Interactive behavior big data mining method and system applied to visual database

Abstract

The present invention relates to the fields of big data, artificial intelligence, visualization and database technology, and provides an interactive behavior big data mining method and system applied to a visual database. Through the present invention, a visualization operation preference analysis report is determined by analyzing and processing from multiple focus areas, which can improve the accuracy of the visualization operation preference analysis report; and when the image description is changed, multiple quantitative image description members are changed into linked non-quantitative image description members, which can avoid the error accumulation when the operation preference is analyzed based on the image description later as much as possible, so that the accuracy and credibility of the visualization operation preference analysis report can be improved, so as to provide a decision-making basis for upgrading and optimizing the operation interaction function of the visualization database or customizing the personalized operation interface.

Description

Technical Field

[0001] The present invention relates to the fields of big data, artificial intelligence, visualization and database technology, and in particular to an interactive behavior big data mining method and system applied to a visualization database. Background Art

[0002] With the development of GUI (Graphical User Interface) technology, its vivid and friendly graphical interface has been widely used in many fields. In the field of database, database designers often spend a lot of energy on good data definition. Similarly, users also have to worry about the strict and boring data manipulation language. In order to solve the above problems, the visual database system based on the graphical user interface constitutes another direction of database development.

[0003] With the continuous maturity of big data and artificial intelligence, the number and types of interactive tasks that visual database systems deal with are increasing day by day. Since a series of interactive functions of visual database systems are implemented based on graphical interfaces, the operation behavior data of users when using the visual database system for database interaction is also in the form of images. In order to achieve the upgrade and optimization of the visual database system and further improve the interactive intelligence of the visual database system, it is very important to accurately and reliably mine the user's operation behavior preferences. Summary of the invention

[0004] The present invention at least provides a method and system for mining interactive behavior big data applied to a visual database.

[0005] The technical solution of the present invention is achieved through at least some of the following embodiments.

[0006] A method for mining interactive behavior big data applied to a visual database, applied to a big data analysis system, the method comprising:

[0007] Obtaining database operation behavior image records that trigger the visual interaction behavior parsing mechanism;

[0008] Performing image descriptor mining processing on the database operation behavior image record to obtain an image wavelet descriptor of the database operation behavior image record; wherein the image wavelet descriptor includes a plurality of quantized image description members;

[0009] Performing image description change processing on the image wavelet descriptor, changing multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtaining linkage operation behavior description knowledge of the image wavelet descriptor; wherein the linkage operation behavior description knowledge includes multiple non-quantized image description members of different focus areas;

[0010] Performing knowledge clustering operations on the linkage operation behavior description knowledge respectively to obtain linkage clustering results of the image wavelet descriptor, wherein the linkage clustering results include clustering results of the multiple different focus surfaces;

[0011] An aggregation analysis operation is performed on the linkage group results to determine a visualized operation preference analysis report of the database operation behavior image record.

[0012] In some example embodiments, the linkage operation behavior description knowledge includes scenario operation behavior description knowledge, event operation behavior description knowledge and business project operation behavior description knowledge; wherein, the scenario operation behavior description knowledge is used to characterize the interaction scenario corresponding to the database operation behavior image record, the event operation behavior description knowledge is used to characterize the behavior event matched by the communication terminal of the database operation behavior image record, and the business project operation behavior description knowledge is used to characterize the database business project that matches the scenario and the behavior event.

[0013] In some example embodiments, performing knowledge clustering operations on the linkage operation behavior description knowledge to obtain linkage clustering results of the image wavelet descriptor includes:

[0014] For the target focus surface, a first AI learning algorithm is used to process the target focus surface operation behavior description knowledge to obtain average description knowledge of the target focus surface operation behavior description knowledge; wherein the average description knowledge is used to characterize the average positioning variable of the target focus surface operation behavior description knowledge in the knowledge relationship network;

[0015] The target focus surface operation behavior description knowledge is processed by a second AI learning algorithm to obtain deviation description knowledge of the target focus surface operation behavior description knowledge; wherein the deviation description knowledge is used to characterize the deviation positioning variable of the target focus surface operation behavior description knowledge in the knowledge relationship network;

[0016] Using the target disturbance feature to perform disturbance feature addition processing on the deviation description knowledge to obtain processed deviation description knowledge;

[0017] Determine the quasi-cluster operation behavior description knowledge corresponding to the target focus surface operation behavior description knowledge by combining the average description knowledge and the processed deviation description knowledge;

[0018] A knowledge clustering operation is performed on the quasi-clustering operation behavior description knowledge to obtain a target attention surface clustering result of the image wavelet descriptor.

[0019] In some example embodiments, after performing an aggregation analysis operation on the linkage grouping results to determine a visualization operation preference analysis report of the database operation behavior image record, the method further includes:

[0020] According to the visualization operation preference analysis report, a database interface optimization case matching the visualization operation preference analysis report is obtained from a database operation behavior image record set as a candidate database interface optimization case;

[0021] When there are multiple candidate database interface optimization cases, obtaining optimization feedback metric values ​​of each candidate database interface optimization case;

[0022] From the candidate database interface optimization cases, the candidate database interface optimization case whose optimization feedback metric value meets the requirements is selected as the database interface optimization case corresponding to the database operation behavior image record.

[0023] In some example embodiments, the method is implemented by a visualization operation preference parsing network, and the debugging steps of the visualization operation preference parsing network include:

[0024] Acquire at least one debugging learning information of the visualization operation preference parsing network, wherein the debugging learning information includes a past database operation behavior image record and a learning annotation corresponding to the past database operation behavior image record;

[0025] Performing image descriptor mining processing on the image records of the past database operation behaviors to obtain image wavelet descriptors of the image records of the past database operation behaviors; wherein the image wavelet descriptors include a plurality of quantized image description members;

[0026] Performing image description change processing on the image wavelet descriptor, changing multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtaining linkage operation behavior description knowledge of the image wavelet descriptor; wherein the linkage operation behavior description knowledge includes multiple non-quantized image description members of different focus areas;

[0027] Performing knowledge clustering operations on the linkage operation behavior description knowledge respectively to obtain linkage clustering results of the image wavelet descriptor, wherein the linkage clustering results include clustering results of the multiple different focus surfaces;

[0028] Performing aggregation and analysis operations on the linkage group results to determine a visualized operation preference analysis report of the image records of the past database operation behaviors;

[0029] Determine the network learning cost variable of the visualization operation preference analysis network by combining the linkage grouping result, the visualization operation preference analysis report, and the learning annotations corresponding to the past database operation behavior image records;

[0030] The network configuration parameters of the visualization operation preference parsing network are improved in combination with the network learning cost variable.

[0031] In some example embodiments, the linkage grouping results include scenario grouping results, behavior event grouping results, and database business project grouping results;

[0032] The learning annotations include scene cluster learning annotations, behavior event cluster learning annotations and database business item cluster learning annotations corresponding to the past database operation behavior image records, as well as visualization operation preference learning annotations corresponding to the past database operation behavior image records.

[0033] In some example embodiments, combining the linkage clustering result, the visualization operation preference analysis report, and the learning annotations corresponding to the past database operation behavior image records to determine the network learning cost variables of the visualization operation preference analysis network includes:

[0034] Determine the first network learning cost variable corresponding to the visualization operation preference parsing network by combining the scene clustering result, the behavior event clustering result, the database business project clustering result, the scene clustering learning annotation, the behavior event clustering learning annotation and the database business project clustering learning annotation; wherein the first network learning cost indicator is used to characterize the accuracy of the linkage clustering result of the visualization operation preference parsing network;

[0035] Determine a second network learning cost variable corresponding to the visualization operation preference parsing network in combination with the visualization operation preference parsing report and the visualization operation preference learning annotation; wherein the second network learning cost indicator is used to characterize the accuracy of the visualization operation preference parsing report of the visualization operation preference parsing network;

[0036] The first network learning cost variable and the second network learning cost variable are combined to determine the network learning cost variable of the visualization operation preference parsing network.

