Enterprise multi-modal data fusion visualization method and system

Abstract

The present application relates to the technical field of data processing, in particular to a business multi-modal data fusion visualization method and system, comprising: obtaining information data of each dimension under each kind of business data in the business multi-modal data; obtaining associated characteristic factors between each dimension and other dimensions under each kind of business data, obtaining the visualization attention degree of each dimension under each kind of business data by comprehensively considering the calling times of the information data and the time information of the calling request response; obtaining the eye-catching degree of different visualization modes in the historical records, obtaining the membership degree according to the information data of each dimension under each kind of business data contained in each visualization mode in the historical records; obtaining the priority degree of the information data for each visualization mode by combining the visualization attention degree, the membership degree and the eye-catching degree; and determining the best visualization mode of the information data according to the priority degree. The present application can adaptively determine the visualization mode with better visualization effect of the business data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for enterprise multimodal data fusion and visualization. Background Technology

[0002] In modern enterprise management, the ability to make data-driven decisions has become a crucial component of corporate competitiveness. Enterprise operations involve multiple data sources, including sales data, product data, and financial data, which typically possess different modalities. Enterprise multimodal data fusion visualization generally refers to collecting data from various data sources, performing initial cleaning and integration, then using data analysis methods for data mining and statistical analysis. The processed data is stored in a database, and finally, it is presented using different visualization methods.

[0003] Existing methods typically use fixed visualization techniques to display different categories of enterprise data based on experience or historical records. This approach does not consider the correlation between data and the display effects of different visualization methods. Although it improves the efficiency of data visualization to some extent, the visualization effect of enterprise data is poor. Summary of the Invention

[0004] To address the technical problem of poor visualization effects of existing methods for various types of enterprise data, the present invention aims to provide a method and system for enterprise multimodal data fusion and visualization. The specific technical solution adopted is as follows:

[0005] In a first aspect, the present invention provides a method for enterprise multimodal data fusion and visualization, the method comprising:

[0006] Obtain information data for each dimension under each type of business data in the enterprise's multimodal data, as well as the number of times the information data is invoked and the time information of the invoked request response;

[0007] Based on the differences in the changing trends and degrees of change between the information data of each dimension under each type of business data and the information data of other dimensions of the same type of business data, the correlation feature factors between each dimension and other dimensions under each type of business data are obtained.

[0008] Based on the correlation feature factors between each dimension under each type of business data and all other dimensions under the same type of business data, and by combining the data distribution of the number of times the information data is called and the time information of the call request response, the visual attention of each dimension under each type of business data is obtained.

[0009] To obtain the attractiveness of different visualization methods in the historical record, based on the distribution and time of information data in each dimension of each business data contained in each visualization method in the historical record, the membership degree of the information data in each visualization method is obtained;

[0010] By combining the visualization attention, the membership degree of the information data in the corresponding dimension of each visualization method, and the eye-catching degree, the priority of the information data for each visualization method is obtained; the optimal visualization method for the information data is determined based on the priority degree.

[0011] Preferably, the step of obtaining the correlation feature factor between each dimension and other dimensions under each type of business data based on the differences in the changing trends and the differences in the degree of change between the information data of each dimension under each type of business data and the information data of other dimensions of the same type of business data specifically includes:

[0012] Take any dimension of information data under any type of business data as the target dimension information data, and take other dimensions of information data under the same type of business data as the target dimension information data as the reference dimension information data.

[0013] Based on the difference between the data volatility of the target dimension and the data volatility of each reference dimension, determine the first difference factor between the target dimension and each reference dimension.

[0014] Based on the difference between the changing trends of information data in the target dimension and the changing trends of information data in each reference dimension, a second difference factor between the target dimension and each reference dimension is determined.

[0015] Based on the first difference factor and the second difference factor, the correlation feature factors between the target dimension and each reference dimension are obtained, and both the first difference factor and the second difference factor are negatively correlated with the correlation feature factors.

[0016] Preferably, determining the second difference factor between the target dimension and each reference dimension based on the difference between the changing trend of the information data in the target dimension and the changing trend of the information data in each reference dimension specifically includes:

[0017] Based on the distribution of all information data in the target dimension, determine each first extreme point of the target dimension; for any reference dimension, based on the distribution of all information data in the reference dimension, determine each second extreme point of the reference dimension.

[0018] The slope of the line connecting each two adjacent first extreme points in the target dimension is recorded as the first slope value of the target dimension; the slope of the line connecting each two adjacent second extreme points in the reference dimension is recorded as the second slope value of the reference dimension.

[0019] A second difference factor between the target dimension and any one of the reference dimensions is determined based on the average difference between each first slope value of the target dimension and each second slope value corresponding to the reference dimension.

[0020] Preferably, the step of obtaining the visual attention of each dimension under each type of business data based on the correlation feature factors between each dimension and all other dimensions under the same type of business data, and by comprehensively considering the data distribution of the number of times the information data is invoked and the time information of the invoked request response, specifically includes:

[0021] Based on the data distribution of the number of times all information data in the target dimension is invoked and the time information of the invoked request response, the data invocation factor of the target dimension is obtained;

[0022] The data correlation factor of the target dimension is obtained by averaging the correlation feature factors between the target dimension and all reference dimensions.

