A fusion processing method and platform for intelligence data

Abstract

The present invention discloses a fusion processing method and platform for intelligence data, and relates to the field of data processing technology. The method includes: constructing an element label table and a scene label table for intelligence data, and establishing an element label database based on historical intelligence data. The currently input intelligence data, intelligence data scenes and their labels are obtained, and the existing element label sets and label data sets are screened out, and the missing element label sets are identified at the same time. The data for completing the missing element labels in the label database is optimized to form an optimal completed label data set, and the scene matching and label matching are considered during the optimization. According to the existing label data set and the optimal completed label data set, the standardized processing scheme for intelligence data is optimized, and finally the intelligence data is standardized to generate a fusion processing result. Thereby achieving the technical effect of improving the efficiency and responsiveness of intelligence processing and improving the accuracy of fusion processing.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a fusion processing method and platform for intelligence data. Background Art

[0002] In intelligence analysis, the diversity and uncertainty of intelligence data make data fusion and processing complicated. Current intelligence processing methods often rely on manual judgment. Faced with raw intelligence data with diverse data sources, inconsistent formats and types, missing labels or incomplete information, there are technical problems such as low processing efficiency and real-time performance, and poor processing accuracy. Summary of the invention

[0003] The present application provides a method and platform for fusion processing of intelligence data, which is used to solve the technical problems of low processing efficiency and real-time performance, and poor processing accuracy existing in the prior art.

[0004] In view of the above problems, the present application provides a method and platform for fusion processing of intelligence data.

[0005] In a first aspect, the present application provides a method for fusion processing of intelligence data, the method comprising:

[0006] Pre-construct a scene label table for multiple intelligence data scenarios, and pre-construct an element label database based on historical intelligence data, wherein the element label database includes label data of multiple element labels of historical intelligence data; obtain the currently input intelligence data and intelligence data scenario, as well as the input label data of the intelligence data, and automatically screen the element labels according to the intelligence data scenario mapping scene label table to obtain an existing element label set and an existing label data set, as well as a missing element label set, wherein a scene label table is constructed according to the intelligence data scenario when the scene label table is not mapped; according to the missing element label set, optimize and complete the label data of multiple missing element labels in the missing element label set in the element label database to obtain an optimal completed label data set, wherein the completion fitness is analyzed according to the completed label data, and optimized completion is performed, and the completion fitness includes a scene matching degree and a label matching degree; according to the existing label data set and the optimal completed label data set, optimize the standardized processing scheme of the intelligence data to obtain an optimal standardized processing scheme, perform standardized processing on the intelligence data, and obtain a fusion processing result of the intelligence data.

[0007] In a second aspect, the present application provides a fusion processing platform for intelligence data, the platform comprising:

[0008] An element label module, the element label module is used to pre-build scene label tables for multiple intelligence data scenes, and to pre-build an element label database based on historical intelligence data, wherein the element label database includes label data of multiple element labels of historical intelligence data.

[0009] An intelligence screening module is used to obtain the currently input intelligence data and intelligence data scenario, as well as the input label data of the intelligence data, and automatically screen element labels according to the intelligence data scenario mapping scene label table to obtain an existing element label set and an existing label data set, as well as a missing element label set, wherein a scene label table is constructed according to the intelligence data scenario when the scene label table is not mapped.

[0010] A label optimization and completion module is used to optimize and complete the label data of multiple missing element labels in the missing element label set in the element label database according to the missing element label set, so as to obtain the optimal completed label data set, wherein the completion fitness is analyzed according to the completed label data, and the optimization and completion are performed, and the completion fitness includes scene matching and label matching.

[0011] The optimization and standardization module is used to optimize the standardized processing scheme of the intelligence data according to the existing label data set and the optimal completed label data set, obtain the optimal standardized processing scheme, perform standardized processing on the intelligence data, and obtain the fusion processing result of the intelligence data.

