A visualization method for business process data analysis

Abstract

The present invention relates to a business process data analysis visualization method, comprising the following steps: obtaining a target time granularity according to the time attribute of the quality management business process data; obtaining a target spatial granularity according to the spatial attribute of the quality management business process data; obtaining a target attribute granularity combination according to the multi-attribute characteristics of the quality management business process; analyzing different granularity levels and organically integrating the analysis results to obtain a multi-granularity fusion analysis result; obtaining visually chaotic data in the multi-granularity fusion analysis result to obtain a visualization analysis result; recommending optimized visual analysis configuration parameters according to the characteristics and analysis objectives of the quality management business process in the visualization analysis result to obtain a final visualization analysis result; and continuously optimizing the analysis granularity selection according to the final visualization analysis result to form an adaptive analysis closed loop. The present invention can effectively improve the analysis efficiency and accuracy of the quality management business process and provide accurate support for decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a business process data analysis visualization method. Background Art

[0002] In the analysis and visualization of the quality management business process, we are faced with the technical difficulty of how to deeply understand the characteristics and laws of the quality management business process data at different granularity levels. First, the quality management business process data usually has a time dimension. How to choose the appropriate time granularity for analysis is a key issue. The choice of time granularity directly affects the grasp of the trend of changes in the quality management business process. Too coarse granularity may cover important short-term changes, and too fine granularity may cause the analysis to be too trivial and difficult to understand the overall trend; secondly, the quality management business process often spans different geographical regions, and the differences between regions may have an important impact on the business process. How to choose the appropriate spatial granularity for analysis, which can reveal the differences between regions without ignoring the similarities between regions due to excessive refinement, is another thorny problem; finally, the quality management business process often involves multiple attribute indicators, and there may be complex correlations between different attributes. How to choose the appropriate attribute granularity for analysis, which can reveal the correlation between attributes without losing the focus of analysis due to too many attributes, is also a problem that needs to be solved urgently.

[0003] In short, how to choose the optimal combination of analysis granularity in multiple dimensions such as time, space, and attributes is the core technical problem faced by the analysis and visualization of quality management business processes. It requires repeated exploration and weighing based on a deep understanding of quality management business scenarios. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the present invention aims to provide a business process data analysis visualization method. The business process data analysis visualization method can effectively improve the analysis efficiency and accuracy of the quality management business process and provide comprehensive and accurate support for decision-making.

[0005] A method for business process data analysis and visualization according to the present invention comprises the following steps:

[0006] S1. Obtain the target time granularity according to the time attribute of the quality management business process data;

[0007] S2. Obtain the target spatial granularity according to the spatial attributes of the quality management business process data;

[0008] S3. Obtain target attribute granularity combination according to the multi-attribute characteristics of the quality management business process;

[0009] S4. Analyze different granularity levels and organically integrate the analysis results to form a comprehensive and three-dimensional quality management business process analysis view and obtain multi-granularity fusion analysis results;

[0010] S5, obtaining visual confusion data in the multi-granularity fusion analysis result, evaluating the visual complexity of different granularity combinations according to the visual confusion problem, and optimizing the visualization layout to obtain the visualization analysis result;

[0011] S6. Recommend optimized visual analysis configuration parameters according to the characteristics of the quality management business process and the analysis objectives in the visual analysis results to obtain the final visual analysis results;

[0012] S7. According to the final visual analysis result, continuously optimize the analysis granularity selection to form an adaptive analysis closed loop.

[0013] Preferably, the step S1 specifically includes:

[0014] Obtain quality management business process data and perform time series decomposition to obtain trend, seasonal and residual parts;

[0015] Determine a candidate time granularity set according to the trend part, the seasonal part and the residual part;

[0016] The sliding window method is used to aggregate the quality management business process data to obtain aggregated data;

[0017] The variance, skewness, and kurtosis are calculated for the aggregated data to obtain the target time granularity in the candidate time granularity set.

[0018] Preferably, the step S2 specifically includes:

[0019] Acquire the quality management business process data, and use the Gaussian mixture function to model the data of each region according to the spatial attributes of the quality management business process data, and extract regional characteristic parameters from it;

[0020] Calculate the Mahalanobis distance according to the characteristic parameters to obtain the inter-region similarity matrix;

[0021] According to the regional similarity matrix, spectral clustering is used to divide the initial spatial units, and the number of clusters is set to the preset maximum spatial granularity;

[0022] Adopting hierarchical clustering method, adjacent highly similar spatial units are merged based on spectral clustering results. When the change rate of Moran's index after merging is less than the preset threshold, the merging is stopped to obtain the adaptively adjusted spatial granularity.

[0023] The boundary detection is performed on the adaptively adjusted spatial units, the significant boundaries between the spatial units are identified through the contour extraction function, the boundary length is calculated to determine the degree of regional difference, and the target spatial granularity is obtained.

[0024] Preferably, the step S3 specifically includes:

[0025] Acquire multi-attribute data of a quality management business process, wherein the multi-attribute data includes quantitative attributes;

[0026] The quantitative attributes are converted into qualitative attributes by using equal frequency, equal width and information gain ratio-based discretization methods to obtain initial attribute granularity;

[0027] According to the initial attribute granularity, candidate association rules are generated by the Apriori algorithm, and the optimal support and confidence thresholds are determined by the grid search method to screen out a set of strong association rules;

[0028] For the set of strongly associated rules, the Jaccard similarity between the rules is calculated, hierarchical clustering is performed, and a rule cluster is obtained by setting a clustering threshold;

[0029] Constructing a dependency graph between attributes according to the structural features of the rule cluster, and calculating the mutual information value of each attribute according to the dependency graph to determine the attribute importance;

[0030] If the attribute importance is lower than the preset threshold, the attribute is merged;

[0031] If the attribute importance is higher than the preset threshold, the attribute is subdivided;

[0032] The adjusted attribute granularity combination is obtained by processing, cross-validation is performed on the adjusted attribute granularity combination, and a validation performance index is calculated;

[0033] Repeat the attribute granularity adjustment and cross-validation steps until the verification performance index reaches the preset conditions and the target attribute granularity combination is obtained.