[0037] In some example embodiments, combining the scene clustering result, the behavior event clustering result, the database business project clustering result, the scene clustering learning annotation, the behavior event clustering learning annotation, and the database business project clustering learning annotation to determine the first network learning cost variable corresponding to the visualization operation preference parsing network includes:

[0038] Determine the network learning cost variable corresponding to the scene cluster component in the visualization operation preference parsing network in combination with the scene clustering result, the scene clustering learning annotation, the behavior event clustering learning annotation and the database business item clustering learning annotation;

[0039] Determine the network learning cost variable corresponding to the behavior event cluster component in the visualization operation preference parsing network by combining the behavior event cluster result, the scene cluster learning annotation, the behavior event cluster learning annotation and the database business project cluster learning annotation;

[0040] Determine the network learning cost variable corresponding to the database business project group component in the visualization operation preference parsing network in combination with the database business project group result, the scenario group learning annotation, the behavior event group learning annotation and the database business project group learning annotation;

[0041] The first network learning cost variable is determined according to the network learning cost variable corresponding to the scenario cluster component, the network learning cost variable corresponding to the behavior event cluster component and the network learning cost variable corresponding to the database business item cluster component.

[0042] In some example embodiments, the method further comprises:

[0043] Based on the set processing strategy, a divide-and-conquer operation is performed on the visualization operation preference learning annotation to obtain the scene cluster learning annotation, the behavior event cluster learning annotation and the database business item cluster learning annotation.

[0044] In some example embodiments, the visualization operation preference parsing network includes:

[0045] Image descriptor mining branch, scene feature mining branch, behavior event feature mining branch, database business project feature mining branch, first AI learning algorithm, second AI learning algorithm, disturbance feature addition branch, scene clustering component, behavior event clustering component, database business project clustering component and aggregation analysis branch;

[0046] The image descriptor mining branch is used to perform image descriptor mining processing on the database operation behavior image record to obtain the image wavelet descriptor corresponding to the database operation behavior image record; the image wavelet descriptor includes multiple quantized image description members;

[0047] The scene feature mining branch is used to perform scene feature change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain scene operation behavior description knowledge of the image wavelet descriptor; wherein the scene operation behavior description knowledge is a non-quantized image description member;

[0048] The behavior event feature mining branch is used to perform behavior event feature change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain event operation behavior description knowledge of the image wavelet descriptor; wherein the event operation behavior description knowledge is a non-quantized image description member;

[0049] The database business item feature mining branch is used to perform database business item feature change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain the business item operation behavior description knowledge of the image wavelet descriptor; wherein the business item operation behavior description knowledge is the non-quantized image description member;

[0050] The first AI learning algorithm is used to process the plurality of focus-plane operation behavior description knowledge respectively, and obtain the average description knowledge corresponding to the plurality of focus-plane operation behavior description knowledge respectively; wherein the average description knowledge is used to characterize the average positioning variable of the focus-plane operation behavior description knowledge in the knowledge relationship network;

[0051] The second AI learning algorithm is used to process the plurality of focus-plane operation behavior description knowledge respectively, and obtain the deviation description knowledge respectively corresponding to the plurality of focus-plane operation behavior description knowledge;

[0052] Among them, the deviation description knowledge is used to characterize the deviation positioning variable of the focus surface operation behavior description knowledge in the knowledge relationship network; the disturbance feature addition branch is used to combine the average description knowledge and the deviation description knowledge corresponding to the multiple focus surface operation behavior description knowledge respectively, add the target disturbance feature, and determine the quasi-cluster operation behavior description knowledge corresponding to the multiple focus surface operation behavior description knowledge respectively; the scene clustering component is used to perform knowledge clustering operation on the quasi-cluster operation behavior description knowledge of the scene focus surface, and obtain the scene focus surface clustering result of the operation behavior description knowledge; the behavior event clustering component is used to perform knowledge clustering operation on the quasi-cluster operation behavior description knowledge of the behavior event focus surface, and obtain the behavior event focus surface clustering result of the operation behavior description knowledge; the database business project clustering component is used to perform knowledge clustering operation on the quasi-cluster operation behavior description knowledge of the database business project focus surface, and obtain the database business project focus surface clustering result of the operation behavior description knowledge; the aggregation analysis branch is used to add the scene clustering result, the behavior event clustering result and the database business project clustering result to determine the visual operation preference analysis report of the database operation behavior image record.

[0053] A big data analysis system includes: a processor, a memory and a network interface; the processor is connected to the memory and the network interface; the network interface is used to provide data communication functions, the memory is used to store program codes, and the processor is used to call the program codes to execute the above method.

[0054] A computer-readable storage medium stores a computer program, which executes an interactive behavior big data mining method applied to a visual database when running.

[0055] A computer program product includes a computer program or a computer executable instruction. When the computer program or the computer executable instruction is executed by a processor, an interactive behavior big data mining method applied to a visual database is implemented.

[0056] According to one embodiment of the present invention, the image description change processing of the database operation behavior image record and the linkage knowledge clustering operation are performed to determine the linkage clustering result of the database operation behavior image record, and then the linkage clustering result is aggregated and analyzed to determine the visualization operation preference analysis report of the database operation behavior image record. The visualization operation preference analysis report is determined by analyzing and processing from multiple focus aspects, which can improve the accuracy of the visualization operation preference analysis report; and, when the image description change processing is performed, multiple quantitative image description members are changed into linked non-quantitative image description members, which can avoid as much as possible the error accumulation when performing operation preference analysis based on the image description later, thereby improving the accuracy and credibility of the visualization operation preference analysis report, so as to provide a decision-making basis for upgrading and optimizing the operation interaction function of the visualization database or customizing the personalized operation interface.

[0057] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present invention and are used together with the specification to illustrate the technical solutions of the present invention. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 A schematic diagram of a big data analysis system provided by an embodiment of the present invention is shown.

[0060] Figure 2 A flow chart of an interactive behavior big data mining method applied to a visual database provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention described and shown herein can be arranged and designed in various different configurations.

[0062] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0063] Figure 1 A schematic diagram of a big data analysis system provided by an embodiment of the present invention is shown, and the big data analysis system 100 includes a processor 110, a memory 120, and a network interface 130. The processor 110 is connected to the memory 120 and the network interface 130. Further, the network interface 130 is used to provide a data communication function, the memory 120 is used to store program code, and the processor 110 is used to call the program code to execute the interactive behavior big data mining method applied to the visual database.

[0064] Figure 2 is a flow chart showing a method for mining interactive behavior big data applied to a visual database that can implement an embodiment of the present invention. The method for mining interactive behavior big data applied to a visual database can be implemented by Figure 1 The big data analysis system 100 shown is implemented, and the method exemplarily includes steps STEP101-STEP105.

[0065] STEP 101, obtain the database operation behavior image record that triggers the visual interaction behavior analysis mechanism.

[0066] In an embodiment of the present invention, the database operation behavior image record can be understood as a series of visual operation behavior data generated by the interaction between the digital client and the database server. The visual operation behavior data includes but is not limited to click operation behavior data, sliding operation behavior data, dragging operation behavior data or box selection operation behavior data, etc.

[0067] Furthermore, the database operation behavior image record that triggers the visual interaction behavior analysis mechanism can be understood as the database operation behavior image record to be analyzed. The visual interaction behavior analysis mechanism can be created based on the number of visual operation behavior data in the database operation behavior image record, can be created based on different digital client identifiers, and can also be created based on different database interaction time periods.

[0068] In an embodiment of the present invention, after collecting the database operation behavior image records corresponding to the digital client, the big data analysis system obtains the database operation behavior image records that trigger the visual interaction behavior analysis mechanism, and then determines the visualization operation preference based on the visualization operation preference analysis report of the database operation behavior image records. Among them, the visualization operation preference can be understood as the operation habits or behavior preferences of the digital client when accessing the visualization database. For example, the visualization operation preference can be "biased towards click operation" or "biased towards box selection operation", etc., and can also be features such as common preferences for certain specific functions. In some examples, the visualization operation preference can be recorded in the form of a feature vector or a feature label.

[0069] For another example, the big data analysis system can obtain the database operation behavior image records that trigger the visual interaction behavior analysis mechanism in real time, or can obtain the database operation behavior image records that trigger the visual interaction behavior analysis mechanism according to the acquisition cycle. The big data analysis system can determine the acquisition rules of the database operation behavior image records that trigger the visual interaction behavior analysis mechanism according to actual needs.

[0070] STEP 102, perform image descriptor mining on the database operation behavior image record to obtain the image wavelet descriptor of the database operation behavior image record.

[0071] In the embodiment of the present invention, after obtaining the database operation behavior image record that triggers the visual interaction behavior analysis mechanism, the big data analysis system performs image descriptor mining processing (which can be understood as image feature mining processing) on ​​the database operation behavior image record to obtain the image wavelet descriptor of the database operation behavior image record. The image wavelet descriptor contains multiple quantized image description members.