[0023] The visual attention level of the target dimension is obtained based on the data retrieval factor and the data association factor of the target dimension; the data retrieval factor is positively correlated with the visual attention level, and the data association factor is negatively correlated with the visual attention level.

[0024] Preferably, obtaining the data invocation factor for the target dimension based on the data distribution of the number of times all information data in the target dimension is invoked and the time information of the invocation request response specifically includes:

[0025] The time information of the target dimension's information data includes the time interval between the request and response each time the target dimension's information data is invoked; the average of all time intervals corresponding to the target dimension's information data is calculated to obtain the target dimension's time feature value;

[0026] The data retrieval factor for the target dimension is determined by the ratio between the sum of the number of times all information data in the target dimension is retrieved and the time feature value. The data retrieval factor is a normalized numerical value.

[0027] Preferably, the step of obtaining the membership degree of the information data in each visualization method based on the distribution and time of information data in each dimension of each business data contained in each visualization method in the historical record specifically includes:

[0028] For any visualization method, the total number of information data in the target dimension in the information data displayed by that visualization method is recorded as the quantity feature value;

[0029] The mean of the time interval between two consecutive displays of information data in the target dimension in the information data displayed by this visualization method is recorded as the interval feature value;

[0030] The product of the proportion of the quantitative feature value and the negative correlation coefficient of the interval feature value is used as the membership degree of the information data of the target dimension in this visualization method.

[0031] Preferably, the step of combining the visualization attention and the membership degree of the information data in the corresponding dimension of each visualization method with the eye-catching degree to obtain the priority of the information data for each visualization method specifically includes:

[0032] For any visualization method, calculate the L2 norm between the visualization attention of the target dimension and the membership degree of the information data of the target dimension in the visualization method. The normalized result of the product between the L2 norm and the eye-catchingness of the visualization method is used as the priority of the information data of the target dimension to the visualization method.

[0033] Preferably, the step of obtaining the eye-catching nature of different visualization methods in historical records specifically includes:

[0034] The eye-catching factor of each visualization method is obtained using a neural network. The neural network is trained by manually labeling the network, and the eye-catching factor ranges from (0,1).

[0035] Preferably, the business data includes sales data, product data, and financial data; the product data includes product sales quantity data, product sales revenue data, and product inventory data; each type of business data contains one type of data corresponding to one dimension of information data.

[0036] Secondly, the present invention provides an enterprise multimodal data fusion visualization system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the computer program, when executed by the processor, implements the steps of an enterprise multimodal data fusion visualization method.

[0037] The embodiments of the present invention have at least the following beneficial effects:

[0038] This invention acquires various types of information data and obtains information on the frequency and timing of data retrieval, providing a data foundation for subsequent feature analysis. First, it quantifies the correlation characteristic factors between different dimensions of information data within the same type of business data, reflecting the degree of correlation between each dimension and other dimensions based on differences in the trends and extent of information data changes. Then, by combining the correlation characteristic factors and the distribution of information data retrieval across various aspects, it determines visualization attention. Visualization attention comprehensively characterizes the importance of the corresponding dimension of information data, as well as its priority and attention for visualization, from both the perspectives of correlation and data usage. Furthermore, it quantifies the eye-catching appeal of each visualization method, reflecting the quality and popularity of the visualization effect, and analyzes the distribution and timing of information data for each dimension in the historical visualization results to determine membership. This membership characterizes the degree of membership of each dimension of information data when visualized using a particular visualization method, and the quality of its effect. Ultimately, by combining the results of these feature analyses, we can determine the priority of displaying information data in each dimension using a specific visualization method. This allows us to adaptively determine the best visualization method for each type of data, resulting in better visualization effects. A more suitable visualization format can more intuitively display the characteristics and trends of the data, helping decision-makers make more scientific and rational decisions. Attached Figure Description

[0039] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of the steps of an enterprise multimodal data fusion and visualization method provided by the present invention;

[0041] Figure 2 This is a flowchart of the steps for obtaining the correlation feature factors between the target dimension and the reference dimension provided by the present invention;

[0042] Figure 3 This is a schematic diagram of the first fitted curve of the product sales quantity data provided by the present invention;

[0043] Figure 4 This is a schematic diagram of the second fitted curve of the product sales data provided by the present invention;

[0044] Figure 5 This is a flowchart of the steps of the target dimension visualization attention acquisition method provided by the present invention. Detailed Implementation

[0045] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the enterprise multimodal data fusion and visualization method and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0047] The following description, in conjunction with the accompanying drawings, details the specific solution of the enterprise multimodal data fusion and visualization method and system provided by this invention.

[0048] Please see Figure 1 The diagram illustrates a flowchart of a multimodal data fusion and visualization method for enterprises according to an embodiment of the present invention. The method includes the following steps:

[0049] Step S100: Obtain information data for each dimension under each type of business data in the enterprise multimodal data, as well as the number of times the information data is called and the time information of the time of the call request response.