[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0013] The embodiment of the present application provides a method and platform for fusion processing of intelligence data, which pre-constructs a scene label table of multiple intelligence data scenarios, and establishes an element label database based on historical intelligence data, including label data of multiple element labels; obtains the currently input intelligence data, intelligence data scenario and its input label data, and screens out existing element label sets and existing label data sets, as well as missing element label sets, wherein a scene label table is constructed according to the intelligence data scenario when the scene label table is not mapped; according to the missing element label set, the label data is optimized and supplemented in the element label database to obtain the optimal supplemented label data set. The supplementation adaptability includes scene matching and label matching; according to the existing label data set and the optimal supplemented label data set, the standardized processing scheme of intelligence data is optimized to obtain the optimal standardized processing scheme, and the intelligence data is standardized to obtain the fusion processing result of the intelligence data. The method and platform for fusion processing of intelligence data disclosed in the present invention solves the technical problems of low processing efficiency and real-time performance and poor processing accuracy, and achieves the technical effect of improving intelligence processing efficiency and responsiveness and improving fusion processing accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of a method for fusion processing of intelligence data is provided for this application;

[0015] Figure 2 A structural diagram of a fusion processing platform for intelligence data is provided for this application.

[0016] Explanation of the accompanying drawings: element labeling module 11, intelligence screening module 12, label optimization and completion module 13, optimization and standardization module 14. DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0020] Embodiment 1: Figure 1 As shown, the present application provides a method for fusion processing of intelligence data, the method comprising:

[0021] Pre-construct a scenario label table for a variety of intelligence data scenarios, and pre-construct an element label database based on historical intelligence data, wherein the element label database includes label data for a variety of element labels of historical intelligence data.

[0022] Optionally, intelligence data refers to the discovered article data. The intelligence data is non-standardized, original data information, and needs to be reasonably collected, processed and organized to form standard literature data in the database.

[0023] Specifically, the element label table of intelligence data is a structured representation of the classification elements of multiple dimensions corresponding to intelligence data, which is used to describe and classify the key elements in intelligence data and is a classification and identification tool for intelligence data. Exemplarily, the element dimensions in the element label table include: Name: the title or name of the intelligence data or research, used to identify the data content; Subject type: describes the main object or subject of the data, such as organization, individual, equipment, etc.; Content type: describes the nature or form of the data, such as text, image, video, etc.; Research question: describes the core problem addressed by the research or data; Research method: describes the research method or technology used; Related field: indicates the academic or application field involved in the data or research.

[0024] Optionally, there is a corresponding relationship between the research question and the research method, such as the research method or technology of “UAV path planning” is “convolutional neural network”.

[0025] Specifically, the element label database is a database that stores and manages the label information of intelligence data, and is used to support the retrieval and analysis of intelligence data. The steps of pre-building the element label database include: data collection, data preprocessing, database design, element label extraction, element label table filling, data storage, database management and maintenance, etc. By building a structured element label database, it is convenient to systematically manage and analyze intelligence data. This database can not only improve the utilization efficiency of intelligence data, but also support the precision of intelligence analysis and decision-making.

[0026] In some embodiments, scenario label tables for multiple intelligence data scenarios are pre-built, and based on historical intelligence data, a factor label database is pre-built, including:

[0027] Construct an element label table for intelligence data, wherein the element label table includes a plurality of element labels; according to intelligence sources including label data with different element label combinations, set a plurality of intelligence data acquisition scenarios, and pre-construct a plurality of scenario label tables according to element labels containing label data and element labels not containing label data in the intelligence data in each intelligence data scenario; obtain a set of historical intelligence data, and extract label data of the plurality of element labels in each historical intelligence data to obtain a plurality of historical label data sets; based on the element label table, construct an index relationship between the historical intelligence data set and a plurality of historical label data sets to obtain an element label database.

[0028] Specifically, based on the element label table of intelligence data and combined with historical intelligence data, a element label database is pre-built. First, the element label table of intelligence data needs to be determined. This table includes a variety of element labels, such as name, subject type, content type, research question, research method, related fields, etc. Then, historical intelligence data is collected from different sources (such as databases, document libraries, patent libraries, etc.). Historical intelligence data may exist in the form of articles, reports, documents, etc. Next, each historical intelligence data is analyzed to extract the element label data contained therein. Exemplarily, it is obtained through natural language processing (NLP), manual annotation, rule matching and other paths.