[0072] For example, the wavelet descriptor has strong robustness to distortion, and can obtain higher image description accuracy with fewer coefficients, and supports multi-level analysis, through which multiple levels of description from coarse to fine can be achieved. Therefore, the feature characterization performance of the image wavelet descriptor can be improved by using the wavelet descriptor to realize the feature characterization of different operation behaviors. Furthermore, the quantitative image description member can be understood as a discrete description element or description variable.

[0073] Exemplarily, after obtaining the database operation behavior image record, the big data analysis system performs image descriptor mining on the database operation behavior image record, obtains the operation behavior description details carried by the database operation behavior image record, and then outputs an image wavelet descriptor composed of each operation behavior description detail. At this time, each operation behavior description detail is independent of each other and discontinuous in the image wavelet descriptor, that is, the operation behavior description details are quantitative image description members.

[0074] Under some exemplary design ideas, the big data analysis system obtains the above image wavelet descriptor through AI image processing. Exemplarily, after obtaining the above database operation behavior image record, the big data analysis system performs an image block extraction operation on the database operation behavior image record to obtain each image block in the database operation behavior image record, and then selects important image blocks as core image blocks based on the confidence of each image block, and then determines the operation behavior description details carried by the database operation behavior image record based on the core image block. Among them, the confidence of the image block is determined based on the detailed content of the image block and the heat of the image block.

[0075] In other exemplary design ideas, the big data analysis system obtains the above image wavelet descriptor through the image descriptor mining algorithm. Exemplarily, after obtaining the above database operation behavior image record, the big data analysis system loads the database operation behavior image record into the image descriptor mining algorithm, and then obtains the image wavelet descriptor generated by the image descriptor mining algorithm.

[0076] STEP 103, performing image description change processing on the image wavelet descriptor, changing multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtaining linkage operation behavior description knowledge of the image wavelet descriptor.

[0077] In an embodiment of the present invention, after obtaining the above image wavelet descriptor, the big data analysis system performs image description change processing on the image wavelet descriptor, changes multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtains the linkage operation behavior description knowledge of the image wavelet descriptor.

[0078] Among them, the above linkage operation behavior description knowledge includes multiple non-quantitative image description members of different focus aspects (different dimensions). Exemplarily, in an embodiment of the present invention, the linkage operation behavior description knowledge includes scene operation behavior description knowledge, event operation behavior description knowledge and business item operation behavior description knowledge. Based on this, the linkage operation behavior description knowledge can be understood as multi-dimensional operation behavior description knowledge.

[0079] For example, non-quantized image description members can be understood as continuous image description elements or image description variables, and image description change processing can be understood as multi-dimensional feature coding processing or classification coding processing.

[0080] Exemplarily, in an embodiment of the present invention, after obtaining the above image wavelet descriptor, the big data analysis system uses a scene feature mining branch (scene feature mining component) to perform image description change processing on the image wavelet descriptor from the scene focus surface, and then obtains the scene operation behavior description knowledge generated by the scene feature mining branch. Among them, the scene operation behavior description knowledge is used to characterize the interaction scene corresponding to the database operation behavior image record. For example, the interaction scene can be a data call interaction scene or a data migration interaction scene, etc., and can also be a data sharing interaction scene.

[0081] Exemplarily, in an embodiment of the present invention, after acquiring the above image wavelet descriptor, the big data analysis system uses a behavioral event feature mining branch to perform image description change processing on the image wavelet descriptor from the behavioral event focus surface, and then obtains the event operation behavior description knowledge generated by the behavioral event feature mining branch. Among them, the event operation behavior description knowledge is used to characterize the behavioral events matched by the communication terminal (i.e., the above digital client) of the database operation behavior image record. For example, the behavioral event can be a behavioral event contained in the above database operation behavior image record, such as a page click event and a file box selection event. Based on this, the subsequent visualization operation preference analysis report for the database operation behavior image record can include multiple visualization operation preference analysis reports.

[0082] Exemplarily, in an embodiment of the present invention, after obtaining the above image wavelet descriptor, the big data analysis system uses the database business project feature mining branch to perform image description change processing on the image wavelet descriptor from the database business project focus aspect, and then obtains the business project operation behavior description knowledge generated by the database business project feature mining branch. Among them, the business project operation behavior description knowledge is used to characterize the database business project that matches the above interaction scenario and the above behavior event, and the number of the database business project can be one or more. For example, the database business project can be a database business project included in the above database operation behavior image record, such as an encrypted access business project, an identity authentication call project, etc. Based on this, the subsequent visualization operation preference analysis report for the database operation behavior image record obtained can include multiple visualization operation preference analysis reports.

[0083] STEP 104, perform knowledge clustering operations on the linkage operation behavior description knowledge to obtain the linkage clustering result of the image wavelet descriptor.

[0084] In the embodiment of the present invention, after acquiring the above linkage operation behavior description knowledge, the big data analysis system performs knowledge clustering operations on the linkage operation behavior description knowledge to obtain linkage clustering results of the image wavelet descriptor. The linkage clustering results include clustering results of multiple different focus areas.

[0085] For example, corresponding to the aforementioned linkage operation behavior description knowledge, the linkage grouping results include scene grouping results, behavior event grouping results and database business project grouping results. Further, the knowledge grouping operation can be understood as feature classification processing, so that classification results of different dimensions can be obtained.

[0086] STEP 105: Perform aggregate analysis on the linkage group results to determine a visualized operation preference analysis report of the database operation behavior image record.

[0087] In the embodiment of the present invention, after obtaining the above linkage grouping results, the big data analysis system performs an aggregate analysis operation (aggregate identification processing) on ​​the linkage grouping results to determine a visualization operation preference analysis report of the database operation behavior image record. The visualization operation preference analysis report is used to characterize the visualization operation preference of the digital client for the above database operation behavior image record.

[0088] Exemplarily, when the above linkage grouping results include scenario grouping results, behavior event grouping results, and database business project grouping results, the big data analysis system adds analysis operations to the scenario grouping results, behavior event grouping results, and database business project grouping results, and then determines a visual operation preference analysis report of the database operation behavior image record based on the added analysis operations.

[0089] It can be seen that through the image description change processing of the database operation behavior image record and the linkage knowledge clustering operation, the linkage clustering result of the database operation behavior image record is determined, and then the linkage clustering result is aggregated and analyzed to determine the visualization operation preference analysis report of the database operation behavior image record. The visualization operation preference analysis report is determined by analyzing and processing from multiple focus aspects, which can improve the accuracy of the visualization operation preference analysis report; moreover, when the image description change is processed, multiple quantitative image description members are changed into linked non-quantitative image description members, which can avoid as much as possible the error accumulation when performing operation preference analysis based on the image description later. This can improve the accuracy and credibility of the visualization operation preference analysis report, so as to provide a decision-making basis for upgrading and optimizing the operation interaction function of the visualization database or customizing the personalized operation interface.

[0090] In some examples, the target focus area (target dimension) is taken as an example to explain the rules for obtaining the target focus area grouping results. The target focus area can be any focus area, such as the above-mentioned scenario focus area, behavior event focus area, database business project focus area, etc.

[0091] Under some possible design ideas, the above STEP104 includes the following contents: 1) For the target focus surface, the first AI learning algorithm is used to process the target focus surface operation behavior description knowledge, and obtain the average description knowledge of the target focus surface operation behavior description knowledge; 2) The second AI learning algorithm is used to process the target focus surface operation behavior description knowledge, and obtain the deviation description knowledge of the target focus surface operation behavior description knowledge; 3) The target perturbation feature is used to add perturbation features to the deviation description knowledge to obtain the processed deviation description knowledge; 4) Based on the average description knowledge and the processed deviation description knowledge, the quasi-cluster operation behavior description knowledge corresponding to the target focus surface operation behavior description knowledge is determined; 5) The quasi-cluster operation behavior description knowledge is subjected to knowledge clustering operation to obtain the target focus surface clustering result of the image wavelet descriptor.