[0050] In enterprise multimodal data, business data comes in a variety of forms. Visualizing various types of business data during numerical analysis can provide a more intuitive way to analyze data trends. Different types of business data require different visualization methods in different scenarios, rather than fixed visualization methods, to ensure the visualization effect of enterprise data.

[0051] In this embodiment, data feature analysis and visualization effect analysis are performed on the multiple dimensions of information data contained in each type of business data, providing a data foundation for subsequent data analysis. In this embodiment, business data includes sales data, product data, and financial data. Each type of business data contains multiple dimensions of information data to present the information related to the enterprise's business in a more detailed manner.

[0052] For example, product data includes product sales quantity data, product sales revenue data, and product inventory data. Sales data includes sales order quantity, sales order amount, and sales price. Financial data includes profit data, cost data, and asset data. Under each type of business data, one type of data corresponds to one dimension of information data; that is, under the product data category, product sales quantity data belongs to one dimension of information data.

[0053] It should be noted that the business data of the enterprise in this embodiment needs to be acquired in compliance with relevant industry regulations and business privacy. Furthermore, the data acquired is data within a certain time period from the database, meaning the data has a certain temporal sequence. Data is collected at fixed time intervals, and the implementer can set the length of the time period according to the specific implementation scenario, such as 30 days. This means that in this embodiment, one piece of information data can be collected every day. Simultaneously, to prevent the impact of dimensional issues on data analysis results when analyzing different data, this embodiment performs data preprocessing on each piece of information data. The preprocessing operation may include standardization, and the standardization method is a well-known technique, which will not be described in detail here.

[0054] Furthermore, in the process of enterprise data visualization, information data of different dimensions under different types of business data should have different response priorities. Information data with higher priority occupies a larger proportion in the actual access and retrieval process, and thus this type of information data is relatively more important. Therefore, it is also necessary to analyze the retrieval of information data, which can reflect the usage of this type of information data.

[0055] Specifically, in this embodiment, information on the retrieval of all information data under each dimension within a certain time period is obtained. This includes the number of times the information data is retrieved and the time information of the request and response. Specifically, this includes the total number of times all information data under a certain dimension is retrieved and the time interval between the request and response for each retrieval. For example, the total number of times product sales quantity data under the product data dimension is retrieved, and the time interval between the request and response for each retrieval. It can be understood that querying, reading, and writing data of a certain type of data can all be considered as having a usage operation record for that type of data, and thus can all be recorded as the number of retrieval operations. The retrieved data can be directly obtained through the enterprise's system access data records and log data.

[0056] Step S200: Based on the differences in the changing trends and degrees of change between the information data of each dimension under each type of business data and the information data of other dimensions of the same type of business data, obtain the correlation feature factors between each dimension and other dimensions under each type of business data.

[0057] In the process of visualization priority analysis, there may be multiple data types with the same priority, and there may be certain correlations between the data. When the correlation between different types of data is high, it means that the phenomena reflected by a certain type of data and other types of data are relatively close or similar, which will overshadow the importance of these types of data. That is, compared with the data with low correlation, the visualization importance or priority of this type of data is low. Therefore, it is necessary to first analyze the data correlation between each type of data and other types of data.

[0058] In this embodiment, an arbitrary data type is used as an example. Specifically, information data from any dimension of any type of business data is taken as the target dimension, and information data from other dimensions belonging to the same type of business data as the target dimension are taken as reference dimensions. By analyzing the data differences between the target dimension and each reference dimension in two aspects, the correlation characteristic factors between the target dimension and each reference dimension are determined. The first aspect refers to the difference in the changing trends of the information data between the target dimension and the reference dimensions, and the second aspect refers to the difference in the degree of change of the information data between the target dimension and the reference dimensions.

[0059] It is understood that this embodiment uses product data as an example for explanation. For example, if the dimension containing the product sales quantity data under the product data is taken as the target dimension, then the dimensions containing other types of data under the product data, other than the product sales quantity data, are all reference dimensions.

[0060] In this embodiment, as Figure 2 As shown, the method for obtaining the correlation feature factors between the target dimension and the reference dimension can be implemented by steps S201 to S203.

[0061] Step S201: Based on the difference between the data fluctuation degree of the information data of the target dimension and the data fluctuation degree of the information data of each reference dimension, determine the first difference factor between the target dimension and each reference dimension.

[0062] In this embodiment, an arbitrary reference dimension is used as an example. For instance, the i-th dimension under any type of business data is taken as the target dimension, and the v-th dimension under that type of business data other than the i-th dimension is taken as the arbitrary reference dimension.

[0063] Each data collection point in the target dimension corresponds to one piece of information data, and similarly, each data collection point in the reference dimension corresponds to one piece of information data. By calculating the variance of all information data in the target dimension, the degree of data fluctuation in the target dimension can be reflected. Similarly, by calculating the variance of all information data in the reference dimension, the degree of data fluctuation in the reference dimension can be reflected. That is, this embodiment uses variance to characterize the degree of data fluctuation of a set of data. In other embodiments, implementers can also choose other parameters for characterization according to the specific implementation scenario, such as standard deviation.