[0029] Specifically, according to different combinations of element labels in the historical intelligence data analysis results, scenarios for obtaining multiple intelligence data are set. For example, different data scenarios include different combinations of labels. For example, the labels included in the literature scenario (such as keywords, subject majors, languages, fields, authors, classification numbers, publication dates, etc.) are different from the label combinations included in the announcement or information disclosure scenario (such as publishing department, publishing time, publishing font size, publishing agency, etc.).

[0030] Furthermore, the element labels of intelligence data are associated and classified with specific historical intelligence data to build an index relationship, and the intelligence data set is associated with the label data set to establish an element label database. The constructed element label database can display the content of each element label of each piece of intelligence data, which is convenient for subsequent data analysis and utilization.

[0031] Acquire the currently input intelligence data and intelligence data scenario, as well as the input label data of the intelligence data, perform automatic element label screening according to the intelligence data scenario mapping scene label table, obtain an existing element label set and an existing label data set, and a missing element label set, wherein when the scene label table is not mapped, construct a scene label table according to the intelligence data scenario.

[0032] In some embodiments, currently input intelligence data and intelligence data scenarios, as well as input label data of the intelligence data, are obtained, and automatic element label screening is performed according to the intelligence data scenario mapping scene label table to obtain an existing element label set and an existing label data set, as well as a missing element label set, wherein when the scene label table is not mapped, a scene label table is constructed according to the intelligence data scenario, including:

[0033] Acquire currently input intelligence data and intelligence data scenarios, and acquire input label data of the intelligence data, wherein the input label data includes a plurality of input element labels and a plurality of input label data; perform mapping in a plurality of scene label tables according to the intelligence data scenarios, and when a mapped scene label table exists, automatically perform element label screening and matching on the input label data; when a mapped scene label table does not exist, construct a corresponding scene label table according to the intelligence data scenario for storage, and automatically perform element label screening and matching on the input label data, use the plurality of input element labels as an existing element label set, use the plurality of input label data as an existing label data set, and obtain a missing element label set.

[0034] Specifically, first, the input intelligence data and intelligence data scenarios are obtained through the human-computer interaction device and the corresponding data interface, and then the inherent labels of the input intelligence data are extracted to obtain the input label data. Among them, the input label data includes multiple input element labels (such as title, content type, research question, etc.) and their corresponding multiple input label data (specific label content, such as specific research methods, fields involved, etc.).

[0035] Specifically, calling a predefined scenario label table corresponding to the input intelligence data scenario, and screening the input label data includes: matching existing element labels, that is, checking whether the input element labels already exist in the scenario label table, and if so, adding these labels to the existing element label set. In other words, the input label data included in the scenario label table is included in the existing label data set.

[0036] Specifically, by comparing the intelligence data scenario with the scenario label table, the labels that are not included in the input label data but defined in the scenario label table are identified. Then, based on the above-mentioned unidentified intelligence data scenario, the scenario label table is parsed and constructed, and the element labels corresponding to the intelligence data scenario that cannot be matched are the missing element label set. In other words, the unidentified intelligence data scenario is the newly added intelligence data scenario. Through the above steps, the intelligence data under the input newly added intelligence data scenario is parsed, and the input label data it has is classified and divided, so that a new scenario label table is established, and the scenario label table stored in the pre-built multiple intelligence data scenarios is updated, which improves the coverage of the intelligence data scenario, and at the same time helps to quickly call it when encountering the same intelligence data scenario in the future, further ensuring the comprehensiveness and consistency of the intelligence data, and providing a good foundation for subsequent data processing and analysis.

[0037] According to the missing element label set, in the element label database, the label data of multiple missing element labels in the missing element label set are optimized and completed to obtain the optimal completed label data set, wherein the completion fitness is analyzed according to the completed label data to perform optimized completion, and the completion fitness includes scene matching and label matching.

[0038] Specifically, the label data included in the above-mentioned input intelligence data may be missing due to various reasons, such as human omissions or poor integrity of the original data. Therefore, it is necessary to perform corresponding completion and optimization on the missing label data so that each document has data under the element label to facilitate subsequent retrieval.