[0034] Preferably, the step S4 specifically includes:

[0035] According to the determined target time granularity, the target space granularity and the target attribute granularity, the quality management business process data is resampled and aligned, missing data is processed by linear interpolation, and fine-grained data is compressed by an aggregation function to obtain a multidimensional data set of uniform granularity;

[0036] The multidimensional data set is used to construct a multidimensional data cube, wherein the multidimensional data cube uses time, space and attributes as dimensions and quality indicators as measurement values, and realizes data fusion at different granularity levels through slicing operations, dicing operations and rotation operations;

[0037] Analyze the multidimensional data cube, and reduce the dimension of the analysis result by principal component analysis, set a preset variance information ratio, extract key feature vectors, and construct a feature space;

[0038] The analysis results of different granularities are mapped into the feature space, and the similarity of the analysis results of different granularities is quantified by calculating the Euclidean distance;

[0039] A hierarchical clustering algorithm is used to cluster the analysis results of different granularities in the feature space, and a multi-level quality management business process analysis view is constructed based on the clustering results;

[0040] A three-dimensional scatter plot is created, in which the x-axis represents time, the y-axis represents space, the z-axis represents attribute dimension, and the size and color of the points represent quality indicators, forming a comprehensive and three-dimensional quality management business process analysis view and obtaining multi-granularity fusion analysis results.

[0041] Preferably, the step S5 specifically includes:

[0042] Obtain visual clutter data from the multi-granularity fusion analysis results, and use image processing functions to extract data point density, color distribution, and element overlap rate features of a visualization chart;

[0043] Combining features into feature vectors, calculating Mahalanobis distances between the feature vectors, and obtaining visual confusion indexes for each particle size combination;

[0044] According to the visual confusion index and other visual features, a visual complexity evaluation model is constructed, wherein the visual complexity evaluation model uses data density, number of color types, overlapping area ratio and visual confusion index as input variables;

[0045] The multivariate linear regression method was used to fit the visual complexity function and calculate the visual complexity scores of different granularity combinations;

[0046] The information complexity of different granularity combinations is calculated by information entropy, and the information complexity is combined with the visual complexity score to comprehensively evaluate the complexity of each granularity combination, and a genetic algorithm is used to optimize the visual layout;

[0047] The optimized layout scheme is visualized and rendered, and multiple charts are arranged using the grid layout function. The grid structure is dynamically adjusted according to the number and importance of charts to obtain visual analysis results.

[0048] Preferably, the step S6 specifically includes:

[0049] Extracting a feature vector of the quality management business process from the visual analysis result, wherein the feature vector includes a data type, a time span, a spatial range, and an attribute dimension;

[0050] The K-means clustering algorithm is used to classify business processes with similar characteristics, and the silhouette coefficient method is used to determine the optimal number of clusters to obtain the classification results of business characteristics;

[0051] A CART decision tree model is constructed based on the classification results of the business characteristics, and the clustering label and the original feature vector are combined as input features to identify and analyze the type;

[0052] According to the identified analysis target, a matching basic template is selected from the preset visualization configuration template library, and the template parameters are optimized through the fuzzy comprehensive evaluation method;

[0053] The optimized visualization configuration parameters are applied to the original data, and an adaptive rendering algorithm is used to generate visualization charts. The appropriate chart type is dynamically selected according to the data characteristics, and the visual encoding method is adjusted to output the final visualization analysis results that meet the business characteristics and analysis objectives.

[0054] Preferably, the step S7 specifically includes:

[0055] Extracting business process features from the final visualization analysis results, wherein the business process features include time series patterns, spatial distribution features, and multi-dimensional attribute relationships;

[0056] The STL decomposition method is used to separate the trend, seasonality and residual components of the time series, and the decomposition results of the time series are obtained. The Moran's I index is calculated based on the decomposition results to obtain the key attribute set.

[0057] Constructing a multi-objective optimization model for the key attribute set, wherein the multi-objective optimization model takes information entropy, interpretability and computational complexity as optimization objectives;

[0058] Solving the multi-objective optimization model by particle swarm algorithm to obtain the optimal granularity combination;

[0059] Reconstructing the time series data according to the optimal granularity combination to obtain optimized time series data;

[0060] The sliding window technology is used to monitor the characteristics of the optimized time series data, and to update the constraints and objective function of the multi-objective optimization model.

[0061] The business process data analysis visualization method described in the present invention has the following advantages:

[0062] A business process data analysis visualization method of the present invention can comprehensively and three-dimensionally display the quality management business process by combining the granularity analysis of the three dimensions of time, space and attributes, which helps users to deeply understand the business process from multiple angles and find potential problems and improvement points; through the careful selection of time, space and attribute granularity, it can accurately capture the key information and data features in the business process; through the continuous optimization of the analysis granularity selection and the visual analysis configuration parameters, it can adaptively adapt to the changes in the business process and ensure the accuracy and effectiveness of the analysis results; through the provided visual analysis results, it can intuitively display the key information and data features of the business process, provide strong data support for decision makers, help decision makers make decisions quickly, and improve decision efficiency and accuracy; through the continuous optimization of the analysis granularity selection and the visual analysis configuration parameters, it continuously iterates and improves the analysis process, which helps to ensure the continuous optimization and accuracy of the analysis results, and provides strong support for the continuous improvement of the quality management business process. Therefore, the method can effectively improve the analysis efficiency and accuracy of the quality management business process and provide comprehensive and accurate support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flow chart of a business process data analysis visualization method described in the present invention. DETAILED DESCRIPTION

[0064] like Figure 1 As shown, a business process data analysis visualization method according to the present invention comprises the following steps:

[0065] S1. Obtain the target time granularity according to the time attribute of the quality management business process data;

[0066] S2. Obtain the target spatial granularity according to the spatial attributes of the quality management business process data;

[0067] S3. Obtain target attribute granularity combination according to the multi-attribute characteristics of the quality management business process;

[0068] S4. Analyze different granularity levels and organically integrate the analysis results to form a comprehensive and three-dimensional quality management business process analysis view and obtain multi-granularity fusion analysis results;

[0069] S5. Obtain visual clutter data from the multi-granularity fusion analysis results, evaluate the visual complexity of different granularity combinations based on the visual clutter problem that occurs, and optimize the visualization layout to obtain visualization analysis results;

[0070] S6. Recommend optimized visual analysis configuration parameters based on the characteristics of the quality management business process and the analysis objectives in the visual analysis results to obtain the final visual analysis results;

[0071] S7. Based on the final visual analysis results, continuously optimize the analysis granularity selection to form an adaptive analysis closed loop.