[0064] Furthermore, by comparing the differences in data fluctuation between the target dimension and the reference dimension, the correlation between the target dimension and the reference dimension can be quantified. The greater the difference in data fluctuation between the two, the smaller the correlation between them.

[0065] Specifically, the absolute value of the difference between the variance of all information data in the target dimension and the variance of all information data in the reference dimension is taken as the first difference factor between the target dimension and the reference dimension. In other words, the first difference factor characterizes the difference in data fluctuation between the target dimension and the reference dimension. The larger the value, the greater the difference in data fluctuation.

[0066] Step S202: Based on the difference between the changing trend of the information data of the target dimension and the changing trend of the information data of each reference dimension, determine the second difference factor between the target dimension and each reference dimension.

[0067] In this embodiment, the slope of the straight line between local data is used to reflect the local change trend of information data in the corresponding dimension. Then, by analyzing the differences in the change trend between the information data in the target dimension and the information data in the reference dimension, the correlation characteristics of the second aspect are quantitatively characterized.

[0068] First, based on the distribution of all information data in the target dimension, determine each first extreme point of the target dimension; for any reference dimension, based on the distribution of all information data in the reference dimension, determine each second extreme point of the reference dimension. The method for obtaining the extreme points of a set of data is a well-known technique and will not be discussed in detail here.

[0069] More specifically, a Cartesian coordinate system is constructed with the data collection time point as the x-axis and the information data of the target dimension at each data collection time point as the y-axis. Data fitting is then performed to obtain a first fitted curve. Simultaneously, a Cartesian coordinate system is constructed with the data collection time point as the x-axis and the information data of the reference dimension at each data collection time point as the y-axis. Data fitting is then performed to obtain a second fitted curve. Data analysis of the first and second fitted curves respectively yields the first extreme point of the target dimension and the second extreme point of the reference dimension. For example... Figure 3 It shows the first fitted curve for product sales volume data, where i represents the i-th dimension of the business data, such as... Figure 4 It shows the second fitted curve of product sales data, where v represents the v-th dimension of business data, i ≠ v, and i and v belong to the same type of business data.

[0070] Then, the slope of the line connecting each pair of adjacent first extreme points in the target dimension is calculated and recorded as the first slope value of the target dimension; the slope of the line connecting each pair of adjacent second extreme points in the reference dimension is calculated and recorded as the second slope value of the reference dimension. That is, each pair of adjacent first extreme points corresponds to a first slope value, which reflects the local change trend of the first fitted curve at that position; and each pair of adjacent second extreme points corresponds to a second slope value, which reflects the local change trend of the second fitted curve at that position.

[0071] For example, Figure 3 In the middle, K i,1 This represents the first slope value corresponding to the first local position of the i-th dimension, i.e. the target dimension, under the business data. It is also the first slope value. As can be seen from the figure, this first slope value reflects the general trend of data change from 0 to 3 days. Figure 4 Zhong K v,1 This represents the second slope value corresponding to the first local position of the v-th dimension (i.e., the reference dimension) under the business data. It is also the first second slope value. As can be seen from the figure, this second slope value reflects the general trend of data changes from 0 to 3 days.

[0072] Finally, based on the average difference between each first slope value of the target dimension and each second slope value corresponding to the reference dimension, a second difference factor between the target dimension and any one of the reference dimensions is determined.

[0073] The first slope value reflects the local variation trend among information data in the target dimension, while the second slope value reflects the local variation trend among information data in the reference dimension. By comparing the differences between the first and second slope values ​​at corresponding positions, the differences in local variation trends between the target dimension and the reference dimension can be characterized. Furthermore, by averaging the data, the second difference factor ultimately characterizes the differences in local variation trends between the target dimension and the reference dimension. The larger the value, the weaker the correlation between the target dimension and the reference dimension.

[0074] It should be noted that when the number of the first slope value and the second slope value are different, the slope value with the fewest values ​​is used as the reference for difference analysis. Also, since the information data in this embodiment has been standardized and collected on a daily basis, therefore... Figure 3 and Figure 4 The vertical axis in the graph has no unit, and the horizontal axis represents time, with the unit being days.

[0075] Step S203: Based on the first difference factor and the second difference factor, obtain the correlation feature factor between the target dimension and each reference dimension, wherein both the first difference factor and the second difference factor are negatively correlated with the correlation feature factor.

[0076] The first difference factor reflects the difference between the target dimension and the reference dimension in terms of volatility, while the second difference factor reflects the difference between the target dimension and the reference dimension in terms of trend. By combining the data difference characteristics of the two aspects, the correlation characteristic factor between the target dimension and the reference dimension can be quantified. That is, the greater the difference, the smaller the correlation characteristic factor. Therefore, the difference analysis results should be negatively correlated.

