be displayed first based on the user's preferences. For example, if the user prefers charts, chart data will be displayed first.

[0080] Based on historical behavior: Prioritize the data types and features that users access most frequently based on their historical behavior. For example, if a user frequently views sales data, sales data will be displayed first.

[0081] Based on the current context: Dynamically adjust the display priority based on the user's current context. For example, if the user uses the system on a weekday, work-related data may be displayed first; if the user uses the system on a mobile device, data views suitable for small screens may be displayed first.

[0082] Then, by combining the above rules, the display priority of each data is calculated, and the weighted scoring method or other algorithms can be used to determine the final priority. Priority rule definition:

[0083] Based on preferences: Determine which data types or functions should be displayed first based on the user's preferences. For example, if the user prefers charts, then chart data will be displayed first.

[0084] Based on historical behavior: Prioritize the data types and features that users access most frequently based on their historical behavior. For example, if a user frequently views sales data, sales data will be displayed first.

[0085] Based on the current context: Dynamically adjust the display priority based on the user's current context. For example, if the user uses the system during the workday, work-related data may be displayed first; if the user uses the system on a mobile device, data views suitable for small screens may be displayed first.

[0086] Priority calculation: Combine the above rules to calculate the display priority of each data item. You can use weighted scoring or other algorithms to determine the final priority; dynamically adjust the display priority of each data item in the analysis results based on the user's preferences, historical behavior, and current context, so as to ensure that the user sees the most relevant and important information.

[0087] S113: Displaying the preliminary analysis result to the user in a three-dimensional image based on the adjusted display priority.

[0088] Among them, the data in the preliminary analysis results are reorganized according to the adjusted display priority to prepare for the three-dimensional stereoscopic image display of holographic projection technology. It is necessary to design the layout and style of the three-dimensional stereoscopic image to ensure that high-priority data is more prominent visually, for example, using different colors, sizes and positions to represent data of different priorities.

[0089] Through the above steps S111 to S113, not only the accuracy and reliability of the analysis results are improved, but also a highly personalized user experience is provided, ensuring that users can quickly and intuitively understand complex data analysis results.

[0090] The method also includes:

[0091] S114, determining different perspective dimensions that need to be analyzed based on the multidimensional data to be analyzed.

[0092] Among them, different perspective dimensions that need to be analyzed are determined based on data characteristics and business needs. Common dimensions include:

[0093] Time dimension: Analyze the changing trends of data over time, identify periodic patterns, seasonal changes, etc.

[0094] Spatial dimension: Analyze the distribution of data in different geographical locations and identify regional differences and hot spots.

[0095] Category dimension: Analyze the data characteristics of different categories and identify the differences and associations between categories.

[0096] Other dimensions: Other dimensions can be introduced according to specific needs, such as user behavior dimension, product performance dimension, etc.

[0097] S115, performing multi-perspective analysis on the multi-dimensional data to be analyzed based on different perspective dimensions, and obtaining preliminary analysis results corresponding to each perspective.

[0098] Specifically, time dimension analysis: extract time series data, arrange them in chronological order, use time series analysis methods (such as ARIMA, LSTM) to identify trends, periodicity and seasonality of the data, detect outliers in the time series, identify abnormal events or emergencies, and then predict future trends based on historical data to provide decision support.

[0099] Spatial dimension analysis: Extract data containing geographic location information, use geographic information system (GIS) technology to draw the spatial distribution map of the data, identify hot spots through heat maps or other visualization tools, and finally use clustering algorithms (such as K-means, DBSCAN) to identify geographic areas with similar characteristics.

[0100] Category dimension analysis: extract data containing category information, use classification algorithms (such as decision trees and random forests) to classify the data, use association rule mining algorithms (such as Apriori and FP-Growth) to discover the association between categories, and finally compare the differences between different categories to identify key features.

[0101] Other dimension analysis:

[0102] User behavior dimension: Analyze user operation records, click streams and other data to identify user behavior patterns.

[0103] Product performance dimension: Analyze product performance indicators, such as response time, failure rate, etc., to evaluate product quality.

[0104] By comprehensively analyzing data from multiple perspectives, complex relationships and patterns in the data can be revealed, providing deeper data insights.