[0039] Optionally, in order to select the most suitable completion data, the completion fitness is calculated. Completion fitness includes the following two aspects: scene matching and label matching. Among them, scene matching is to match the label information such as the background and field of the intelligence data with the label data in the feature label database to determine the confidence level of the scene. Label matching is an evaluation indicator that evaluates the consistency and relevance of the content of potential completion data based on the existing label data, ensuring the logical and content coherence between the completion data and the existing data.

[0040] Optionally, based on the calculated completion fitness, optimize and complete the label data of multiple missing feature labels. Select the data with the highest scores in scene matching and label matching as the optimal completed label dataset to ensure the integrity and consistency of the data. The completed dataset has complete feature label data, which is helpful for subsequent retrieval, classification and analysis.

[0041] In some embodiments, according to the missing element label set, optimizing and completing label data of multiple missing element labels in the missing element label set in the element label database includes:

[0042] In the element label database, the label data sets of multiple missing element labels in the missing element label set are indexed and obtained to construct a label data completion space; in the label data completion space, a first completion label data set is randomly selected to generate, and the first completion label data set includes multiple first completion label data of the multiple missing element labels; according to the existing label data set, the first label matching degree of the first completion label data set is analyzed, and according to the historical scene intelligence data of the intelligence data scene, the first scene matching degree of the first completion label data set is analyzed, and the first completion fitness is calculated; continue to optimize and complete the label data of multiple missing element labels in the label data completion space until convergence, output the completion label data set with the largest completion fitness, and obtain the optimal completion label data set.

[0043] Specifically, first, using the missing element label set obtained above as an index constraint, multiple groups of label data corresponding to multiple missing element labels in the missing element label set are retrieved and called to generate a label data set, which represents all possible label data of the missing element label set. Then, two matching degree analyses are performed, including: analyzing the label matching degree of the first completed label data set based on the existing label data set. This process can be completed through data matching and similarity calculation methods, such as cosine similarity, Jaccard similarity, etc. According to the historical scene intelligence data of the intelligence data scene, label matching is performed by traversing multiple label data in the first completed label data set, and based on the matching results, the scene matching degree of the first completed label data set is analyzed. Exemplarily, the more label data that matches the historical scene intelligence data with the first completed label data set, the higher the corresponding scene matching degree.

[0044] Specifically, the label matching degree and the scene matching degree are scored and the first completion fitness degree is calculated. The completion fitness degree comprehensively reflects the degree of adaptation between the label data and the current intelligence data, and is a quantitative representation of the matching.

[0045] Furthermore, the label data of multiple missing element labels are optimized and completed in the label data completion space, including selecting the label data with the highest completion fitness corresponding to each missing element label. Finally, the completed label data set with the largest completion fitness is output as the optimal completed label data set. The optimal completed label data set can provide more accurate and complete information for subsequent decisions, thereby improving the quality of decision-making.

[0046] The above method steps can ensure that the completed label data is highly matched with the original label data set and scene intelligence data, thereby improving the integrity and accuracy of the entire label data, by analyzing the label matching degree and scene matching degree in the label data completion space for multiple optimization completions.

[0047] In some implementations, analyzing the first label matching degree of the first complementary label data set according to the existing label data set, analyzing the first scene matching degree of the first complementary label data set according to the historical scene intelligence data of the intelligence data scene, and calculating and obtaining the first complementary adaptation degree includes:

[0048] Based on the label data processing record when the element label of the intelligence data has a missing element label set, multiple sample existing label data sets and multiple sample complementary label data sets are collected, and the sample label matching degree set is annotated to obtain the sample label matching degree set; the multiple sample existing label data sets, multiple sample complementary label data sets and sample label matching degree set are used to train the label matching analyzer; the label matching analyzer is used to perform label matching degree analysis on the existing label data set and the first complementary label data set to obtain the first label matching degree; according to the label data processing record when the intelligence data is processed in the intelligence data scenario, a scene matching analyzer is trained, wherein multiple scene existing label data sets and multiple scene complementary label data sets are collected, and the scene label matching degree set is annotated to obtain the scene label matching degree set, and the scene matching analyzer is trained; the scene matching analyzer is used to perform scene matching degree analysis on the existing label data set and the first complementary label data set to obtain the first scene matching degree; the first label matching degree and the first scene matching degree are weightedly calculated to obtain the first complementary adaptation degree.