[0077] As a concrete example, the formula for calculating the correlation feature factor between the target dimension and a reference dimension can be expressed as:

[0078]

[0079] Among them, W i,v This represents the correlation feature factor between the target dimension and the v-th reference dimension, where i represents the i-th dimension under any type of business data, and v represents the v-th reference dimension belonging to the same type of business data as the i-th dimension, i ≠ v; K i,n K represents the nth first slope value in the target dimension. v,n This represents the nth slope value in the reference dimension, N1 represents the minimum number of slope values ​​in the target and reference dimensions, and σ i σ represents the degree of data fluctuation in the target dimension, which is also the variance of all information data under the target dimension;v It represents the degree of data fluctuation in the reference dimension, that is, the variance of all information data under the reference dimension. exp represents an exponential function with the natural constant e as the base.

[0080] |σ i -σ v | indicates the first difference factor. The second difference factor is represented by a negative exponentiation exp(-x) to negatively correlate the product of the first and second difference factors. The correlation feature factor characterizes the data correlation characteristics between the target dimension and the reference dimension. The larger the value, the more similar the fitting curves of the target dimension and the reference dimension are not only in terms of change trend, but also in terms of fluctuation degree, thus indicating that the correlation between the target dimension and the reference dimension is greater.

[0081] Step S300: Based on the correlation feature factors between each dimension under each type of business data and all other dimensions under the same type of business data, and by comprehensively considering the data distribution of the number of times the information data is called and the time information of the call request response, the visual attention of each dimension under each type of business data is obtained.

[0082] In the process of analyzing the visualization priority of information data for each dimension under each type of business data, the more times information is called, the higher the frequency of use of that type of information data, and thus the greater its importance. This reflects, to some extent, the urgency of visualizing that type of information data. The higher the value of the correlation feature factor between the target dimension and other dimensions, the stronger the correlation between that type of information data and other dimensions. This indicates a stronger substitutability of that type of data; that is, other dimensions with a strong correlation can replace the data of this dimension for feature analysis when conducting enterprise data analysis. Consequently, the importance is lower, and the urgency of visualizing that type of information data is lower. Based on this, this embodiment quantifies the visualization attention of information data under the corresponding dimension by combining the correlation feature factor and the information distribution of information data called for the corresponding dimension.

[0083] In this embodiment, we will still use information data of any dimension under any type of business data as an example for explanation, such as... Figure 5 As shown, the method for obtaining the visual attention of the target dimension can be implemented by steps S301 to S303.

[0084] Step S301: Based on the data distribution of the number of times all information data in the target dimension is called and the time information of the call request response, obtain the data call factor of the target dimension.

[0085] The more frequently all information data under the target dimension is accessed, the more frequently the corresponding data type of the target dimension is retrieved, analyzed, and used by relevant personnel within the enterprise, thus indicating a higher degree of importance for the corresponding data type of the target dimension. The information regarding the access of all information data under the target dimension includes two aspects, as described in step S100. The first aspect includes the number of times the data is accessed, specifically the total number of times all information data under the target dimension is accessed. The second aspect includes the time information of the access request and response, specifically the time interval between the request and response each time all information data under the target dimension is accessed.

[0086] Under the target dimension, there is a time interval each time information data is called. By comprehensively analyzing the time intervals of all call operations, we can reflect the data call response of the corresponding type in the target dimension. That is, we can calculate the average of all time intervals corresponding to the information data in the target dimension to obtain the time feature value of the target dimension. This time feature value reflects the frequency of use of the corresponding type of data in the target dimension.

[0087] Furthermore, the data retrieval factor for the target dimension is determined by the ratio between the sum of the number of times all information data in the target dimension is retrieved and the time feature value. This data retrieval factor is a normalized value. In this embodiment, the result of normalizing this ratio is used as the data retrieval factor for the target dimension. The normalization method is a well-known technique and will not be described in detail here. The data retrieval factor characterizes the retrieval frequency and usage of information data in the target dimension. The more times the information data in the target dimension is retrieved and the shorter the response time, the larger the corresponding data retrieval factor value, indicating that the data is used more frequently or retrieved more often.

[0088] Step S302: Obtain the average value of the correlation feature factors between the target dimension and all reference dimensions to obtain the data correlation factor of the target dimension.

[0089] There is a correlation feature factor between the target dimension and each reference dimension, which reflects the correlation characteristics of the information data between the target dimension and each reference dimension. The average value of all correlation feature factors corresponding to the target dimension is used as the data correlation factor of the target dimension. The data correlation factor integrates the correlation characteristics between the target dimension and all other reference dimensions, and reflects the magnitude of the correlation in the overall comprehensive situation of the information data of the target dimension.

[0090] Step S303: Based on the data retrieval factor and the data association factor of the target dimension, obtain the visual attention of the target dimension.