[0072] Furthermore, in this embodiment, step S1 specifically includes:

[0073] Obtain quality management business process data and perform time series decomposition to obtain trend, seasonal and residual parts;

[0074] Determine a candidate time granularity set according to the trend part, the seasonal part and the residual part;

[0075] The sliding window method is used to aggregate the quality management business process data to obtain aggregated data;

[0076] Calculate the variance, skewness and kurtosis of the aggregated data to obtain the target time granularity in the candidate time granularity set;

[0077] Further, the quality management business process data is resampled according to the target time granularity to obtain resampled data;

[0078] Exponential smoothing is performed on the resampled data to obtain smoothed data;

[0079] Calculate the residual sequence of quality management business process data and smoothed data;

[0080] Apply DBSCAN algorithm to cluster the residual sequence and obtain the clustering result;

[0081] Identify short-term abnormal changes based on clustering results, combine short-term abnormal changes with smoothed data to obtain evaluation results;

[0082] Select the optimal time granularity based on the evaluation results. The optimal time granularity is used to retain key short-term change information and reflect the overall trend;

[0083] Specifically, the quality management business process data is decomposed into three parts: trend, seasonality and residual. The initial time granularity is determined by analyzing the change rate of the trend part and the cycle length of the seasonal part. The autocorrelation function is applied to the residual part to calculate the optimal lag period. The candidate time granularity set is obtained by comprehensive consideration.

[0084] For the quality management specific indicators such as product qualification rate and customer complaint rate, the STL decomposition method is used to decompose the time series, and the seasonal cycle parameter is set to the length of the business cycle, such as 7 days or 30 days;

[0085] Use the fft module of the numpy library to perform Fourier transform on the decomposed data, analyze the spectrum characteristics, and determine whether there is a significant periodic pattern. If so, add the corresponding period to the candidate time granularity set;

[0086] For each granularity in the candidate time granularity set, the sliding window method is used to aggregate the original data, and the variance, skewness, and kurtosis of the aggregated data are calculated;

[0087] Use the stats module in the scipy library to calculate these statistics, such as the kurtosis() function to calculate kurtosis;

[0088] The variance reflects the degree of data fluctuation, the skewness reflects the asymmetry of the distribution, and the kurtosis indicates the degree of sharpness of the distribution;

[0089] Combined with the changing trends of these statistics, select the time granularity that can retain the original data characteristics to the greatest extent;

[0090] Resample the data according to the selected time granularity, and apply exponential smoothing method to the resampled data to perform exponential smoothing to eliminate the impact of short-term random fluctuations and extract long-term trend information;

[0091] Use the exponential smoothing class in the statsmodels library, set trend and seasonal parameters, and calculate the residual series of the original data and smoothed data for subsequent anomaly detection;

[0092] The DBSCAN algorithm is applied to cluster the residual sequence to identify short-term abnormal changes;

[0093] Use the NearestNeighbors method in the sklearn library to calculate the optimal eps value and set the min_samples parameter;

[0094] Combine the clustering results with the long-term trend information to comprehensively evaluate the overall trend and short-term change characteristics of the data;

[0095] Based on the evaluation results, the optimal time granularity that can reflect the overall trend is selected as the final target time granularity while retaining key short-term change information;

[0096] Here is an example:

[0097] The quality management business process data is decomposed into time series. Taking product qualification rate as an example, the STL decomposition method is used to decompose the data into three parts: trend, seasonality and residual.

[0098] The seasonal period parameter is set to 30 days to reflect the monthly business cycle. The trend analysis shows that the change rate is 0.5% / month. The seasonal period is 30 days. The autocorrelation function of the residual shows that the best lag period is 7 days. Based on this, the candidate time granularity set is determined as {1 day, 7 days, 30 days};

[0099] Using numpy's fft module to perform Fourier transform on the decomposed data, we found that there is a significant peak at the 30-day period, which further verifies the rationality of 30 days as a candidate time granularity.

[0100] For the candidate time granularity set, a sliding window with a width of 7 days is used to aggregate the original data;

[0101] The statistics of the aggregated data are calculated using the stats module of scipy. The variance of the 1-day granularity is 0.0025, the skewness is 0.15, and the kurtosis is 2.8; the variance of the 7-day granularity is 0.0018, the skewness is 0.08, and the kurtosis is 3.1; the variance of the 30-day granularity is 0.0010, the skewness is 0.03, and the kurtosis is 3.3. Considering that the 7-day granularity can effectively reduce the impact of noise while retaining the data fluctuation characteristics, 7 days is selected as the preliminary time granularity;

[0102] Resample the selected 7-day time granularity data, use the ExponentialSmoothing class in the statsmodels library for exponential smoothing, set the trend parameter to additive trend and the seasonal parameter to multiplicative seasonality;

[0103] The smoothed data show that the overall product qualification rate presents a slow upward trend, with obvious monthly fluctuations. The residual sequence of the original data and the smoothed data is calculated, and the residual standard deviation is 0.015;

[0104] The DBSCAN algorithm is applied to cluster the residual sequence. The best eps value is calculated using the NearestNeighbors method of sklearn to be 0.03, and min_samples is set to 3;

[0105] The clustering results identified five outliers, corresponding to the time periods of production line adjustments and raw material batch changes;

[0106] A comprehensive assessment of long-term trends and short-term anomalies was conducted, and 7 days was determined as the final target time granularity, which can not only reflect the overall quality trend but also capture key short-term changes.

[0107] Furthermore, in this embodiment, step S2 specifically includes:

[0108] Acquire the quality management business process data, and use the Gaussian mixture function to model the data of each region according to the spatial attributes of the quality management business process data, and extract regional characteristic parameters from it;

[0109] Calculate the Mahalanobis distance based on the feature parameters to obtain the inter-region similarity matrix;

[0110] According to the regional similarity matrix, spectral clustering is used to divide the initial spatial units, and the number of clusters is set to the preset maximum spatial granularity;

[0111] Adopting hierarchical clustering method, adjacent highly similar spatial units are merged based on spectral clustering results. When the change rate of Moran's index after merging is less than the preset threshold, the merging is stopped to obtain the adaptively adjusted spatial granularity.

[0112] The boundary detection is performed on the adaptively adjusted spatial units. The significant boundaries between the spatial units are identified through the contour extraction function. The boundary length is calculated to determine the degree of regional difference and obtain the target spatial granularity. If the number of spatial units is greater than the preset threshold, the boundary length threshold is increased and the adjacent units with weak boundaries are merged.