[0049] Specifically, firstly, a plurality of sample existing label data sets and a plurality of sample completed label data sets are collected, and a sample label matching degree set is obtained by annotating. Then, the sample label matching degree set is used as training data to train a label matching analyzer. The label matching analyzer is constructed based on a machine learning model (such as a neural network, a support vector machine, etc.). Then, the trained label matching analyzer is used to perform label matching degree analysis on the existing label data set and the first completed label data set, thereby obtaining a first label matching degree.

[0050] Similarly, multiple scene existing label data sets and multiple scene completion label data sets are collected, and the scene label matching degree sets are annotated and obtained, and based on the same method principle as the above training label matching analyzer, a scene matching analyzer is constructed and trained. Then, the trained scene matching analyzer is used to perform scene matching degree analysis on the existing label data set and the first completion label data set, thereby obtaining the first scene matching degree.

[0051] Through the above-mentioned construction method based on neural network and sample data training, a scene matching analyzer and a label matching analyzer are obtained, which omits the complex process of defining and adjusting similarity evaluation rules, and is able to automatically learn and understand complex patterns in the data, thereby improving the process of similarity evaluation and evaluation system construction, and better adapting to various different data and tasks, thereby improving processing efficiency and accuracy.

[0052] In some implementations, continuing to optimize and complete the label data of multiple missing element labels in the label data completion space includes:

[0053] Continue to randomly select and generate a second supplementary label data set in the label data completion space; randomly exchange and update multiple first supplementary label data and multiple second supplementary label data of the multiple missing element labels in the first supplementary label data set and the second supplementary label data set to obtain a third supplementary label data set; analyze and calculate to obtain the second supplementary fitness and the third supplementary fitness, and retain the supplementary label data set corresponding to the largest supplementary fitness among the first supplementary fitness, the second supplementary fitness and the third supplementary fitness; continue to randomly select and generate a new supplementary label data set in the label data completion space, randomly exchange and update it with the retained supplementary label data set, and perform iterative optimization until the convergence optimization times are reached, output the supplementary label data set with the largest supplementary fitness, and obtain the optimal supplementary label data set.

[0054] Specifically, through a heuristic optimization method, the optimal complementary label data set is found in the label data completion space. First, in the label data completion space, a second complementary label data set is randomly selected to generate. Then, multiple first complementary label data and multiple second complementary label data of multiple missing element labels in the first complementary label data set and the second complementary label data set are randomly exchanged and updated to obtain a third complementary label data set. Next, the second complementary label data set and the third complementary label data set are analyzed by the aforementioned scene matching analyzer and label matching analyzer, and the second complementary fitness and the third complementary fitness are calculated. Then, the first complementary fitness, the second complementary fitness and the third complementary fitness are compared, and the complementary label data set with the largest complementary fitness is retained. Then, in the label data completion space, the above process is repeated, and a new complementary label data set is randomly selected to generate, and the complementary label data set is randomly exchanged and updated with the reserved complementary label data set. And analysis and comparison are performed to achieve iterative optimization until the preset number of convergence optimizations is reached, and the optimal complementary label data set, that is, the optimal complementary label data set, is output.

[0055] Through the above method steps, the optimal completed label data set can be effectively found in the label data completion space, thereby improving the completeness and accuracy of the label data. Due to the use of heuristic optimization methods, it is possible to avoid falling into the local optimal solution to a certain extent, making it more likely to find the global optimal solution.

[0056] According to the existing label data set and the optimal completed label data set, the standardized processing scheme of the intelligence data is optimized to obtain the optimal standardized processing scheme, the intelligence data is standardized to obtain the fusion processing result of the intelligence data.