[0091] Specifically, the data retrieval factor of the target dimension is positively correlated with the visualization attention, while the data association factor of the target dimension is negatively correlated with the visualization attention. In this embodiment, the formula for calculating the visualization attention of the target dimension can be expressed as:

[0092]

[0093] Among them, S i F represents the visual attention level of the target dimension. i C represents the data retrieval factor for the target dimension, and W represents the total number of all reference dimensions corresponding to the target dimension. i,v The denoting feature factor represents the correlation between the target dimension and the v-th reference dimension, where i represents the i-th dimension under any type of business data, and v represents the v-th reference dimension that belongs to the same type of business data as the i-th dimension, i≠v; exp() represents an exponential function with the natural constant e as the base.

[0094] Data call factor F i The larger the value, the more frequently the data is used; the higher the data association factor. The smaller the value of S, the weaker the correlation between the information data in the target dimension and the information data in other dimensions, the less substitutable the information data in the target dimension, and the greater the importance of the corresponding data, thus increasing the visual attention S of the target dimension. i The larger the value, the greater the visual attention level of the target dimension, representing the level of attention, importance, and priority given to the visualization of information and data within that target dimension.

[0095] Step S400: Obtain the attractiveness of different visualization methods in the historical records. Based on the distribution and time of information data in each dimension of each business data contained in each visualization method in the historical records, obtain the membership degree of the information data in each visualization method.

[0096] For businesses, common data visualization methods include line charts, bar charts, pie charts, scatter plots, heatmaps, and 3D graphs, among others. Different visualization methods offer different visual effects and enjoy varying degrees of popularity. When different visualization methods are used to visualize information from business data, employees show varying levels of interest and preference for the different results. Historical records show that each dimension of information under each type of business data has been visualized. By analyzing the distribution of different types of information data under each visualization method, and by reviewing historical records, we can understand the visual effects and popularity of different visualization methods. Ultimately, we can quantify the degree to which each type of information data is visualized using each visualization method.

[0097] First, it is necessary to evaluate the visualization effects of different visualization methods. In this embodiment, this is achieved through a neural network. Specifically, the neural network is used to obtain the eye-catching value of each visualization method. The neural network is trained using manual labeling, and the eye-catching value ranges from (0,1). The eye-catching value characterizes the quality of the visualization effect and the degree of popularity of the corresponding visualization method.

[0098] More specifically, in this embodiment, the training information for the neural network includes: feature information for different visualization methods, including bar charts, pie charts, line charts, scatter plots, and heatmaps; each feature is scored manually, with higher scores indicating greater visual appeal; the neural network is a five-layer fully connected neural network model with a classification network structure and a cross-entropy loss function; the training ratio is set to 7:3, and the visual appeal of the visualization is trained using the fully connected network model. The input to the classification network is bar charts, pie charts, line charts, scatter plots, and heatmaps, and the output is the visual appeal score for each visualization method.

[0099] During the training of the classification network, gradient descent is used for training until the loss function converges, which completes the training of the classification network. The value range of the eye-catching index is (0,1). The eye-catching index represents the effect and popularity of the corresponding visualization method. The higher the eye-catching index, the higher the popularity of the visualization display form and the better the visualization effect.

[0100] Then, in the historical records, each visualization method contains a variety of information data. By statistically analyzing the proportion of information data under each visualization method, we can reflect the amount of data that is visualized using each visualization method in the historical records. The larger the proportion of data, the more it indicates that this type of information data tends to be displayed using this visualization method in the historical records.

[0101] Specifically, for any visualization method, the total number of information data in the target dimension displayed by that visualization method is recorded as the quantity feature value; the average of the time intervals between two consecutive displays of information data in the target dimension displayed by that visualization method is recorded as the interval feature value. It should be noted that this time interval records the time interval between two consecutive visualizations of information data in the target dimension using the same visualization method in the historical record. For example, in the historical record, a certain visualization method was used for 100 visualizations. The second, sixth, and seventh visualizations all visualized information data in the target dimension. Each visualization was recorded with a time, and thus the time interval between the second and sixth visualizations, and the time interval between the sixth and seventh visualizations, can be obtained. The average of these two time intervals is the interval feature value, which reflects the frequency of visualization of information data in the target dimension under this visualization method in terms of time distribution.

[0102] Furthermore, the product of the proportion of the quantitative feature value and the negative correlation coefficient of the interval feature value is used as the membership degree of the information data in the target dimension in this visualization method. As a specific example, taking any visualization method as an example, the formula for calculating the membership degree of the information data in the target dimension in the u-th visualization method can be expressed as:

[0103]

[0104] Among them, Gh u,i G represents the membership degree of the target dimension information data in the u-th visualization method. u,i G represents the total amount of information data in the target dimension contained in the information data displayed by the u-th visualization method, i.e., the quantity feature value; u ΔT represents the total amount of information data displayed in the u-th visualization method; ΔT represents the average time interval between two consecutive displays of information data in the target dimension in the u-th visualization method, i.e., the interval feature value; i represents the i-th dimension under any type of business data, which is the target dimension.