[0105] S116, synthesizing the preliminary analysis results of different perspectives to generate a corresponding multi-perspective analysis report, and presenting the multi-perspective analysis results to the user in the form of a three-dimensional image.

[0106] The preliminary analysis results from each perspective are summarized to form a comprehensive data set, and then the results from different perspectives are compared to identify consistency and differences, and the key findings from each perspective are combined for comprehensive analysis to extract the overall conclusion. The key findings from each perspective are described in detail in the report, including trends, patterns, anomalies, etc., and then specific conclusions and suggestions are put forward based on the comprehensive analysis results to guide business decisions. Charts, graphs and other visualization tools can be used to intuitively display the analysis results.

[0107] The multi-view analysis results are converted into a 3D model suitable for holographic projection display, and the holographic projection technology is used to render the 3D image in real time, and the display content is dynamically adjusted according to the user's behavior and context. When the data changes, the 3D image is updated in real time through the incremental update mechanism to ensure that the user sees the latest analysis results.

[0108] Through these steps, not only the accuracy and reliability of the analysis results are improved, but also a highly personalized user experience is provided, ensuring that users can quickly and intuitively understand complex data analysis results.

[0109] After obtaining the preliminary analysis results, the method further includes:

[0110] S101, retrieve pre-set expert rules.

[0111] In this process, domain experts define a series of rules based on their experience and knowledge. These rules can be based on business logic, industry standards, or best practices in specific situations. The defined expert rules are then stored in a rule base. The rule base can be a database, file system, or other form of storage medium. After obtaining the preliminary analysis results, the relevant expert rules are retrieved from the rule base. This may involve selecting an appropriate set of rules based on the data type, analysis task, or user needs.

[0112] S102, evaluating the preliminary analysis results according to the rules defined by the expert rules through the inference engine to obtain corresponding evaluation results.

[0113] Among them, the reasoning engine first parses the expert rules retrieved from the rule base, understands the specific content and logic of each rule, matches the preliminary analysis results with each rule, and checks whether the rule conditions are met. If the conditions are met, the corresponding rule action is executed. The reasoning engine generates an evaluation result based on the application of the rules. It should be pointed out here that the evaluation results may include verification of the preliminary analysis results, correction suggestions or new analysis conclusions.

[0114] S103, comparing the preliminary analysis result with the evaluation result, and if the comparison results are inconsistent, adjusting the preliminary analysis result.

[0115] Among them, the preliminary analysis results are compared item by item with the evaluation results generated by the inference engine, which may involve numerical comparison, pattern matching or other forms of comparison methods; the differences between the preliminary analysis results and the evaluation results are identified, and the differences may include numerical deviations, classification errors, omissions or misjudgments; all differences found are recorded to provide a basis for subsequent adjustments.

[0116] By comparing preliminary analysis results with evaluation results, potential problems and inconsistencies can be identified to ensure the accuracy and consistency of analysis results.

[0117] According to the specific circumstances of the discrepancy, formulate an adjustment strategy, which may include numerical correction, classification adjustment, supplementation of missing data or re-running of part of the analysis process; According to the formulated adjustment strategy, correct the preliminary analysis results. This may involve manual adjustment or automatic adjustment, depending on the nature and complexity of the discrepancy; Verify the adjusted analysis results to ensure that the adjusted results meet the expert rules and expected goals; Update the adjusted analysis results into the system and prepare them for further presentation or report generation.

[0118] Based on the above steps S101 to S103, by calling the pre-set expert rules, the reasoning engine is used to automatically evaluate the preliminary analysis results, and the preliminary analysis results are compared with the evaluation results to identify potential problems and inconsistencies, and then the preliminary analysis results are adjusted according to the comparison results, so as to ensure the accuracy and reliability of the final analysis results. At the same time, it integrates the knowledge and experience of domain experts and complies with industry standards and best practices.

[0119] After obtaining the preliminary analysis results, the method further includes:

[0120] S104, calling a corresponding anomaly detection algorithm, using the anomaly detection algorithm to detect potential anomalies in the preliminary analysis results, and generating a corresponding anomaly detection report, wherein the anomaly detection report marks all detected anomalies.