[0057] In some embodiments, optimizing the standardized processing scheme of intelligence data based on the existing label data set and the optimal completed label data set to obtain the optimal standardized processing scheme includes:

[0058] Based on the existing labeled data set and the optimal completed labeled data set, a complete labeled data set of the intelligence data is generated; a standardized processing solution library for standardized processing of the intelligence data is obtained; based on the complete labeled data set, the standardized processing solution is optimized within the standardized processing solution library to obtain the optimal standardized processing solution.

[0059] Specifically, intelligence data from different sources (such as articles, documents, reports, etc.) may have different formats and need to be properly standardized to facilitate subsequent retrieval, call, and analysis of the intelligence data.

[0060] Specifically, first, a complete label data set of intelligence data is generated based on the existing label data set and the optimal complete label data set. This complete label data set contains all intelligence data and corresponding labels; then, a standardized processing solution library for intelligence data standardization is obtained, which contains a series of processing solutions, such as completing keywords, supplementing summaries, adjusting charts, adjusting fonts, etc. Then, based on the complete label data set, the solutions in the standardized processing solution library are optimized, including selecting the most suitable processing solution, or adjusting the parameters of the existing solution, etc. In order to find an optimal standardized processing solution, the standardized processing effect of intelligence data is the best. Finally, according to the optimal standardized processing solution obtained. Standardization is performed to make the format of articles from different sources consistent and improve the consistency and availability of data.

[0061] Through the above steps, the standardized processing of intelligence data can be effectively optimized, and the processing efficiency and accuracy can be improved. At the same time, due to the use of a complete labeled data set, this process can also better understand and process complex patterns in the data, thereby improving the effect of standardized processing.

[0062] In some implementations, optimizing the standardized processing scheme within the standardized processing scheme library to obtain the optimal standardized processing scheme includes:

[0063] According to the historical data of intelligence data standardization processing, multiple sample complete label data sets and sample standardization processing solution sets are collected, and the sample processing fitness sets are annotated; the multiple sample complete label data sets, sample standardization processing solution sets and sample processing fitness sets are used to train the standardization processing analyzer; a first standardization processing solution is randomly selected from the standardization processing solution library; the standardization processing analyzer is used to perform standardization processing analysis on the first standardization processing solution and the complete label data set to obtain a first processing fitness; the standardization processing solution optimization is continued in the standardization processing solution library until convergence, and the standardization processing solution with the largest processing fitness is output to obtain the optimal standardization processing solution.

[0064] Optionally, the processing scheme is optimized by training a standardized processing analyzer, including: collecting multiple sample complete label data sets and sample standardized processing scheme sets based on historical data of intelligence data standardized processing, and annotating and obtaining a sample processing fitness set. The sample processing fitness set is the training data set of the standardized processing analyzer mentioned above.

[0065] Optionally, multiple sample complete label data sets, sample standardization processing scheme sets and sample processing fitness sets are used to train a standardization processing analyzer. The standardization processing analyzer acquires the ability to evaluate the fitness of different standardization processing schemes by learning the complex mapping relationship between the above multiple data sets and sets.

[0066] Furthermore, a standardized processing scheme is randomly selected from the standardized processing scheme library as the first standardized processing scheme, and a standardized processing analyzer is used to perform standardized processing analysis on the first standardized processing scheme and the complete label data set, thereby obtaining a first processing fitness. Then, in the standardized processing scheme library, the standardized processing scheme is continuously optimized, including selecting a new processing scheme, or adjusting the parameters of an existing scheme, etc., until a preset convergence condition is reached, and then the standardized processing scheme with the largest processing fitness is output, i.e., the optimal standardized processing scheme.

[0067] Through the above steps, the standardization processing scheme can be effectively optimized to improve processing efficiency and accuracy. In addition, machine learning methods can also better understand and process complex patterns in data, thereby improving the effect of standardization processing.

[0068] Embodiment 2: Based on the same inventive concept as the method for fusion processing of intelligence data in the above embodiment, Figure 2 As shown, the present application provides a fusion processing platform for intelligence data, the platform comprising:

[0069] The element label module 11 is used to pre-build scenario label tables for various intelligence data scenarios, and to pre-build an element label database based on historical intelligence data, wherein the element label database includes label data for various element labels of historical intelligence data.