[0105] This indicates the percentage of information data in the target dimension within the information data displayed using the u-th visualization method. The larger the value of , the more target dimension information data is contained in the information data displayed using the u-th visualization method, and the higher the visualization frequency. This indicates that in actual enterprise data visualization history, the target dimension information data is better visualized using the u-th visualization method, and has a relatively high acceptance rate among enterprise personnel. Therefore, Ghu,i The larger the value of , the higher the membership degree of the target dimension's information data to the u-th visualization method.

[0106] Membership degree characterizes the degree of membership of information data of this type when displayed using the u-th visualization method. The larger the value, the greater the degree of membership, and thus the better the effect of displaying information data of this type using the u-th visualization method.

[0107] Step S500: Combining the visualization attention and the membership degree of the information data in the corresponding dimension of each visualization method, as well as the eye-catching degree, the priority of the information data for each visualization method is obtained; the optimal visualization method for the information data is determined based on the priority degree.

[0108] Under different visualization methods, the visualization effects of information data of different dimensions under different types of business data are different. In order to determine the best visualization display method, each visualization method has a certain priority for each type of information data. In order to more accurately evaluate the priority, this embodiment comprehensively considers two aspects of data characteristics and quantifies the degree of tendency of the visualization effect of the visualization method.

[0109] Firstly, by analyzing the degree of attention paid to the visualization of each type of information data within a certain period, we can reflect the attention, importance, and priority of visualizing that type of data. If the correlation between information data is poor and it is called more frequently within a certain period, then the higher the importance of the corresponding type of information data, the greater the priority for visualization.

[0110] Secondly, by analyzing the proportion and temporal distribution of data under each different visualization method in historical records, we construct the relationship between visualization methods and various types of information data. We use membership degrees to reflect the degree of preference of each type of information data for each visualization method. The higher the membership degree, the better the visualization effect of that type of information data using the corresponding visualization method, and the greater its priority for visualization.

[0111] Based on this, for any visualization method, the L2 norm between the visualization attention of the target dimension and the membership degree of the information data of the target dimension in the visualization method is calculated, and the normalized result of the product between the L2 norm and the eye-catchingness of the visualization method is used as the priority of the information data of the target dimension to the visualization method.

[0112] As a concrete example, taking any visualization method as an illustration, the formula for calculating the priority of the target dimension information data with respect to the u-th visualization method can be expressed as:

[0113]

[0114] Among them, R i,u α represents the priority of the target dimension's information data for the u-th visualization method. u S represents the eye-catching power of the u-th visualization method. i Gh represents the visual attention level of the target dimension. u,i represents the membership degree of the target dimension information data in the u-th visualization method, and th() represents the hyperbolic tangent function.

[0115] use By fusing the features of visualization attention and membership, a higher priority is given to the u-th visualization method when both values ​​are larger. The attention-grabbing factor α corresponding to the chosen visualization method is then calculated. u As a coefficient for feature synthesis, the higher the eye-catching value, the better the visualization effect of the corresponding visualization method, which in turn indicates that the visualization method has a higher priority for data processing.

[0116] Priority indicates the degree to which information data is visualized using a corresponding visualization method. The higher the value, the better the effect of visualizing the information data using that visualization method, and the higher the priority.

[0117] At this point, each dimension of information data under each type of business data has a corresponding priority level among each visualization method, reflecting the effectiveness of processing the corresponding data using the corresponding visualization method. Visualization methods can be filtered for each type of information data separately.

[0118] For example, for information data in the target dimension, the visualization method with the highest priority among all visualization methods is taken as the best visualization method for the information data in the target dimension.

[0119] Visualizing data of corresponding types using optimal visualization methods achieves the best visualization results and can, to some extent, avoid system latency caused by large amounts of data, thus improving system performance and visualization effectiveness. Appropriate visualization methods help businesses more effectively understand data, reveal patterns and trends, and support data-driven decision-making.

[0120] This invention also provides an enterprise multimodal data fusion visualization system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When executed by the processor, the computer program implements the steps of an enterprise multimodal data fusion visualization method. Since an embodiment of an enterprise multimodal data fusion visualization method has already been described in detail, the process will not be repeated here.