[0113] Specifically, for the spatial attributes of the quality management business process data, a Gaussian mixture function is used to model the data of each region, from which regional characteristic parameters are extracted, and the similarity between regions is quantified by calculating the Mahalanobis distance of the characteristic parameters to obtain a regional similarity matrix;

[0114] According to the regional similarity matrix, spectral clustering is used to divide the initial spatial units, the number of clusters is set to the preset maximum spatial granularity, spatial autocorrelation analysis is performed on the clustering results, and the Moran index is calculated to measure the degree of spatial association;

[0115] A hierarchical clustering method was used to gradually merge adjacent highly similar spatial units based on the spectral clustering results. The Moran's index was recalculated after each merger, and the Moran's index change rate was calculated, which was defined as the difference between the current Moran's index and the previous Moran's index divided by the previous Moran's index.

[0116] When the exponential change rate is less than a preset threshold, the merging is stopped to obtain the adaptively adjusted spatial granularity;

[0117] The boundary detection is performed on the adaptively adjusted spatial units. The contour extraction function is used to identify the significant boundaries between the spatial units. The degree of regional difference is determined by calculating the boundary length. The target spatial granularity is determined by combining the degree of difference and the number of spatial units.

[0118] If the number of spatial units is greater than the preset threshold, the boundary length threshold is increased to merge adjacent units with weak boundaries to avoid over-refinement and finally obtain the target spatial granularity;

[0119] Here is an example:

[0120] Analyze the spatial attribute data of the quality management business process, taking the product defect rate as an example;

[0121] Gaussian mixture function is used to model the data of each region and extract characteristic parameters such as mean and covariance;

[0122] For each of the 10 production areas, 5 feature parameters are extracted to form a 50-dimensional feature vector. The Mahalanobis distance is used to calculate the similarity between regions, and a 10x10 similarity matrix is ​​constructed. Based on this matrix, the spectral clustering algorithm is applied to perform initial spatial unit division, and the maximum spatial granularity is set to 5;

[0123] The Moran index was calculated for the clustering results, with an initial value of 0.75, indicating strong spatial autocorrelation;

[0124] Then, the hierarchical clustering method was used to gradually merge the five initial units. The Moran index was recalculated after each merger. For example, the change rate was calculated from 0.75 to 0.72, and the change rate was -4%. The threshold was set to -5%. When the change rate reached -5.2%, the merger was stopped, and three adaptively adjusted spatial units were obtained.

[0125] The boundaries of the three spatial units are detected, and the contour extraction function is used to identify the boundaries and calculate the boundary length. Assuming that the boundary lengths between the three units are 100, 80, and 60 pixels respectively, the boundary length threshold is set to 70, and the weak boundary adjacent units smaller than the threshold are merged, and finally two spatial units are obtained as the target spatial granularity;

[0126] This approach not only reveals significant regional differences but also avoids over-refinement, achieving adaptive adjustment of spatial granularity.

[0127] Furthermore, in this embodiment, step S3 specifically includes:

[0128] Obtain multi-attribute data of the quality management business process, where the multi-attribute data includes quantitative attributes;

[0129] The quantitative attributes are converted into qualitative attributes by using equal frequency, equal width and information gain ratio-based discretization methods to obtain the initial attribute granularity;

[0130] According to the initial attribute granularity, the candidate association rules are generated by the Apriori algorithm, and the grid search method is used to determine the optimal support and confidence thresholds to screen out the strong association rule set;

[0131] For the set of strongly associated rules, the Jaccard similarity between the rules is calculated, hierarchical clustering is performed, and the rule clusters are obtained by setting the clustering threshold;

[0132] According to the structural characteristics of the rule cluster, a dependency graph between attributes is constructed, and the mutual information value of each attribute is calculated based on the dependency graph to determine the attribute importance;

[0133] If the attribute importance is lower than the preset threshold, the attribute is merged;

[0134] If the attribute importance is higher than the preset threshold, the attribute is subdivided;

[0135] The adjusted attribute granularity combination is obtained by processing, cross-validation is performed on the adjusted attribute granularity combination, and a validation performance index is calculated;

[0136] Repeat the attribute granularity adjustment and cross-validation steps until the verification performance index reaches the preset conditions and the target attribute granularity combination is obtained;

[0137] Specifically, for the multi-attribute data of the quality management business process, a variety of discretization methods are used to convert quantitative attributes into qualitative attributes, including equal-frequency and equal-width discretization and the selection of optimal partition points based on the information gain ratio. The initial attribute granularity is determined by comparing the information retention degree of different discretization results;

[0138] Based on the initial attribute granularity, the candidate association rules are generated by the Apriori algorithm, and the optimal support and confidence thresholds are determined by the grid search method to screen out the strong association rule set;

[0139] Perform hierarchical clustering on the selected strong association rules, calculate the Jaccard similarity between the rules as the clustering basis, obtain the rule clusters by setting the clustering threshold, and construct the dependency relationship diagram between the attributes according to the structural characteristics of the rule clusters to reveal the internal connection between the attributes;

[0140] Based on the dependency graph, the importance of each attribute is calculated using mutual information. The attribute importance and association rule strength are combined to dynamically adjust the attribute granularity, merge low-importance attributes, and subdivide high-importance attributes, and finally determine the target attribute granularity combination;

[0141] In the process of adjusting the attribute granularity, the performance of different granularity combinations is evaluated through the cross-validation method, and the optimal target attribute granularity combination is selected;

[0142] Here is an example:

[0143] For the multi-attribute data of the quality management business process, taking the product production line as an example, it includes 10 quantitative attributes such as temperature, pressure, and raw material purity, and 5 qualitative attributes;

[0144] Firstly, the equal frequency, equal width and information gain ratio based discretization methods are used to process quantitative attributes. For temperature attribute, the equal frequency method is used to divide the data into 5 intervals, the equal width method is used to divide the data into 6 intervals, and the information gain ratio method selects 4 optimal division points.

[0145] By comparing the information retention of different discretization results, the information gain ratio method is selected as the initial granularity of the temperature attribute. This process is repeated for other attributes to finally determine the initial attribute granularity combination.

[0146] Based on the initial granularity, the candidate association rules are generated by the Apriori algorithm, the support threshold range is set to [0.1, 0.5], the confidence threshold range is set to [0.6, 0.9], the step size is 0.1, and the optimal threshold combination is determined by grid search to be 0.2 for support and 0.7 for confidence.

[0147] 100 strong association rules were selected, such as "high temperature & high pressure → unqualified product (support 0.22, confidence 0.75)";

[0148] The Jaccard similarity of these 100 rules was calculated, the similarity matrix was constructed, and hierarchical clustering was applied to divide the rules into 10 clusters;

[0149] Based on the structural characteristics of the clusters, an attribute dependency graph was constructed, revealing a strong correlation between the “temperature” and “pressure” attributes;

[0150] The mutual information of each attribute is calculated, and the importance of "temperature" and "raw material purity" are found to be 0.8 and 0.6 respectively;

[0151] Combining importance and association rule strength, the "temperature" attribute is subdivided into 6 intervals, and the attributes with low importance are merged;

[0152] The performance of different granularity combinations was evaluated through 5-fold cross validation, and the target attribute granularity combination consisting of 8 quantitative attributes (2 refined and 4 merged) and 5 qualitative attributes was finally determined.