[0121] Among them, according to the data characteristics and analysis requirements, select the appropriate anomaly detection algorithm. Common anomaly detection algorithms include statistical methods (such as Z-score, IQR), machine learning methods (such as isolation forest, local outlier factor LOF, One-ClassSVM), etc.; according to the characteristics of the data set and historical experience, set the parameters of the anomaly detection algorithm. For example, for statistical-based methods, you may need to set a threshold; for machine learning-based methods, you may need to adjust the model's hyperparameters. You can call the selected anomaly detection algorithm from the pre-built algorithm library and prepare the input data.

[0122] Secondly, the data in the preliminary analysis results are input into the anomaly detection algorithm, and the anomaly detection algorithm is run to identify potential outliers in the data. The algorithm will determine which data points deviate from the normal range based on the set rules or models. Then an anomaly detection report is generated, which contains all detected outliers and related information, such as the specific value of the outlier, location, degree of anomaly, etc., and all detected outliers are marked in the report so that users can quickly identify and handle them.

[0123] S105, obtaining abnormal feedback from the target user on the abnormal detection report.

[0124] Among them, the target user's feedback on the anomaly detection report is collected through the system interface, email, questionnaire, etc. The feedback content may include the user's confirmation of the anomaly, correction suggestions, further explanation requests, etc. Then, the user's feedback text is parsed using natural language processing (NLP) technology or other methods to extract key information, for example, to identify whether the user confirms that a certain anomaly is a real anomaly, or the user has made specific correction suggestions. The user's feedback can be classified, such as confirmation of anomaly, denial of anomaly, and correction suggestions.

[0125] S106, combining the anomaly detection report with the anomaly feedback, and re-evaluating the anomaly value to determine whether there is any omission or misjudgment.

[0126] Among them, the user's anomaly feedback is combined with the anomaly detection report to form a comprehensive data set. This may involve updating the tags in the anomaly detection report and adding user confirmation or correction information.

[0127] During the re-evaluation process, if the user confirms that an outlier is a true anomaly, the anomaly mark is retained; if the user denies that an outlier is an anomaly, the anomaly mark is removed and the user's reasons are recorded; if the user makes specific correction suggestions, the parameters or rules of the anomaly detection algorithm are adjusted according to the suggestions and the algorithm is re-run. The adjusted anomaly detection results are verified to ensure that the new results meet the user's feedback and expectations, and then an updated anomaly detection report can be generated and the final results fed back to the user.

[0128] Based on executing the above steps S104 to S106, a detailed anomaly detection report is generated by calling a suitable anomaly detection algorithm, collecting and combining professional feedback from target users, re-evaluating and adjusting the anomaly values, thereby improving the accuracy of anomaly detection, integrating the user's professional knowledge and experience, and ensuring that the anomaly detection results are more reliable and practical.

[0129] It should be noted here that the preliminary analysis results include multiple prediction values, so the method also includes:

[0130] S107, performing uncertainty assessment on each predicted value in the preliminary analysis result and generating a corresponding confidence level.

[0131] Among them, according to the data characteristics and analysis requirements, choose the appropriate uncertainty quantification method. Common methods include Bayesian method, Monte Carlo sampling, Bootstrap method, etc. If the Bayesian method is used, the model needs to be trained in Bayesian to obtain the posterior distribution; if Monte Carlo sampling or Bootstrap method is used, multiple sampling is required to estimate the distribution of the predicted value.

[0132] For each predicted value, calculate its confidence. For example, in the Bayesian method, you can calculate the posterior probability distribution of the predicted value and extract the confidence interval or confidence level from it; in Monte Carlo sampling or Bootstrap method, you can calculate the standard deviation or percentile of the predicted value through the results of multiple sampling. Then record the confidence of each predicted value to form a confidence list or matrix.

[0133] By performing uncertainty assessment on each predicted value and generating corresponding confidence levels, it provides users with quantitative indicators of the reliability of the prediction results and helps them understand the uncertainty of the prediction results.

[0134] S108, sorting based on all confidence levels, and highlighting the results with higher confidence levels when displayed on a visual interface.

[0135] Among them, all prediction values ​​are sorted from high to low according to confidence, with high-confidence prediction values ​​in front and low-confidence prediction values ​​in the back; the visual interface is designed to ensure that high-confidence results can be highlighted. For example, different colors (such as green for high confidence and red for low confidence), font sizes, icons, etc. can be used to distinguish results with different confidence levels; in addition, during the user's interaction with the system, the visual interface is updated in real time to ensure that the latest confidence ranking results are always displayed. By highlighting high-confidence results in the visual interface, users can quickly identify and focus on the most reliable parts, thereby improving decision-making efficiency.