[0092] In an embodiment of the present invention, average description knowledge is used to characterize the average positioning variable of the target focus surface operation behavior description knowledge in the knowledge relationship network. Deviation description knowledge is used to characterize the deviation positioning variable of the target focus surface operation behavior description knowledge in the knowledge relationship network. Taking the target focus surface as an example, in an embodiment of the present invention, after acquiring the target focus surface operation behavior description knowledge, the big data analysis system uses a first AI learning algorithm (average value determination algorithm) to process the target focus surface operation behavior description knowledge to obtain the average description knowledge of the target focus surface operation behavior description knowledge, and uses a second AI learning algorithm (variance determination algorithm) to process the target focus surface operation behavior description knowledge to obtain the deviation description knowledge of the target focus surface operation behavior description knowledge.

[0093] Furthermore, the big data analysis system uses the target disturbance feature to perform disturbance feature addition processing on the deviation description knowledge to obtain the processed deviation description knowledge. Further, based on the above average description knowledge and the processed deviation description knowledge, the quasi-clustering operation behavior description knowledge corresponding to the target dimension feature is determined, and the quasi-clustering operation behavior description knowledge is subjected to knowledge clustering operation to obtain the target attention surface clustering result of the image wavelet descriptor. Among them, the above target disturbance feature can be any disturbance feature that satisfies the Gaussian distribution.

[0094] Exemplarily, the big data analysis system performs knowledge clustering operations on the knowledge of the description of the operation behavior of the proposed clustering. For the knowledge of the description of the operation behavior of the proposed clustering of different focus aspects, the clustering components (classifiers) of different focus aspects are used to perform knowledge clustering operations. For example, for the scene focus aspect, the big data analysis system uses the scene clustering component to perform knowledge clustering operations on the knowledge of the description of the operation behavior of the proposed clustering of the scene focus aspect, and obtains the scene clustering results of the operation behavior description knowledge; for the behavior event focus aspect, the big data analysis system uses the behavior event clustering component to perform knowledge clustering operations on the knowledge of the description of the operation behavior of the proposed clustering of the behavior event focus aspect, and obtains the behavior event clustering results of the operation behavior description knowledge; for the database business project focus aspect, the big data analysis system uses the database business project clustering component to perform knowledge clustering operations on the knowledge of the description of the operation behavior of the proposed clustering of the database business project focus aspect, and obtains the database business project clustering results of the operation behavior description knowledge.

[0095] Another embodiment of the present invention provides an interactive behavior big data mining method applied to a visual database, including the following STEP 201-STEP 208.

[0096] STEP 201, obtain the database operation behavior image record that triggers the visual interaction behavior analysis mechanism.

[0097] STEP 202, perform image descriptor mining on the database operation behavior image record to obtain the image wavelet descriptor of the database operation behavior image record.

[0098] STEP 203, performing image description change processing on the image wavelet descriptor, changing multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtaining linkage operation behavior description knowledge of the image wavelet descriptor.

[0099] STEP 204, perform knowledge clustering operations on the linkage operation behavior description knowledge to obtain the linkage clustering result of the image wavelet descriptor.

[0100] STEP 205, perform aggregation and analysis operations on the linkage group results to determine a visualized operation preference analysis report of the database operation behavior image record.

[0101] The above STEP 201 to STEP 205 are the same as STEP 101 to STEP 105.

[0102] STEP 206 , according to the visualization operation preference analysis report, obtain a database interface optimization case that matches the visualization operation preference analysis report from the database operation behavior image record set as a candidate database interface optimization case.

[0103] In an embodiment of the present invention, after obtaining the above visualization operation preference analysis report, the big data analysis system obtains the database interface optimization case that matches the visualization operation preference analysis report from the database operation behavior image record set as an alternative database interface optimization case according to the visualization operation preference analysis report. Among them, the database interface optimization case is a database interface optimization reference for the above database operation behavior image record. Exemplarily, after obtaining the database interface optimization case, the big data analysis system performs targeted optimization and upgrade of the interactive interface of the visualization database according to the database interface optimization case and the visualization operation preference analysis report.

[0104] Exemplarily, the above database operation behavior image record set includes database interface optimization cases corresponding to each visualization operation preference analysis report. In the embodiment of the present invention, since the visualization operation preference analysis report is identified and determined based on the scene grouping results, behavior event grouping results, and database business project grouping results, correspondingly, the database operation behavior image record set records the database interface optimization cases of the visualization operation preference analysis report with the same focus.

[0105] For example, the database operation behavior image record centrally records the database interface optimization case corresponding to the target interaction scenario, and the database interface optimization case corresponding to the target interaction scenario includes the database interface optimization case directly for the target interaction scenario (applied when it is difficult to include the behavior events and database business items that are accurately represented in the visualization operation preference analysis report), and the database interface optimization case corresponding to the target interaction scenario includes the database interface optimization case corresponding to each behavior event (applied when the interaction scenario and behavior event are accurately represented in the visualization operation preference analysis report but it is difficult to accurately represent the database business items), and in addition, for the database interface optimization case corresponding to the target behavior event, the database interface optimization case corresponding to each database business item is included (applied when the interaction scenario, behavior event and database business item are accurately represented in the visualization operation preference analysis report); or, the database operation behavior image record centrally records the database interface optimization case corresponding to the target behavior event, and the database interface optimization case corresponding to the target behavior event includes the database interface optimization case directly for the target behavior event (applied when it is difficult to include the interaction scenario and database business item that are accurately represented in the visualization operation preference analysis report), and the database interface optimization case corresponding to the target behavior event includes each interaction scenario. The database interface optimization cases corresponding to the target interaction scenarios are recorded in a centralized manner, and the database interface optimization cases corresponding to the target database business projects are recorded in a centralized manner. The database interface optimization cases corresponding to the target database business projects include database interface optimization cases directly for the target database business projects (applied when it is difficult to include interaction scenarios and behavior events with accurate representation in the visualization operation preference analysis report), and the database interface optimization cases corresponding to the target database business projects include database interface optimization cases corresponding to the target database business projects (applied when it is difficult to include interaction scenarios and behavior events with accurate representation in the visualization operation preference analysis report), and the database interface optimization cases corresponding to the target database business projects include database interface optimization cases corresponding to the various interaction scenarios (applied when it is difficult to accurately represent interaction scenarios and database business projects in the visualization operation preference analysis report), and the database interface optimization cases corresponding to the target database business projects include database interface optimization cases corresponding to the various interaction scenarios (applied when it is difficult to accurately represent interaction scenarios and database business projects in the visualization operation preference analysis report), and the database interface optimization cases corresponding to the target interaction scenarios include database interface optimization cases corresponding to the various behavior events (applied when it is difficult to accurately represent interaction scenarios, behavior events and database business projects in the visualization operation preference analysis report).Exemplarily, the big data analysis system can record each database interface optimization case in the form of "interaction scenario--behavioral event--database business project" to realize the retrieval of database interface optimization cases. If there is no corresponding interaction scenario for the database interface optimization case, the interaction scenario is represented by "none". Behavioral events and database business projects are similar to interaction scenarios.

[0106] STEP 207, when there are multiple candidate database interface optimization cases, obtain optimization feedback metric values ​​of each candidate database interface optimization case.

[0107] In an embodiment of the present invention, after obtaining the above candidate database interface optimization cases, the big data analysis system determines the processing method for the candidate database interface optimization cases based on the number of the candidate database interface optimization cases. If the number of candidate database interface optimization cases is unique, the big data analysis system directly uses the candidate database interface optimization case as the database interface optimization case corresponding to the above database operation behavior image record; if the number of candidate database interface optimization cases is multiple, the big data analysis system obtains the optimization feedback metric value (quality evaluation value) of each candidate database interface optimization case, and selects the database interface optimization case corresponding to the database operation behavior image record from the candidate database interface optimization cases based on the optimization feedback metric value.

[0108] Among them, the optimization feedback metric is used to characterize the quality of the database interface optimization case. Exemplarily, the optimization feedback metric includes at least one of the following: the number of citations of the database interface optimization case, the feedback opinions contained in the database interface optimization case, the number of references to the database interface optimization case, etc. Among them, the number of citations of the database interface optimization case can be understood as the ratio between the number of views of the database interface optimization case and the number of references to the database interface optimization case. The feedback opinions are used to reflect the quality of the reference experience of the database interface optimization case, etc. The number of references can be understood as the actual number of uses of some optimization solutions in the database interface optimization case rather than the number of views.

[0109] STEP 208, from the candidate database interface optimization cases, select the candidate database interface optimization cases whose optimization feedback metric values ​​meet the requirements as the database interface optimization cases corresponding to the database operation behavior image records.