[0153] Furthermore, in this embodiment, step S4 specifically includes:

[0154] According to the determined target time granularity, target space granularity and target attribute granularity, the quality management business process data is resampled and aligned, the missing data is processed by linear interpolation, and the fine-grained data is compressed by aggregation functions to obtain a multidimensional data set with uniform granularity;

[0155] Use multidimensional data sets to construct multidimensional data cubes. Multidimensional data cubes use time, space and attributes as dimensions and quality indicators as measurement values. They achieve data fusion at different granularity levels through slicing, dicing and rotation operations.

[0156] Analyze the multidimensional data cube, and reduce the dimension of the analysis results through principal component analysis, set the preset variance information ratio, extract key feature vectors, and construct feature space;

[0157] The analysis results of different granularities are mapped into the feature space, and the similarity of the analysis results of different granularities is quantified by calculating the Euclidean distance;

[0158] A hierarchical clustering algorithm is used to cluster the analysis results of different granularities in the feature space, and a multi-level quality management business process analysis view is constructed based on the clustering results;

[0159] Create a three-dimensional scatter plot, where the x-axis represents time, the y-axis represents space, the z-axis represents attribute dimensions, and the size and color of the points represent quality indicators, forming a comprehensive and three-dimensional quality management business process analysis view and obtaining multi-granularity fusion analysis results;

[0160] Specifically, according to the determined target time granularity, target space granularity and target attribute granularity, the quality management business process data is resampled and aligned, the missing data is processed by linear interpolation, and the fine-grained data is compressed by aggregation function to obtain a multidimensional data set with uniform granularity;

[0161] Construct a multidimensional data cube, taking time, space and attributes as dimensions and quality indicators as measurement values, and achieve data fusion at different granularity levels through slicing, dicing and rotation operations to obtain multi-angle quality management business process analysis results;

[0162] The principal component analysis method is applied to the analysis results of the multidimensional data cube to reduce the dimension, set to retain 95% of the variance information, extract the key feature vectors, construct the feature space, map the analysis results of different granularities into the feature space, and quantify the similarity of the analysis results of different granularities by calculating the Euclidean distance;

[0163] Based on the mapping results of the feature space, a hierarchical clustering algorithm is used to cluster the analysis results of different granularities. A multi-level quality management business process analysis view is constructed based on the clustering results. The analysis results in the same cluster are fused through the weighted average method based on information entropy to obtain multi-granularity fusion analysis results.

[0164] Create a three-dimensional scatter plot, with the x-axis representing time, the y-axis representing space, the z-axis representing attribute dimensions, and the size and color of the points representing quality indicators, to form a comprehensive and three-dimensional analysis view of the quality management business process;

[0165] Here is an example:

[0166] Taking the quality management business process of an automobile manufacturing company as an example, a multi-granularity fusion analysis of production line data is performed;

[0167] First, the original data are resampled and aligned according to the determined target time granularity (daily), target spatial granularity (production line level), and target attribute granularity (10 key quality indicators);

[0168] Use linear interpolation to handle missing data, for example, interpolate missing hourly temperature data into 24-hour complete data;

[0169] The hourly data is compressed into daily data through the average aggregation function to obtain a multidimensional data set of uniform granularity. A multidimensional data cube is constructed with time (365 days), space (5 production lines) and attributes (10 indicators) as dimensions and quality score as the measurement value;

[0170] Perform slicing operations, for example, select data for July;

[0171] Perform slicing operations, such as analyzing all indicators of production line 2;

[0172] Perform rotation operations, such as swapping the time dimension with the space dimension;

[0173] Apply principal component analysis to the cube analysis results, set to retain 95% variance information, extract 3 principal components from 10 original indicators, and construct the feature space;

[0174] The analysis results of different granularities (such as monthly, quarterly, and annual) are mapped to the feature space and the Euclidean distance is calculated. It is found that the distance between the quarterly granularity and the annual granularity is 0.15, while the distance between the monthly granularity and the annual granularity is 0.38;

[0175] A hierarchical clustering algorithm is used to cluster the mapping results to obtain three levels of analysis views. The weights are calculated using information entropy. For example, the weights of the monthly, quarterly, and annual granularities are 0.5, 0.3, and 0.2, respectively. The weighted average is used to obtain the fusion analysis results.

[0176] Finally, a three-dimensional scatter plot is created, with the x-axis representing 365 days, the y-axis representing 5 production lines, and the z-axis representing 10 quality indicators. The size of the dots represents the quality score, and the colors represent different production lines, forming a comprehensive and three-dimensional analysis view.

[0177] Furthermore, in this embodiment, step S5 specifically includes:

[0178] Obtain visually cluttered data from multi-granularity fusion analysis results, and use image processing functions to extract data point density, color distribution, and element overlap characteristics of visualization charts;

[0179] Combining features into feature vectors, calculating the Mahalanobis distance between feature vectors, and obtaining the visual confusion index of each granularity combination;

[0180] According to the visual confusion index and other visual features, a visual complexity evaluation model is constructed. The visual complexity evaluation model uses data density, number of color types, overlapping area ratio and visual confusion index as input variables.

[0181] The multivariate linear regression method was used to fit the visual complexity function and calculate the visual complexity scores of different granularity combinations;

[0182] The information complexity of different granularity combinations is calculated by information entropy, and the information complexity is combined with the visual complexity score to comprehensively evaluate the complexity of each granularity combination. The genetic algorithm is used to optimize the visual layout. The genetic algorithm encodes the layout parameters into chromosomes and uses the inverse of the complexity evaluation result as the fitness function.

[0183] Visualize the optimized layout scheme and use the grid layout function to arrange multiple charts. Dynamically adjust the grid structure according to the number and importance of charts to obtain visual analysis results.