[0121] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for enterprise multimodal data fusion and visualization, characterized in that, The method includes the following steps: Obtain information data for each dimension under each type of business data in the enterprise's multimodal data, as well as the number of times the information data is invoked and the time information of the invoked request response; Based on the differences in the changing trends and degrees of change between the information data of each dimension under each type of business data and the information data of other dimensions of the same type of business data, the correlation feature factors between each dimension and other dimensions under each type of business data are obtained. Based on the correlation feature factors between each dimension under each type of business data and all other dimensions under the same type of business data, and by combining the data distribution of the number of times the information data is called and the time information of the call request response, the visual attention of each dimension under each type of business data is obtained. To obtain the attractiveness of different visualization methods in the historical record, based on the distribution and time of information data in each dimension of each business data contained in each visualization method in the historical record, the membership degree of the information data in each visualization method is obtained; By combining the visualization attention, the membership degree of the information data in the corresponding dimension of each visualization method, and the eye-catching degree, the priority of the information data for each visualization method is obtained; and the optimal visualization method for the information data is determined based on the priority degree. The method for obtaining the visualized attention includes: Take any dimension of information data under any type of business data as the target dimension information data, and take other dimensions of information data under the same type of business data as the target dimension information data as the reference dimension information data. Based on the data distribution of the number of times all information data in the target dimension is invoked and the time information of the invoked request response, the data invocation factor of the target dimension is obtained; The data correlation factor of the target dimension is obtained by averaging the correlation feature factors between the target dimension and all reference dimensions. The visual attention level of the target dimension is obtained based on the data retrieval factor and the data association factor of the target dimension; the data retrieval factor is positively correlated with the visual attention level, and the data association factor is negatively correlated with the visual attention level. The methods for obtaining the membership degree include: For any visualization method, the total number of information data in the target dimension in the information data displayed by that visualization method is recorded as the quantity feature value; The mean of the time interval between two consecutive displays of information data in the target dimension in the information data displayed by this visualization method is recorded as the interval feature value; The product of the proportion of the quantitative feature value and the negative correlation coefficient of the interval feature value is used as the membership degree of the information data of the target dimension in this visualization method. The calculation formula is as follows: ； in, This represents the membership degree of the information data in the target dimension in the u-th visualization method. This represents the total number of information data in the target dimension contained in the information data displayed by the u-th visualization method, i.e., the quantity feature value; This represents the total amount of information data displayed using the u-th visualization method; The interval characteristic value represents the average time interval between two consecutive displays of information data in the target dimension in the information data displayed by the u-th visualization method; i represents the i-th dimension under any type of business data, which is the target dimension.

2. The enterprise multimodal data fusion and visualization method according to claim 1, characterized in that, The method involves obtaining correlation feature factors between each dimension of each type of business data and other dimensions based on the differences in the changing trends and degrees of change between the information data of each dimension under each type of business data and the information data of other dimensions of the same type of business data. Specifically, this includes: Based on the difference between the data volatility of the target dimension and the data volatility of each reference dimension, determine the first difference factor between the target dimension and each reference dimension. Based on the difference between the changing trends of information data in the target dimension and the changing trends of information data in each reference dimension, a second difference factor between the target dimension and each reference dimension is determined. Based on the first difference factor and the second difference factor, the correlation feature factors between the target dimension and each reference dimension are obtained, and both the first difference factor and the second difference factor are negatively correlated with the correlation feature factors.

3. The enterprise multimodal data fusion and visualization method according to claim 2, characterized in that, The step of determining a second difference factor between the target dimension and each reference dimension based on the difference between the changing trends of information data in the target dimension and the changing trends of information data in each reference dimension specifically includes: Based on the distribution of all information data in the target dimension, determine each first extreme point of the target dimension; for any reference dimension, based on the distribution of all information data in the reference dimension, determine each second extreme point of the reference dimension. The slope of the line connecting each two adjacent first extreme points in the target dimension is recorded as the first slope value of the target dimension; the slope of the line connecting each two adjacent second extreme points in the reference dimension is recorded as the second slope value of the reference dimension. A second difference factor between the target dimension and any one of the reference dimensions is determined based on the average difference between each first slope value of the target dimension and each second slope value corresponding to the reference dimension.

4. The enterprise multimodal data fusion and visualization method according to claim 1, characterized in that, The data invocation factor for the target dimension is obtained by analyzing the data distribution based on the number of times all information data in the target dimension is invoked and the time information of the invocation requests and responses. Specifically, this includes: The time information of the target dimension's information data includes the time interval between the request and response each time the target dimension's information data is invoked; the average of all time intervals corresponding to the target dimension's information data is calculated to obtain the target dimension's time feature value; The data retrieval factor for the target dimension is determined by the ratio between the sum of the number of times all information data in the target dimension is retrieved and the time feature value. The data retrieval factor is a normalized numerical value.

5. The enterprise multimodal data fusion and visualization method according to claim 2, characterized in that, The priority of the information data for each visualization method is obtained by combining the visualization attention, the membership degree of the information data in the corresponding dimension of each visualization method, and the eye-catching degree, specifically including: For any visualization method, calculate the L2 norm between the visualization attention of the target dimension and the membership degree of the information data of the target dimension in the visualization method. The normalized result of the product between the L2 norm and the eye-catchingness of the visualization method is used as the priority of the information data of the target dimension to the visualization method.

6. The enterprise multimodal data fusion and visualization method according to claim 1, characterized in that, The method of determining the attractiveness of different visualization methods in historical records specifically includes: The eye-catching factor of each visualization method is obtained using a neural network. The neural network is trained by manually labeling the network, and the eye-catching factor ranges from (0,1).

7. The enterprise multimodal data fusion and visualization method according to claim 1, characterized in that, The business data includes sales data, product data, and financial data; the product data includes product sales quantity data, product sales revenue data, and product inventory data; each type of business data contains one type of data corresponding to one dimension of information data.

8. An enterprise multimodal data fusion visualization system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the enterprise multimodal data fusion and visualization method as described in any one of claims 1-7.

Images