[0110] In the embodiment of the present invention, after obtaining the above optimization feedback metric value, the big data analysis system selects the candidate database interface optimization case whose optimization feedback metric value meets the requirements from the candidate database interface optimization cases as the database interface optimization case corresponding to the database operation behavior image record. Exemplarily, the above requirement can be a single requirement in the optimization feedback metric value or a joint requirement in the optimization feedback metric value.

[0111] It can be seen that obtaining the database interface optimization cases corresponding to the database operation behavior image records through the visual operation preference analysis report can provide a reference for database interface optimization, and select the database interface optimization cases that meet the requirements according to the optimization feedback metric values ​​of each candidate database interface optimization case to ensure the applicability of the database interface optimization case.

[0112] A method for debugging a visual operation preference parsing network provided by an embodiment of the present invention includes STEP 301 to STEP 307.

[0113] STEP 301, obtain at least one debugging learning information of the visualization operation preference parsing network.

[0114] The visual operation preference parsing network can be understood as a residual network that determines the visual operation preference parsing report of the database operation behavior image record through the image wavelet descriptor of the database operation behavior image record. In an embodiment of the present invention, before debugging the visual operation preference parsing network, the big data parsing system obtains at least one debugging learning information (training sample) of the visual operation preference parsing network. The debugging learning information includes past database operation behavior image records and learning annotations (training labels) corresponding to the past database operation behavior image records.

[0115] STEP 302, perform image descriptor mining on the image records of past database operation behaviors to obtain image wavelet descriptors of the image records of past database operation behaviors.

[0116] In the embodiment of the present invention, after obtaining the above-mentioned past database operation behavior image record, the big data analysis system performs image descriptor mining processing on the past database operation behavior image record to obtain the image wavelet descriptor of the past database operation behavior image record. The image wavelet descriptor includes multiple quantitative image description members.

[0117] Exemplarily, the big data parsing system performs image descriptor mining processing on past database operation behavior image records through the image descriptor mining branch in the visualization operation preference parsing network, and then obtains the image wavelet descriptor output by the image descriptor mining algorithm.

[0118] STEP 303, perform image description change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain linkage operation behavior description knowledge of the image wavelet descriptor.

[0119] In the embodiment of the present invention, after obtaining the above image wavelet descriptor, the big data analysis system performs image description change processing on the image wavelet descriptor, changes multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtains linkage operation behavior description knowledge of the image wavelet descriptor. The linkage operation behavior description knowledge includes multiple non-quantized image description members of different focus areas.

[0120] Exemplarily, the big data analysis system uses feature mining components of different focus planes in the visualization operation preference analysis network to perform image description change processing on the image wavelet descriptor from different focus planes, and then obtains multiple non-quantitative image description members of different focus planes generated by the feature mining components of different focus planes.

[0121] STEP 304, perform knowledge clustering operations on the linkage operation behavior description knowledge to obtain the linkage clustering results of the image wavelet descriptor.

[0122] In the embodiment of the present invention, after acquiring the above linkage operation behavior description knowledge, the big data analysis system performs knowledge clustering operations on the linkage operation behavior description knowledge to obtain linkage clustering results of the image wavelet descriptor. The linkage clustering results include clustering results of multiple different focus areas.

[0123] Exemplarily, when the big data analysis system obtains the linkage clustering results, the first AI learning algorithm in the visualization operation preference analysis network is used to process the multiple focus-side operation behavior description knowledge separately, and the average description knowledge corresponding to the multiple focus-side operation behavior description knowledge is obtained. The second AI learning algorithm in the visualization operation preference analysis network is used to process the multiple focus-side operation behavior description knowledge separately, and the deviation description knowledge corresponding to the multiple focus-side operation behavior description knowledge is obtained. Afterwards, the disturbance feature in the visualization operation preference analysis network is used to add branches, and based on the average description knowledge and deviation description knowledge corresponding to the multiple focus-side operation behavior description knowledge, the target disturbance feature is added to determine the quasi-clustering operation behavior description knowledge corresponding to the multiple focus-side operation behavior description knowledge, and then the clustering components of different focus-sides in the visualization operation preference analysis network are used to perform knowledge clustering operations on the quasi-clustering operation behavior description knowledge of the corresponding focus-side, respectively, to obtain the above linkage clustering results. Among them, the above average description knowledge is used to characterize the average positioning variable of a certain focus-side operation behavior description knowledge in the knowledge relationship network, and the above deviation description knowledge is used to characterize the deviation positioning variable of a certain focus-side operation behavior description knowledge in the knowledge relationship network.

[0124] STEP 305: Perform aggregate analysis on the linkage group results to determine a visualized operation preference analysis report of past database operation behavior image records.

[0125] In the embodiment of the present invention, after obtaining the above linkage grouping results, the big data analysis system performs aggregation analysis operations on the linkage grouping results to determine a visualized operation preference analysis report of the past database operation behavior image records.

[0126] Exemplarily, the big data analysis system adds recognition processing to the linkage group results through the aggregate analysis branch in the visual operation preference analysis network, and determines a visual operation preference analysis report of the past database operation behavior image records.

[0127] STEP 306, based on the linkage grouping results, the visualization operation preference analysis report, and the learning annotations corresponding to the past database operation behavior image records, determine the network learning cost variables of the visualization operation preference analysis network.

[0128] In the embodiment of the present invention, the big data analysis system determines the network learning cost variable of the visualization operation preference analysis network based on the above linkage grouping results, the above visualization operation preference analysis report, and the learning annotations corresponding to the above past database operation behavior image records. Among them, the network learning cost (loss function) is used to characterize the performance of the visualization operation preference analysis network.

[0129] STEP307, improve the network configuration parameters of the visualization operation preference parsing network based on the network learning cost variable.

[0130] In an embodiment of the present invention, after obtaining the above network learning cost variables, the big data analysis system improves the network configuration parameters of the visualization operation preference analysis network based on the network learning cost variables, and uses the above debugging learning information to continue to debug the visualization operation preference analysis network after the network configuration parameters are improved until the network learning cost variables converge.

[0131] It can be seen that by debugging the visualization operation preference parsing network through debugging learning information, during the debugging process, when the image description is changed, multiple quantized image description members are changed into linked non-quantized image description members, which can significantly reduce the loss of details of the visualization operation preference parsing network during application and improve the accuracy of the visualization operation preference parsing report.

[0132] Exemplarily, in an embodiment of the present invention, the above linkage clustering results include scene clustering results, behavior event clustering results, and database business project clustering results, and the above learning annotations include scene clustering learning annotations, behavior event clustering learning annotations, and database business project clustering learning annotations corresponding to past database operation behavior image records, as well as visualization operation preference learning annotations corresponding to past database operation behavior image records. Among them, the learning annotations are the real results corresponding to the visualization operation preference analysis network. Exemplarily, the big data analysis system performs a divide-and-conquer operation on the visualization operation preference learning annotations based on the set processing strategy to obtain scene clustering learning annotations, behavior event clustering learning annotations, and database business project clustering learning annotations.

[0133] In some possible design ideas, the above STEP306 includes: 1) Based on the scene clustering results, the behavior event clustering results, and the database business project clustering results, combined with the scene clustering learning annotations, the behavior event clustering learning annotations, and the database business project clustering learning annotations, determine the first network learning cost variable corresponding to the visualization operation preference parsing network; 2) Based on the visualization operation preference parsing report and the visualization operation preference learning annotations, determine the second network learning cost variable corresponding to the visualization operation preference parsing network; 3) Based on the first network learning cost variable and the second network learning cost variable, determine the network learning cost variable of the visualization operation preference parsing network. The first network learning cost indicator is used to characterize the accuracy of the linkage clustering results of the visualization operation preference parsing network. The second network learning cost indicator is used to characterize the accuracy of the visualization operation preference parsing report of the visualization operation preference parsing network.

[0134] In the embodiment of the present invention, to ensure the performance of each cluster component in the visualization operation preference analysis network, the big data analysis system determines the first network learning cost variable corresponding to the visualization operation preference analysis network based on the scene cluster results, the behavior event cluster results, and the database business project cluster results, combined with the scene cluster learning annotations, the behavior event cluster learning annotations, and the database business project cluster learning annotations; to ensure the accuracy of the visualization operation preference analysis report in the visualization operation preference analysis network, the big data analysis system determines the second network learning cost variable corresponding to the visualization operation preference analysis network based on the visualization operation preference analysis report and the visualization operation preference learning annotations. Furthermore, the big data analysis system determines the network learning cost variable of the visualization operation preference analysis network based on the first network learning cost variable and the second network learning cost variable.