[0184] Specifically, for visually cluttered data in the multi-granularity fusion analysis results, image processing functions are used to extract data point density, color distribution, and element overlap rate characteristics of visualization charts;

[0185] Quantify color distribution by calculating color histograms, identify overlapping areas using contour detection algorithms, and calculate data point density;

[0186] These features are combined into feature vectors, and the Mahalanobis distance between feature vectors is calculated to obtain the visual confusion index of each granularity combination;

[0187] Based on the visual confusion index and other visual features, a visual complexity evaluation model was constructed. The data density, number of color types, overlapping area ratio factor and visual confusion index were used as input variables. The visual complexity function was fitted using the multivariate linear regression method to calculate the visual complexity scores of different granularity combinations. At the same time, the information entropy was used to calculate the information complexity of different granularity combinations. Combined with the visual complexity score, the complexity of each granularity combination was comprehensively evaluated.

[0188] According to the complexity evaluation results, a genetic algorithm is applied to optimize the visualization layout. The layout parameters such as chart type, color scheme, element size, etc. are encoded as chromosomes. The inverse of the complexity evaluation result is used as the fitness function. The population size is set to 100, the crossover rate is 0.8, and the mutation rate is 0.1. The layout scheme is iteratively optimized through selection, crossover, and mutation operations.

[0189] Visually render the optimized layout scheme, use the grid layout function to arrange multiple charts, dynamically adjust the grid structure according to the number and importance of charts, balance information density and readability by adjusting parameters such as chart spacing and alignment, and finally generate optimized visual analysis results;

[0190] Here is an example:

[0191] Taking the visual analysis of quality management of a manufacturing enterprise as an example, the analysis results of 10 different granularity combinations were visually optimized;

[0192] First, use image processing functions to extract the features of each chart, and calculate the color distribution through the color histogram. For example, the main colors of the chart are red (60%), blue (30%), and green (10%).

[0193] Using the contour detection algorithm to identify overlapping areas, the element overlap rate of the chart is calculated to be 15%. Through data point density analysis, it is found that every 100 square pixels in the chart contain 8 data points;

[0194] Combining these features, the Mahalanobis distance was calculated to obtain the visual clutter index, for example, the index for daily granularity data was 0.8 and for monthly granularity data was 0.5;

[0195] The visual confusion index and other features are input into the multivariate linear regression model to fit the visual complexity function: complexity = 0.4*visual confusion index+0.3*data density+0.2*number of color types+0.1*element overlap rate. At the same time, the information entropy is calculated. For example, the entropy value of daily granularity data is 4.2, and that of monthly granularity data is 3.8.

[0196] Combining visual complexity and information entropy, we get a comprehensive complexity score, apply genetic algorithm to optimize the layout, and encode chromosomes containing parameters such as chart type (e.g. bar chart, line chart), color scheme (e.g. warm colors, cool colors), and element size (e.g. large, medium, small);

[0197] The population size is set to 100, and 50 generations are iterated. The layout with the highest fitness is: the daily granularity data with high complexity is represented by a line chart, the color is cold, and the element size is set to small; the monthly granularity data with low complexity is represented by a bar chart, the color is warm, and the element size is set to medium;

[0198] Finally, the grid layout function is used to allocate more space to high-complexity charts based on the complexity scores of the 10 charts. For example, daily granularity data occupies a 2x2 grid, and monthly granularity data occupies a 1x1 grid. The spacing and alignment are dynamically adjusted to generate the final optimized visualization results.

[0199] Furthermore, in this embodiment, step S6 specifically includes:

[0200] Extracting feature vectors of the quality management business process from the visual analysis results. The feature vectors include data type, time span, spatial range, and attribute dimension.

[0201] The K-means clustering algorithm is used to classify business processes with similar characteristics, and the silhouette coefficient method is used to determine the optimal number of clusters to obtain the classification results of business characteristics;

[0202] A CART decision tree model is constructed based on the classification results of business characteristics, and the clustering label and the original feature vector are combined as input features to identify and analyze the type;

[0203] According to the identified analysis target, a matching basic template is selected from the preset visualization configuration template library, and the template parameters are optimized through the fuzzy comprehensive evaluation method;

[0204] Apply the optimized visualization configuration parameters to the original data, use the adaptive rendering algorithm to generate visualization charts, dynamically select the appropriate chart type according to the data characteristics, adjust the visual encoding method, and output the final visualization analysis results that meet the business characteristics and analysis goals;

[0205] Specifically, the feature vector of the quality management business process is extracted from the visual analysis results. The feature vector includes data type, time span, spatial range and attribute dimension.

[0206] The K-means clustering algorithm is used to classify business processes with similar characteristics, and the silhouette coefficient method is used to determine the optimal number of clusters to obtain the classification results of business characteristics;

[0207] Based on the classification results of business characteristics, a CART decision tree model is constructed to identify the analysis target. The clustering label and the original feature vector are combined as input features, and the analysis target type (such as trend analysis, anomaly detection, and comparative analysis) is used as the output label. The model is trained to automatically identify the analysis target of the current business process.

[0208] According to the identified analysis target, a matching basic template is selected from the preset visualization configuration template library, and the template parameters are optimized through the fuzzy comprehensive evaluation method. The factor set (data distribution, readability, information density), comment set (excellent, good, medium, poor) and weight set are established, and the fuzzy comprehensive evaluation matrix is ​​calculated to obtain the optimal visualization configuration parameter combination, including chart type, color scheme, layout structure, and interaction method.

[0209] Apply the optimized visualization configuration parameters to the original data, use the adaptive rendering algorithm to generate visualization charts, dynamically select the appropriate chart type according to the data characteristics (for example, line charts for time series data and bar charts for categorical data), adjust the visual encoding method (for example, mapping values ​​to visual attributes such as color depth and size), and output the final visualization analysis results that meet the business characteristics and analysis objectives;

[0210] Here is an example:

[0211] Taking the visual analysis of quality management of an automobile manufacturing company as an example, feature vectors are extracted from the analysis results. The feature vectors include data type (continuous, discrete), time span (day, week, month), spatial scope (workshop, production line, workstation) and attribute dimension (number of quality indicators);

[0212] The K-means clustering algorithm was used to classify 100 business processes, and the optimal number of clusters was determined to be 4 by calculating the silhouette coefficient;

[0213] The clustering results show that category 1 is short-term continuous data (accounting for 35%), category 2 is long-term discrete data (accounting for 25%), category 3 is multi-dimensional comprehensive data (accounting for 20%), and category 4 is spatial distribution data (accounting for 20%);

[0214] Based on the clustering results, a CART decision tree model is constructed. The input features include cluster labels and original feature vectors, and the output label is the analysis target type.