[0136] S109, generating a corresponding explanation report for the result with the lower confidence ranking.

[0137] Among them, the prediction values ​​with lower confidence are screened out from the sorted results, and an explanation report is generated:

[0138] Possible reasons: Analyze the possible reasons for low confidence, such as poor data quality, inappropriate model parameters, interference from external factors, etc.; Influencing factors: Identify the main factors affecting confidence, such as missing data, outliers, model overfitting or underfitting, etc.; Further verification methods: Propose recommended further verification methods, such as increasing the data sample size, adjusting model parameters, introducing more features, performing cross-validation, etc.

[0139] Finally, the explanation report will be formatted to make it easy to read and understand. The report can contain charts, text descriptions, and specific steps for recommendations.

[0140] By evaluating the uncertainty of each predicted value and generating a confidence level, high-reliability results are highlighted in the visualization interface based on confidence ranking, while providing a detailed explanatory report (including causes, influencing factors, and improvement suggestions) for low-confidence results, thereby improving the transparency, reliability, and interpretability of the prediction results, helping users better understand and optimize the analysis results.

[0141] The present application embodiment discloses a holographic multi-dimensional data analysis device, referring to Figure 2 , the devices include but are not limited to:

[0142] The preliminary analysis result acquisition module 200 acquires the corresponding multidimensional data to be analyzed, pre-processes the multidimensional data to be analyzed, retrieves the pre-trained multidimensional relationship model, and inputs the pre-processed multidimensional data to be analyzed into the multidimensional relationship model for analysis, so as to acquire the corresponding preliminary analysis results; wherein the pre-processing includes adopting an adaptive data cleaning algorithm and a dynamic anomaly detection mechanism, and the multidimensional relationship model integrates a graph neural network, time series analysis, and a hybrid model;

[0143] The preliminary analysis result display module 210 determines the corresponding target user, retrieves the historical behavior and current context corresponding to the target user, and displays the preliminary analysis result to the user in a three-dimensional image through real-time dynamic update based on the historical behavior, current context and holographic projection technology. When the multi-dimensional data to be analyzed changes, the visualization interface corresponding to the preliminary analysis result is updated through the incremental update mechanism;

[0144] The target analysis result acquisition module 220 obtains the corresponding feedback data during the interaction between the target user and the visualization interface, and determines whether to trigger the optimization mechanism based on the feedback data. If triggered, optimization is performed through a preset optimization mechanism based on the feedback data and the preliminary analysis results to obtain the corresponding target analysis results, wherein the feedback data includes the evaluation of the visualization interface results, the interaction behavior records, and the improvement suggestions proposed by the user.

[0145] Furthermore, the device includes but is not limited to:

[0146] A key feature extraction module obtains the target user's preference settings, and uses the preference settings, historical behaviors, and current context to extract corresponding key features, where the key features include the target user's most frequently accessed data types, most frequently used functions, and preferred time periods;

[0147] Display priority adjustment module, used to adjust the display priority of each data in the preliminary analysis results based on key features;

[0148] The preliminary analysis results are presented to the user in a three-dimensional stereoscopic image based on the adjusted display priority.

[0149] Furthermore, the device includes but is not limited to:

[0150] A different perspective dimension determination module is used to determine the different perspective dimensions that need to be analyzed based on the multidimensional data to be analyzed, wherein the different perspective dimensions include time dimension, space dimension, and category dimension;

[0151] The multi-perspective analysis module is used to perform multi-perspective analysis on the multi-dimensional data to be analyzed based on different perspective dimensions, and obtain the preliminary analysis results corresponding to each perspective. The multi-perspective analysis includes time dimension analysis, space dimension analysis, category dimension analysis, and other dimension analysis;

[0152] The multi-perspective analysis report generation module integrates the preliminary analysis results of different perspectives to generate the corresponding multi-perspective analysis report, and presents the multi-perspective analysis results to the user in the form of a three-dimensional stereo image. The multi-perspective analysis report contains the key findings and conclusions of each perspective.