[0135] Exemplarily, when obtaining the first network learning cost variable, based on the scene clustering results, combined with the scene clustering learning annotations, the behavior event clustering learning annotations and the database business project clustering learning annotations, determine the network learning cost variable corresponding to the scene clustering component in the visualization operation preference analysis network; based on the behavior event clustering results, combined with the scene clustering learning annotations, the behavior event clustering learning annotations and the database business project clustering learning annotations, determine the network learning cost variable corresponding to the behavior event clustering component in the visualization operation preference analysis network; based on the database business project clustering results, combined with the scene clustering learning annotations, the behavior event clustering learning annotations and the database business project clustering learning annotations, determine the network learning cost variable corresponding to the database business project clustering component in the visualization operation preference analysis network; and then determine the first network learning cost variable according to the network learning cost variable corresponding to the scene clustering component, the network learning cost variable corresponding to the behavior event clustering component and the network learning cost variable corresponding to the database business project clustering component. Among them, the big data analysis system can sum the network learning cost variable corresponding to the scene clustering component, the network learning cost variable corresponding to the behavior event clustering component and the network learning cost variable corresponding to the database business project clustering component to obtain the first network learning cost variable.

[0136] Exemplarily, the visualization operation preference parsing network includes: an image descriptor mining branch, a scene feature mining branch, a behavior event feature mining branch, a database business project feature mining branch, a first AI learning algorithm, a second AI learning algorithm, a disturbance feature addition branch, a scene clustering component, a behavior event clustering component, a database business project clustering component and an aggregation parsing branch.Among them, the image descriptor mining branch is used to perform image descriptor mining on the database operation behavior image record to obtain the image wavelet descriptor corresponding to the database operation behavior image record, and the image wavelet descriptor contains multiple quantized image description members; the scene feature mining branch is used to perform scene feature change processing on the image wavelet descriptor, and change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members to obtain the scene operation behavior description knowledge of the image wavelet descriptor, and the scene operation behavior description knowledge is a non-quantized image description member; the behavior event feature mining branch is used to perform behavior event feature change processing on the image wavelet descriptor, and change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members. The image description member is changed into a non-quantitative image description member to obtain the event operation behavior description knowledge of the image wavelet descriptor, and the event operation behavior description knowledge is a non-quantitative image description member; the database business project feature mining branch is used to perform database business project feature change processing on the image wavelet descriptor, and multiple quantitative image description members in the image wavelet descriptor are changed into non-quantitative image description members to obtain the business project operation behavior description knowledge of the image wavelet descriptor, and the business project operation behavior description knowledge is a non-quantitative image description member; the first AI learning algorithm is used to process multiple focus surface operation behavior description knowledge separately, and obtain the average description knowledge corresponding to the multiple focus surface operation behavior description knowledge respectively. The average description knowledge is used to characterize the average positioning variable of the attention surface operation behavior description knowledge in the knowledge relationship network; the second AI learning algorithm is used to process multiple attention surface operation behavior description knowledge separately, and obtain the deviation description knowledge corresponding to the multiple attention surface operation behavior description knowledge respectively, and the deviation description knowledge is used to characterize the deviation positioning variable of the attention surface operation behavior description knowledge in the knowledge relationship network; the disturbance feature addition branch is used to add target disturbance features based on the average description knowledge and deviation description knowledge corresponding to the multiple attention surface operation behavior description knowledge respectively, and determine the quasi-cluster operation behavior description knowledge corresponding to the multiple attention surface operation behavior description knowledge respectively; the scene clustering component is used to perform quasi-cluster operation behavior description on the scene attention surface A knowledge clustering operation is performed to describe the knowledge, and the scenario focus-side clustering result of the operation behavior description knowledge is obtained; the behavior event clustering component is used to perform knowledge clustering operations on the quasi-clustered operation behavior description knowledge of the behavior event focus-side, and obtain the behavior event focus-side clustering result of the operation behavior description knowledge; the database business project clustering component is used to perform knowledge clustering operations on the quasi-clustered operation behavior description knowledge of the database business project focus-side, and obtain the database business project focus-side clustering result of the operation behavior description knowledge; the aggregation analysis branch is used to add recognition to the scenario clustering results, the behavior event clustering results, and the database business project clustering results, and determine the visual operation preference analysis report of the database operation behavior image record.

[0137] For example, the big data analysis system loads the database operation behavior image record into the visualization operation preference analysis network, and the image descriptor mining branch in the visualization operation preference analysis network performs image descriptor mining on the database operation behavior image record to obtain the image wavelet descriptor of the database operation behavior image record. Furthermore, the scene feature mining branch performs scene feature change processing on the image wavelet descriptor to obtain scene operation behavior description knowledge; the behavior event feature mining branch performs behavior event feature change processing on the image wavelet descriptor to obtain event operation behavior description knowledge; the database business project feature mining branch performs database business project feature change processing on the image wavelet descriptor to obtain business project operation behavior description knowledge. Afterwards, the average description knowledge and deviation description knowledge corresponding to each focus surface operation behavior description knowledge are obtained respectively through the first AI learning algorithm and the second AI learning algorithm, and the quasi-cluster operation behavior description knowledge of the scene focus surface, the quasi-cluster operation behavior description knowledge of the behavior event focus surface, and the quasi-cluster operation behavior description knowledge of the database business project focus surface are obtained by adding branches with perturbation features. Afterwards, the scenario clustering component is used to perform knowledge clustering operations on the description of the proposed clustering operation behavior to determine the scenario clustering results; the behavior event clustering component is used to perform knowledge clustering operations on the description of the proposed clustering operation behavior to determine the behavior event clustering results; the database business project clustering component is used to perform knowledge clustering operations on the description of the proposed clustering operation behavior to determine the database business project clustering results. Afterwards, the scenario clustering results, the behavior event clustering results, and the database business project clustering results are added and fused, and the feature decoding component is used to perform feature decoding on the added fusion results to obtain a visual operation preference analysis report of the database operation behavior image record.

[0138] Taking the data access scenario as an example, the technical solution of the embodiment of the present invention is described:

[0139] STEP 401, obtain the database operation behavior image record that triggers the visual interaction behavior analysis mechanism.

[0140] STEP 402, loading the database operation behavior image record into the visualization operation preference parsing network, and obtaining the visualization operation preference parsing report output by the visualization operation preference parsing network.

[0141] Exemplarily, the visualization operation preference parsing network includes an image descriptor mining branch, a scene feature mining branch, a behavior event feature mining branch, a database business project feature mining branch, a first AI learning algorithm, a second AI learning algorithm, a disturbance feature addition branch, a scene clustering component, a behavior event clustering component, a database business project clustering component and an aggregation parsing branch.

[0142] The image descriptor mining branch performs image descriptor mining processing on the database operation behavior image record to obtain the image wavelet descriptor corresponding to the database operation behavior image record, and the image wavelet descriptor contains multiple quantized image description members.

[0143] The scene feature mining branch performs scene feature change processing on the image wavelet descriptor, changes multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtains the scene operation behavior description knowledge of the image wavelet descriptor, which is a non-quantized image description member; the behavior event feature mining branch performs behavior event feature change processing on the image wavelet descriptor, changes multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtains the event operation behavior description knowledge of the image wavelet descriptor, which is a non-quantized image description member; the database business project feature mining branch is used to perform database business project feature change processing on the image wavelet descriptor, changes multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtains the business project operation behavior description knowledge of the image wavelet descriptor, which is a non-quantized image description member.

[0144] The first AI learning algorithm processes multiple focus-plane operation behavior description knowledge separately to obtain the average description knowledge corresponding to the multiple focus-plane operation behavior description knowledge, and the average description knowledge is used to characterize the average positioning variable of the focus-plane operation behavior description knowledge in the knowledge relationship network; the second AI learning algorithm processes multiple focus-plane operation behavior description knowledge separately to obtain the deviation description knowledge corresponding to the multiple focus-plane operation behavior description knowledge, and the deviation description knowledge is used to characterize the deviation positioning variable of the focus-plane operation behavior description knowledge in the knowledge relationship network.