[0215] After training, the model achieved an accuracy rate of 85% on the test set and successfully identified that the analysis target of the current business process was trend analysis;

[0216] According to the recognition results, a time series analysis template is selected from the template library, and the fuzzy comprehensive evaluation method is used to optimize the parameters;

[0217] Establish a set of factors (data distribution, readability, information density), a set of comments (excellent, good, fair, poor), and a set of weights (0.4, 0.3, 0.3), and calculate the optimal configuration: line chart, blue gradient color scheme, grid layout, and zoom interaction;

[0218] Apply the optimized configuration to the original data to generate a visual chart. For the detected time series data, a line chart is used for display. The x-axis represents time (span of 30 days) and the y-axis represents the quality indicator value (range 0-100). The value size is mapped to the line thickness (1-5 pixels), and the abnormal points are marked in red.

[0219] The final output of the visual analysis results clearly shows the changing trend of the quality indicators within 30 days, highlights 3 abnormal points, and uses zoom interaction to view detailed data.

[0220] Furthermore, in this embodiment, step S7 specifically includes:

[0221] Extract business process features from the final visualization analysis results. Business process features include time series patterns, spatial distribution features, and multi-dimensional attribute relationships.

[0222] The STL decomposition method is used to separate the trend, seasonality and residual components of the time series, and the decomposition results of the time series are obtained. The Moran's I index is calculated based on the decomposition results to obtain the key attribute set.

[0223] A multi-objective optimization model is constructed for the key attribute set, and the multi-objective optimization model takes information entropy, interpretability and computational complexity as optimization objectives;

[0224] The multi-objective optimization model is solved by particle swarm algorithm to obtain the optimal granularity combination;

[0225] Reconstruct the time series data according to the optimal granularity combination to obtain optimized time series data;

[0226] The sliding window technology is used to monitor the features of the optimized time series data, and the constraints and objective functions of the multi-objective optimization model are updated. When monitoring the features, if the Euclidean distance of the feature vector changes by more than the preset threshold, the granular optimization process is triggered, and the constraints and objective functions of the multi-objective optimization model are updated after optimization.

[0227] Specifically, business process characteristics are extracted from the final visualization analysis results. Business process characteristics include time series patterns, spatial distribution characteristics, and multidimensional attribute relationships. The STL decomposition method is used to separate the trend, seasonality, and residual components of the time series. The spatial autocorrelation is identified by calculating the Moran's I index. The principal component analysis method is used to reduce the dimension and extract key attributes.

[0228] Use radar charts to visually present the extracted business characteristics and show the strengths and weaknesses of the features in each dimension;

[0229] Based on the extracted features, a multi-objective optimization model is constructed, with information entropy, interpretability and computational complexity as optimization targets, time granularity, spatial granularity and attribute granularity as decision variables, and a weighted method is used to transform multiple targets into a single objective function, with weights set according to business needs;

[0230] The optimal granularity combination is solved by particle swarm algorithm, and the optimization results are preliminarily evaluated to calculate the objective function value and constraint satisfaction;

[0231] Conduct sensitivity analysis on the optimized granularity combination, use variance analysis (ANOVA) to quantify the impact of different granularity factors on the analysis results, identify key granularity factors, and construct granularity influence evaluation indicators. Design an adaptive adjustment strategy based on the granularity influence evaluation results, use sliding window technology to continuously monitor feature changes, trigger the granularity optimization process when the Euclidean distance of the feature vector changes by more than 20%, update the constraints and objective function of the optimization model, realize dynamic adjustment of the analysis granularity, and form a closed-loop optimization mechanism;

[0232] Here is an example:

[0233] Taking the quality management business of an automobile manufacturing company as an example, features are extracted from the final visual analysis results;

[0234] The STL decomposition method is used to decompose the daily product qualification rate into time series, and the trend term (showing a trend of increasing by 0.5% every month), the seasonal term (a cyclical decrease of 2% every Friday) and the residual term are obtained;

[0235] By calculating the Moran's I index (value of 0.75), we identified that product quality has significant spatial clustering. After principal component analysis and dimensionality reduction, we extracted three key attributes (temperature, pressure, and raw material purity), which cumulatively explained 80% of the variance. We used a radar chart to display the business characteristics, and the values ​​of each dimension were: time trend (0.8), spatial clustering (0.75), and attribute association (0.7).

[0236] Construct a multi-objective optimization model, setting information entropy (weight 0.4), interpretability (weight 0.3) and computational complexity (weight 0.3) as optimization objectives;

[0237] Decision variables include time granularity (hours / days / weeks), space granularity (workstation / production line / workshop), and attribute granularity (3 / 5 / 7 attributes);

[0238] The particle swarm algorithm is used to solve the problem and obtain the optimal granularity combination: the time granularity is day, the space granularity is production line, and the attribute granularity is 5 attributes;

[0239] Preliminary evaluation shows that the weighted objective function value of this combination is 0.85. A variance analysis of this granularity combination shows that the time granularity has the greatest impact on the analysis results (F value is 15.6), followed by the attribute granularity (F value is 10.2) and the spatial granularity (F value is 8.7).

[0240] A 50-day sliding window is set to continuously monitor feature changes. When the Euclidean distance changes by more than 20%, the optimization process is triggered. For example, if a sudden increase in the importance of the temperature attribute is detected (the contribution rate increases from 25% to 40%), the attribute granularity is immediately adjusted, and temperature-related indicators are added to form a new granularity combination of 6 attributes to achieve dynamic optimization of the analysis granularity.

[0241] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention.

[0242] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A business process data analysis visualization method, characterized in that: The following steps are involved: S1. Obtain the target time granularity according to the time attribute of the quality management business process data; S2. Obtain the target spatial granularity according to the spatial attributes of the quality management business process data; S3. Obtain target attribute granularity combination according to the multi-attribute characteristics of the quality management business process; S4. Analyze different granularity levels and organically integrate the analysis results to form a comprehensive and three-dimensional quality management business process analysis view and obtain multi-granularity fusion analysis results; S5, obtaining visual confusion data in the multi-granularity fusion analysis result, evaluating the visual complexity of different granularity combinations according to the visual confusion problem, and optimizing the visualization layout to obtain the visualization analysis result; S6. Recommend optimized visual analysis configuration parameters according to the characteristics of the quality management business process and the analysis objectives in the visual analysis results to obtain the final visual analysis results; S7. According to the final visual analysis result, continuously optimize the analysis granularity selection to form an adaptive analysis closed loop.

2. The business process data analysis visualization method according to claim 1, characterized in that: The step S1 specifically includes: Obtain quality management business process data and perform time series decomposition to obtain trend, seasonal and residual parts; Determine a candidate time granularity set according to the trend part, the seasonal part and the residual part; The sliding window method is used to aggregate the quality management business process data to obtain aggregated data; The variance, skewness, and kurtosis are calculated for the aggregated data to obtain the target time granularity in the candidate time granularity set.