[0153] Furthermore, the device includes but is not limited to:

[0154] An expert rule retrieval module is used to retrieve pre-set expert rules, wherein the expert rules contain the experience and knowledge of domain experts;

[0155] An evaluation result acquisition module evaluates the preliminary analysis results according to the rules defined by the expert rules through the inference engine to obtain the corresponding evaluation results;

[0156] The result comparison result is used to compare the preliminary analysis result with the evaluation result. If the comparison results are inconsistent, the preliminary analysis result is adjusted.

[0157] Furthermore, the device includes but is not limited to:

[0158] The anomaly detection report generation module calls the corresponding anomaly detection algorithm, uses the anomaly detection algorithm to detect potential anomalies in the preliminary analysis results, and is used to generate a corresponding anomaly detection report. The anomaly detection report marks all detected anomalies.

[0159] The abnormal feedback acquisition module is used to obtain the abnormal feedback of the target user on the abnormal detection report, wherein the abnormal feedback includes the user's confirmation of the abnormal value, correction suggestions, etc.;

[0160] The anomaly judgment module combines the anomaly detection report with the anomaly feedback and re-evaluates the anomaly values ​​to determine whether there are omissions or misjudgments.

[0161] Furthermore, the device includes but is not limited to:

[0162] A confidence generation module performs uncertainty assessment on each predicted value in the preliminary analysis results to generate corresponding confidence levels;

[0163] The sorting module is used to sort based on all confidence levels. When displayed on the visual interface, the results with the highest confidence level are highlighted.

[0164] The explanation report generation module is used to generate a corresponding explanation report for the result corresponding to the latter confidence ranking, wherein the explanation report includes possible causes, influencing factors and recommended further verification methods.

[0165] An embodiment of the present application further discloses a method for analyzing multi-dimensional data based on holography, comprising a processor, wherein the program of any one of the above-mentioned methods for analyzing multi-dimensional data based on holography is running on the processor.

[0166] The embodiment of the present application also discloses a storage medium storing a program based on any one of the above-mentioned holographic multi-dimensional data analysis methods.

[0167] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A holographic multidimensional data analysis method, characterized in that: include: Obtain corresponding multidimensional data to be analyzed, preprocess the multidimensional data to be analyzed, retrieve a pre-trained multidimensional relationship model, input the preprocessed multidimensional data to be analyzed into the multidimensional relationship model for analysis, and obtain corresponding preliminary analysis results; wherein the preprocessing includes using an adaptive data cleaning algorithm and a dynamic anomaly detection mechanism, and the multidimensional relationship model integrates a graph neural network, time series analysis, and a hybrid model; Determine the corresponding target user, retrieve the historical behavior and current context corresponding to the target user, and present the preliminary analysis result to the user in a three-dimensional image through real-time dynamic update based on the historical behavior, current context and holographic projection technology. When the multi-dimensional data to be analyzed changes, update the visualization interface corresponding to the preliminary analysis result through an incremental update mechanism; Obtain corresponding feedback data during the interaction between the target user and the visualization interface, and determine whether to trigger an optimization mechanism based on the feedback data; if triggered, perform optimization through a preset optimization mechanism based on the feedback data and the preliminary analysis results to obtain corresponding target analysis results, wherein the feedback data includes evaluations of visualization interface results, interactive behavior records, and improvement suggestions proposed by users.

2. The holographic multi-dimensional data analysis method according to claim 1, characterized in that: After retrieving the historical behavior and current context corresponding to the target user, the method further includes: Obtaining the target user's preference settings, and extracting corresponding key features from the preference settings, the historical behavior, and the current context, wherein the key features include the data types most frequently accessed by the target user, the functions most frequently used, and the preferred time periods; Adjusting the display priority of each data in the preliminary analysis result based on the key features; The preliminary analysis result is presented to the user in a three-dimensional stereoscopic image based on the adjusted display priority.