[0145] The disturbance feature addition branch is based on the average description knowledge and deviation description knowledge corresponding to multiple focus-plane operation behavior description knowledge, adds the target disturbance feature, and determines the quasi-cluster operation behavior description knowledge corresponding to multiple focus-plane operation behavior description knowledge.

[0146] The scenario clustering component performs knowledge clustering operations on the quasi-clustering operation behavior description knowledge of the scenario focus surface, and obtains the scenario focus surface clustering results of the operation behavior description knowledge; the behavior event clustering component performs knowledge clustering operations on the quasi-clustering operation behavior description knowledge of the behavior event focus surface, and obtains the behavior event focus surface clustering results of the operation behavior description knowledge; the database business project clustering component is used to perform knowledge clustering operations on the quasi-clustering operation behavior description knowledge of the database business project focus surface, and obtains the database business project focus surface clustering results of the operation behavior description knowledge.

[0147] The aggregation analysis branch adds recognition to the scene clustering results, behavior event clustering results, and database business project clustering results to determine a visual operation preference analysis report of the database operation behavior image record.

[0148] STEP 403, according to the visualization operation preference analysis report, determine whether the visualization operation preference of the digital client is a click-to-access preference. If the visualization operation preference of the digital client is a click-to-access preference, execute STEP 404; if the visualization operation preference of the digital client is not a click-to-access preference, execute STEP 405-608.

[0149] STEP 404 triggers the database interface optimization task based on the click access preference.

[0150] STEP 405, obtaining a database interface optimization case that matches the visualization operation preference analysis report from the database operation behavior image record set as an alternative database interface optimization case.

[0151] STEP 406, when there are multiple candidate database interface optimization cases, obtain optimization feedback metric values ​​of each candidate database interface optimization case.

[0152] STEP 407, from the candidate database interface optimization cases, select the candidate database interface optimization cases whose optimization feedback metric values ​​meet the requirements as the database interface optimization cases corresponding to the database operation behavior image records.

[0153] STEP 408 , performing database interface optimization processing based on click-to-access preference based on the database interface optimization case.

[0154] It can be seen that by debugging the visualization operation preference parsing network through debugging learning information, during the debugging process, when the image description is changed, multiple quantized image description members are changed into linked non-quantized image description members, which can significantly reduce the loss of details of the visualization operation preference parsing network during application and improve the accuracy of the visualization operation preference parsing report.

[0155] Under some other independently implementable design ideas, after selecting the alternative database interface optimization cases whose optimization feedback metrics meet the requirements from the alternative database interface optimization cases as the database interface optimization cases corresponding to the database operation behavior image records, the method also includes: determining the current optimization strategy in combination with the visual operation preference analysis report and the database interface optimization cases corresponding to the database operation behavior image records, and performing interface optimization processing based on the current optimization strategy.

[0156] Under some other independently implementable design ideas, the current optimization strategy is determined by combining the visualization operation preference analysis report and the database interface optimization case corresponding to the database operation behavior image record, and interface optimization processing is performed based on the current optimization strategy, including the steps of: extracting a local optimization reference solution image from the database interface optimization case corresponding to the database operation behavior image record based on the visualization operation preference analysis report; obtaining multiple GUI optimization element data of the local optimization reference solution image; loading the local optimization reference solution image into a first convolutional neural network, and performing optimization detail vector extraction on the local optimization reference solution image to obtain multiple first interface optimization update vectors; loading multiple GUI optimization element data into The second convolutional neural network organizes the various GUI optimization element data loaded into the second convolutional neural network to obtain a plurality of GUI optimization aggregate element data; the second convolutional neural network weights the plurality of GUI optimization element data and the plurality of GUI optimization aggregate element data to obtain GUI optimization weighted element data; the GUI optimization weighted element data are respectively merged with the respective first interface optimization update vectors to obtain a plurality of second interface optimization update vectors; the plurality of second interface optimization update vectors are loaded into the optimization strategy discrimination network for optimization strategy discrimination to obtain an optimization strategy discrimination label; the current optimization strategy is determined by the optimization strategy discrimination label and the interface optimization processing is performed based on the current optimization strategy.

[0157] Under some other independently implementable design ideas, the various GUI optimization element data loaded into the second convolutional neural network are sorted to obtain multiple GUI optimization aggregate element data, including: cross-sorting the various GUI optimization element data loaded into the second convolutional neural network to obtain multiple GUI optimization aggregate element data.

[0158] Under some other independently implementable design ideas, the second convolutional neural network weights the multiple GUI optimization element data and the multiple GUI optimization aggregate element data to obtain GUI optimization weighted element data, including: the second convolutional neural network weights the multiple GUI optimization element data and the multiple GUI optimization aggregate element data to obtain GUI optimization weighted element data.

[0159] It can be seen that through the above technical solution, the GUI optimization elements in the local optimization reference solution image can be comprehensively analyzed, and the current optimization strategy can be accurately determined in combination with the interface optimization update vector, which can improve the accuracy and efficiency of the interface optimization processing.

[0160] Based on the same or similar technical concepts mentioned above, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and the computer program executes an interactive behavior big data mining method applied to a visual database when running.

[0161] Based on the same or similar technical concepts mentioned above, an embodiment of the present invention also provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, a method for mining big data of interactive behaviors applied to a visual database is implemented.

[0162] The embodiment of the present invention provides an interactive behavior big data mining method and system applied to a visual database. The method determines the linkage grouping result of the database operation behavior image record through image description change processing and linkage knowledge grouping operation of the database operation behavior image record, and then aggregates and analyzes the linkage grouping result to determine the visual operation preference analysis report of the database operation behavior image record. The visual operation preference analysis report is determined by analyzing and processing from multiple focus aspects, which can improve the accuracy of the visual operation preference analysis report. Moreover, when the image description is changed, multiple quantitative image description members are changed into linked non-quantitative image description members, which can avoid error accumulation when performing operation preference analysis based on the image description later as much as possible. In this way, the accuracy and credibility of the visual operation preference analysis report can be improved, so as to provide a decision-making basis for upgrading and optimizing the operation interaction function of the visual database or customizing the personalized operation interface.

[0163] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0164] The above disclosure is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the relevant contents of the present invention are still within the scope of the present invention.

Claims

1. A method for mining interactive behavior big data applied to a visual database, characterized in that: Applied to a big data analysis system, the method includes: Obtaining database operation behavior image records that trigger the visual interaction behavior parsing mechanism; Performing image descriptor mining processing on the database operation behavior image record to obtain an image wavelet descriptor of the database operation behavior image record; wherein the image wavelet descriptor includes a plurality of quantized image description members; Performing image description change processing on the image wavelet descriptor, changing multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtaining linkage operation behavior description knowledge of the image wavelet descriptor; wherein the linkage operation behavior description knowledge includes multiple non-quantized image description members of different focus areas, and the non-quantized image description members are continuous image description elements or image description variables; Performing knowledge clustering operations on the linkage operation behavior description knowledge respectively to obtain linkage clustering results of the image wavelet descriptor, wherein the linkage clustering results include clustering results of the multiple different focus surfaces; Performing aggregation and analysis operations on the linkage group results to determine a visualized operation preference analysis report of the database operation behavior image record; The linkage operation behavior description knowledge includes scenario operation behavior description knowledge, event operation behavior description knowledge and business item operation behavior description knowledge; wherein the scenario operation behavior description knowledge is used to characterize the interaction scenario corresponding to the database operation behavior image record, the event operation behavior description knowledge is used to characterize the behavior event matched by the communication terminal of the database operation behavior image record, and the business item operation behavior description knowledge is used to characterize the database business item matched with the scenario and the behavior event; The performing knowledge grouping operations on the linkage operation behavior description knowledge to obtain the linkage grouping results of the image wavelet descriptor includes: For the target focus surface, a first AI learning algorithm is used to process the target focus surface operation behavior description knowledge to obtain average description knowledge of the target focus surface operation behavior description knowledge; wherein the average description knowledge is used to characterize the average positioning variable of the target focus surface operation behavior description knowledge in the knowledge relationship network; The target focus surface operation behavior description knowledge is processed by a second AI learning algorithm to obtain deviation description knowledge of the target focus surface operation behavior description knowledge; wherein the deviation description knowledge is used to characterize the deviation positioning variable of the target focus surface operation behavior description knowledge in the knowledge relationship network; Using the target disturbance feature to perform disturbance feature addition processing on the deviation description knowledge to obtain processed deviation description knowledge; Determine the quasi-cluster operation behavior description knowledge corresponding to the target focus surface operation behavior description knowledge by combining the average description knowledge and the processed deviation description knowledge; A knowledge clustering operation is performed on the quasi-clustering operation behavior description knowledge to obtain a target attention surface clustering result of the image wavelet descriptor.