3. The business process data analysis visualization method according to claim 1, characterized in that: The step S2 specifically includes: Acquire the quality management business process data, and use the Gaussian mixture function to model the data of each region according to the spatial attributes of the quality management business process data, and extract regional characteristic parameters from it; Calculate the Mahalanobis distance according to the characteristic parameters to obtain the inter-region similarity matrix; According to the inter-region similarity matrix, spectral clustering is used to divide the initial spatial units, and the number of clusters is set to the preset maximum spatial granularity; Adopting hierarchical clustering method, adjacent highly similar spatial units are merged based on spectral clustering results. When the change rate of Moran's index after merging is less than the preset threshold, the merging is stopped to obtain the adaptively adjusted spatial granularity. The boundary detection is performed on the adaptively adjusted spatial units, the significant boundaries between the spatial units are identified through the contour extraction function, the boundary length is calculated to determine the degree of regional difference, and the target spatial granularity is obtained.

4. The business process data analysis visualization method according to claim 1, characterized in that: The step S3 specifically includes: Acquire multi-attribute data of a quality management business process, wherein the multi-attribute data includes quantitative attributes; The quantitative attributes are converted into qualitative attributes by using equal frequency, equal width and information gain ratio-based discretization methods to obtain initial attribute granularity; According to the initial attribute granularity, candidate association rules are generated by Apriori algorithm, and the optimal support and confidence thresholds are determined by grid search method to screen out a set of strong association rules; For the set of strongly associated rules, the Jaccard similarity between the rules is calculated, hierarchical clustering is performed, and a rule cluster is obtained by setting a clustering threshold; Constructing a dependency graph between attributes according to the structural features of the rule cluster, and calculating the mutual information value of each attribute according to the dependency graph to determine the attribute importance; If the attribute importance is lower than the preset threshold, the attribute is merged; If the attribute importance is higher than the preset threshold, the attribute is subdivided; The adjusted attribute granularity combination is obtained by processing, cross-validation is performed on the adjusted attribute granularity combination, and a validation performance index is calculated; Repeat the attribute granularity adjustment and cross-validation steps until the verification performance index reaches the preset conditions and the target attribute granularity combination is obtained.

5. The business process data analysis visualization method according to claim 1, characterized in that: The step S4 specifically includes: According to the determined target time granularity, the target space granularity and the target attribute granularity, the quality management business process data is resampled and aligned, missing data is processed by linear interpolation, and fine-grained data is compressed by an aggregation function to obtain a multidimensional data set of uniform granularity; The multidimensional data set is used to construct a multidimensional data cube, wherein the multidimensional data cube uses time, space and attributes as dimensions and quality indicators as measurement values, and realizes data fusion at different granularity levels through slicing operations, dicing operations and rotation operations; Analyze the multidimensional data cube, and reduce the dimension of the analysis result by principal component analysis, set a preset variance information ratio, extract key feature vectors, and construct a feature space; The analysis results of different granularities are mapped into the feature space, and the similarity of the analysis results of different granularities is quantified by calculating the Euclidean distance; A hierarchical clustering algorithm is used to cluster the analysis results of different granularities in the feature space, and a multi-level quality management business process analysis view is constructed based on the clustering results; A three-dimensional scatter plot is created, in which the x-axis represents time, the y-axis represents space, the z-axis represents attribute dimension, and the size and color of the points represent quality indicators, forming a comprehensive and three-dimensional quality management business process analysis view and obtaining multi-granularity fusion analysis results.

6. The method for business process data analysis and visualization according to claim 1, characterized in that: The step S5 specifically includes: Obtain visual clutter data from the multi-granularity fusion analysis results, and use image processing functions to extract data point density, color distribution, and element overlap rate features of a visualization chart; Combining features into feature vectors, calculating Mahalanobis distances between the feature vectors, and obtaining visual confusion indexes for each particle size combination; According to the visual confusion index and other visual features, a visual complexity evaluation model is constructed, wherein the visual complexity evaluation model uses data density, number of color types, overlapping area ratio and visual confusion index as input variables; The multivariate linear regression method was used to fit the visual complexity function and calculate the visual complexity scores of different granularity combinations; The information complexity of different granularity combinations is calculated by information entropy, and the information complexity is combined with the visual complexity score to comprehensively evaluate the complexity of each granularity combination, and a genetic algorithm is used to optimize the visual layout; The optimized layout scheme is visually rendered, multiple charts are arranged using the grid layout function, and the grid structure is dynamically adjusted according to the number and importance of charts to obtain visual analysis results.

7. The method for business process data analysis and visualization according to claim 1, characterized in that: The step S6 specifically includes: Extracting a feature vector of the quality management business process from the visual analysis result, wherein the feature vector includes a data type, a time span, a spatial range, and an attribute dimension; The K-means clustering algorithm is used to classify business processes with similar characteristics, and the silhouette coefficient method is used to determine the optimal number of clusters to obtain the classification results of business characteristics; A CART decision tree model is constructed based on the classification results of the business characteristics, and the clustering label and the original feature vector are combined as input features to identify and analyze the type; According to the identified analysis target, a matching basic template is selected from the preset visualization configuration template library, and the template parameters are optimized through the fuzzy comprehensive evaluation method; The optimized visualization configuration parameters are applied to the original data, and an adaptive rendering algorithm is used to generate visualization charts. The appropriate chart type is dynamically selected according to the data characteristics, and the visual encoding method is adjusted to output the final visualization analysis results that meet the business characteristics and analysis objectives.

8. The business process data analysis visualization method according to claim 1, characterized in that: The step S7 specifically includes: Extracting business process features from the final visualization analysis results, wherein the business process features include time series patterns, spatial distribution features, and multi-dimensional attribute relationships; The STL decomposition method is used to separate the trend, seasonality and residual components of the time series, and the decomposition results of the time series are obtained. The Moran's I index is calculated based on the decomposition results to obtain the key attribute set. Constructing a multi-objective optimization model for the key attribute set, wherein the multi-objective optimization model takes information entropy, interpretability and computational complexity as optimization objectives; Solving the multi-objective optimization model by particle swarm algorithm to obtain the optimal granularity combination; Reconstructing the time series data according to the optimal granularity combination to obtain optimized time series data; The sliding window technology is used to monitor the characteristics of the optimized time series data, and to update the constraints and objective function of the multi-objective optimization model.

Images