3. The holographic multi-dimensional data analysis method according to claim 1, characterized in that: The method also includes: Determine different perspective dimensions that need to be analyzed based on the multidimensional data to be analyzed, where the different perspective dimensions include time dimension, space dimension, and category dimension; Based on the different perspective dimensions, a multi-perspective analysis is performed on the multi-dimensional data to be analyzed, and a preliminary analysis result corresponding to each perspective is obtained, wherein the multi-perspective analysis includes time dimension analysis, space dimension analysis, category dimension analysis, and other dimension analysis; The preliminary analysis results from different perspectives are synthesized to generate a corresponding multi-perspective analysis report, which is presented to the user in the form of a three-dimensional image. The multi-perspective analysis report includes the key findings and conclusions of each perspective.

4. The holographic multi-dimensional data analysis method according to claim 3 is characterized in that: After obtaining the preliminary analysis results, the method further includes: Retrieving pre-set expert rules, wherein the expert rules include the experience and knowledge of domain experts; Evaluate the preliminary analysis results according to the rules defined by the expert rules through the inference engine to obtain corresponding evaluation results; The preliminary analysis result is compared with the evaluation result. If the comparison results are inconsistent, the preliminary analysis result is adjusted.

5. The holographic multi-dimensional data analysis method according to claim 4 is characterized in that: After obtaining the preliminary analysis results, the method further includes: Retrieve the corresponding anomaly detection algorithm, use the anomaly detection algorithm to detect potential anomalies in the preliminary analysis results, generate a corresponding anomaly detection report, and mark all detected anomalies in the anomaly detection report; Obtaining abnormal feedback from the target user on the abnormal detection report, wherein the abnormal feedback includes the user's confirmation of the abnormal value, correction suggestions, etc.; The anomaly detection report is combined with the anomaly feedback, and the anomaly values ​​are re-evaluated to determine whether there are any omissions or misjudgments.

6. The holographic multi-dimensional data analysis method according to claim 5, characterized in that: The method also includes: Conduct uncertainty assessment on each predicted value in the preliminary analysis results and generate corresponding confidence levels; Sorting is performed based on all confidence levels. When displayed in the visual interface, the results with the highest confidence levels are highlighted. For the results with the lower confidence ranking, a corresponding explanation report is generated, wherein the explanation report includes possible causes, influencing factors and recommended further verification methods.

7. A holographic multi-dimensional data analysis device, characterized in that: include: A preliminary analysis result acquisition module acquires the corresponding multidimensional data to be analyzed, preprocesses the multidimensional data to be analyzed, retrieves a pre-trained multidimensional relationship model, and inputs the preprocessed multidimensional data to be analyzed into the multidimensional relationship model for analysis, so as to obtain the corresponding preliminary analysis results; wherein the preprocessing includes adopting an adaptive data cleaning algorithm and a dynamic anomaly detection mechanism, and the multidimensional relationship model integrates a graph neural network, time series analysis, and a hybrid model; The preliminary analysis result display module determines the corresponding target user, retrieves the historical behavior and current context corresponding to the target user, and uses the preliminary analysis result to display the three-dimensional image to the user through real-time dynamic update based on the historical behavior, current context and holographic projection technology. When the multi-dimensional data to be analyzed changes, the visualization interface corresponding to the preliminary analysis result is updated through the incremental update mechanism; The target analysis result acquisition module obtains the corresponding feedback data during the interaction between the target user and the visualization interface, and determines whether to trigger the optimization mechanism based on the feedback data. If triggered, optimization is performed through a preset optimization mechanism based on the feedback data and the preliminary analysis results to obtain the corresponding target analysis results, wherein the feedback data includes the evaluation of the visualization interface results, the interaction behavior records, and the improvement suggestions proposed by the user.

8. The holographic multi-dimensional data analysis device according to claim 7, characterized in that: The device also includes: A key feature extraction module, which obtains the target user's preference settings, and uses the preference settings, the historical behavior, and the current context to extract corresponding key features, wherein the key features include the target user's most frequently accessed data types, most frequently used functions, and preferred time periods; A display priority adjustment module, used to adjust the display priority of each data in the preliminary analysis result based on the key feature; The preliminary analysis result is presented to the user in a three-dimensional stereoscopic image based on the adjusted display priority.

9. A holographic multidimensional data analysis method, characterized in that: It comprises a processor, in which runs a program based on the holographic multi-dimensional data analysis method as described in any one of claims 1-6.

10. A storage medium, characterized in that: A program based on a holographic multi-dimensional data analysis method as described in any one of claims 1 to 6 is stored.

Images