2. The method according to claim 1, characterized in that After performing an aggregation analysis operation on the linkage group results to determine a visualized operation preference analysis report of the database operation behavior image record, the method further includes: According to the visualization operation preference analysis report, a database interface optimization case matching the visualization operation preference analysis report is obtained from a database operation behavior image record set as a candidate database interface optimization case; When there are multiple candidate database interface optimization cases, obtaining optimization feedback metric values ​​of each candidate database interface optimization case; From the candidate database interface optimization cases, the candidate database interface optimization case whose optimization feedback metric value meets the requirements is selected as the database interface optimization case corresponding to the database operation behavior image record.

3. The method according to claim 1, characterized in that The method is implemented by a visual operation preference parsing network, and the debugging steps of the visual operation preference parsing network include: Acquire at least one debugging learning information of the visualization operation preference parsing network, wherein the debugging learning information includes a past database operation behavior image record and a learning annotation corresponding to the past database operation behavior image record; Performing image descriptor mining processing on the image records of the past database operation behaviors to obtain image wavelet descriptors of the image records of the past database operation behaviors; wherein the image wavelet descriptors include a plurality of quantized image description members; Performing image description change processing on the image wavelet descriptor, changing multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtaining linkage operation behavior description knowledge of the image wavelet descriptor; wherein the linkage operation behavior description knowledge includes multiple non-quantized image description members of different focus areas; Performing knowledge clustering operations on the linkage operation behavior description knowledge respectively to obtain linkage clustering results of the image wavelet descriptor, wherein the linkage clustering results include clustering results of the multiple different focus surfaces; Performing aggregation and analysis operations on the linkage group results to determine a visualized operation preference analysis report of the image records of the past database operation behaviors; Determine the network learning cost variable of the visualization operation preference analysis network by combining the linkage grouping result, the visualization operation preference analysis report, and the learning annotations corresponding to the past database operation behavior image records; The network configuration parameters of the visualization operation preference parsing network are improved in combination with the network learning cost variable.

4. The method according to claim 3, characterized in that The linkage grouping results include scenario grouping results, behavior event grouping results and database business project grouping results; The learning annotations include scene group learning annotations, behavior event group learning annotations, and database business item group learning annotations corresponding to the past database operation behavior image records, as well as visualization operation preference learning annotations corresponding to the past database operation behavior image records; The step of combining the linkage grouping result, the visualization operation preference analysis report, and the learning annotations corresponding to the past database operation behavior image records to determine the network learning cost variables of the visualization operation preference analysis network includes: Determine the first network learning cost variable corresponding to the visualization operation preference parsing network by combining the scene clustering result, the behavior event clustering result, the database business project clustering result, the scene clustering learning annotation, the behavior event clustering learning annotation and the database business project clustering learning annotation; wherein the first network learning cost indicator is used to characterize the accuracy of the linkage clustering result of the visualization operation preference parsing network; Determine a second network learning cost variable corresponding to the visualization operation preference parsing network in combination with the visualization operation preference parsing report and the visualization operation preference learning annotation; wherein the second network learning cost indicator is used to characterize the accuracy of the visualization operation preference parsing report of the visualization operation preference parsing network; Determine the network learning cost variable of the visualization operation preference parsing network by combining the first network learning cost variable and the second network learning cost variable; The combining of the scene clustering results, the behavior event clustering results, the database business project clustering results, the scene clustering learning annotations, the behavior event clustering learning annotations, and the database business project clustering learning annotations to determine the first network learning cost variable corresponding to the visualization operation preference parsing network includes: Determine the network learning cost variable corresponding to the scene cluster component in the visualization operation preference parsing network in combination with the scene clustering result, the scene clustering learning annotation, the behavior event clustering learning annotation and the database business item clustering learning annotation; Determine the network learning cost variable corresponding to the behavior event cluster component in the visualization operation preference parsing network by combining the behavior event cluster result, the scene cluster learning annotation, the behavior event cluster learning annotation and the database business project cluster learning annotation; Determine the network learning cost variable corresponding to the database business project group component in the visualization operation preference parsing network in combination with the database business project group result, the scenario group learning annotation, the behavior event group learning annotation and the database business project group learning annotation; The first network learning cost variable is determined according to the network learning cost variable corresponding to the scenario cluster component, the network learning cost variable corresponding to the behavior event cluster component and the network learning cost variable corresponding to the database business item cluster component.

5. The method according to claim 4, characterized in that The method further comprises: Based on the set processing strategy, a divide-and-conquer operation is performed on the visualization operation preference learning annotation to obtain the scene group learning annotation, the behavior event group learning annotation, and the database business project group learning annotation; Wherein, the visualization operation preference parsing network includes: Image descriptor mining branch, scene feature mining branch, behavior event feature mining branch, database business project feature mining branch, first AI learning algorithm, second AI learning algorithm, disturbance feature addition branch, scene clustering component, behavior event clustering component, database business project clustering component and aggregation analysis branch; The image descriptor mining branch is used to perform image descriptor mining processing on the database operation behavior image record to obtain the image wavelet descriptor corresponding to the database operation behavior image record; the image wavelet descriptor includes multiple quantized image description members; The scene feature mining branch is used to perform scene feature change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain scene operation behavior description knowledge of the image wavelet descriptor; wherein the scene operation behavior description knowledge is a non-quantized image description member; The behavior event feature mining branch is used to perform behavior event feature change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain event operation behavior description knowledge of the image wavelet descriptor; wherein the event operation behavior description knowledge is a non-quantized image description member; The database business item feature mining branch is used to perform database business item feature change processing on the image wavelet descriptor, change multiple quantized image description members in the image wavelet descriptor into non-quantized image description members, and obtain the business item operation behavior description knowledge of the image wavelet descriptor; wherein the business item operation behavior description knowledge is the non-quantized image description member; The first AI learning algorithm is used to process the plurality of focus-plane operation behavior description knowledge respectively, and obtain the average description knowledge corresponding to the plurality of focus-plane operation behavior description knowledge respectively; wherein the average description knowledge is used to characterize the average positioning variable of the focus-plane operation behavior description knowledge in the knowledge relationship network; The second AI learning algorithm is used to process the plurality of focus-plane operation behavior description knowledge respectively, and obtain the deviation description knowledge respectively corresponding to the plurality of focus-plane operation behavior description knowledge; Among them, the deviation description knowledge is used to characterize the deviation positioning variable of the focus surface operation behavior description knowledge in the knowledge relationship network; the disturbance feature addition branch is used to combine the average description knowledge and the deviation description knowledge corresponding to the multiple focus surface operation behavior description knowledge respectively, add the target disturbance feature, and determine the quasi-cluster operation behavior description knowledge corresponding to the multiple focus surface operation behavior description knowledge respectively; the scene clustering component is used to perform knowledge clustering operation on the quasi-cluster operation behavior description knowledge of the scene focus surface, and obtain the scene focus surface clustering result of the operation behavior description knowledge; the behavior event clustering component is used to perform knowledge clustering operation on the quasi-cluster operation behavior description knowledge of the behavior event focus surface, and obtain the behavior event focus surface clustering result of the operation behavior description knowledge; the database business project clustering component is used to perform knowledge clustering operation on the quasi-cluster operation behavior description knowledge of the database business project focus surface, and obtain the database business project focus surface clustering result of the operation behavior description knowledge; the aggregation analysis branch is used to add the scene clustering result, the behavior event clustering result and the database business project clustering result to determine the visual operation preference analysis report of the database operation behavior image record.

6. A big data analysis system, characterized in that: include: A processor, a memory and a network interface; the processor is connected to the memory and the network interface; the network interface is used to provide a data communication function, the memory is used to store program code, and the processor is used to call the program code to execute the interactive behavior big data mining method applied to a visual database as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is run, the method for mining interactive behavior big data applied to a visual database as described in any one of claims 1 to 5 is executed.

8. A computer program product, characterized in that It includes a computer program or a computer executable instruction, which, when executed by a processor, implements the interactive behavior big data mining method applied to a visual database as described in any one of claims 1 to 5.

Images