[0070] The intelligence screening module 12 is used to obtain the currently input intelligence data and intelligence data scenario, as well as the input label data of the intelligence data, and automatically screen the element labels according to the intelligence data scenario mapping scene label table to obtain an existing element label set and an existing label data set, as well as a missing element label set, wherein when the scene label table is not mapped, a scene label table is constructed according to the intelligence data scenario.

[0071] The label optimization and completion module 13 is used to optimize and complete the label data of multiple missing element labels in the missing element label set in the element label database according to the missing element label set, so as to obtain the optimal completed label data set, wherein the completion fitness is analyzed according to the completed label data, and the optimization and completion are performed, and the completion fitness includes scene matching and label matching.

[0072] The optimization and standardization module 14 is used to optimize the standardized processing scheme of the intelligence data according to the existing label data set and the optimal completed label data set, obtain the optimal standardized processing scheme, perform standardized processing on the intelligence data, and obtain the fusion processing result of the intelligence data.

[0073] Among them, the element label module 11 includes:

[0074] The element label table construction unit is used to construct an element label table of intelligence data, wherein the element label table includes multiple element labels.

[0075] The scenario label table construction unit is used to set up a variety of intelligence data scenarios according to the intelligence source of label data including different element label combinations, and pre-construct multiple scenario label tables according to the element labels of the intelligence data containing label data and the element labels not containing label data in each intelligence data scenario.

[0076] The historical label data set extraction unit is used to obtain the historical intelligence data set and extract the label data of the multiple element labels in each historical intelligence data to obtain multiple historical label data sets.

[0077] The element label database index relationship construction unit is used to construct the index relationship between the historical intelligence data set and multiple historical label data sets based on the element label table to obtain the element label database.

[0078] Furthermore, the intelligence screening module 12 includes:

[0079] The intelligence data acquisition and label extraction unit is used to acquire the currently input intelligence data and intelligence data scene, and acquire the input label data of the intelligence data, wherein the input label data includes multiple input element labels and multiple input label data.

[0080] The element label screening and classification unit is used to map the intelligence data scenario in multiple scene label tables, and when a mapped scene label table exists, automatically screen and match the input label data with element labels; when a mapped scene label table does not exist, a corresponding scene label table is constructed according to the intelligence data scenario for storage, and automatically screen and match the input label data with element labels, and the multiple input element labels are used as an existing element label set, and the multiple input label data are used as an existing label data set, and a missing element label set is obtained.

[0081] Furthermore, the tag optimization and completion module 13 includes:

[0082] The missing element label data set construction unit is used to index and obtain the label data sets of multiple missing element labels in the missing element label set in the element label database, and construct a label data completion space.

[0083] The first complement label data set generating unit is used to randomly select and generate a first complement label data set in the label data complement space, wherein the first complement label data set includes a plurality of first complement label data of the plurality of missing element labels.

[0084] The completion label matching analysis unit is used to analyze the first label matching degree of the first completion label data set according to the existing label data set, analyze the first scene matching degree of the first completion label data set according to the historical scene intelligence data of the intelligence data scene, and calculate and obtain the first completion adaptation degree.

[0085] The optimal completed label data set optimization unit is used to continue to optimize and complete the label data of multiple missing element labels in the label data completion space until convergence, output the completed label data set with the largest completion fitness, and obtain the optimal completed label data set.

[0086] Furthermore, the tag matching degree analysis unit in the tag optimization and completion module 13 includes:

[0087] The sample label data set collection and annotation unit is used to record the label data processing when the element label of the intelligence data is missing the element label set, collect multiple sample existing label data sets and multiple sample complete label data sets, and annotate to obtain the sample label matching degree set.

[0088] The label matching analyzer training unit is used to train the label matching analyzer by using the multiple sample existing label data sets, the multiple sample completed label data sets and the sample label matching degree set.

[0089] The first label matching degree analysis unit is used to use the label matching analyzer to perform label matching degree analysis on the existing label data set and the first completed label data set to obtain a first label matching degree.

[0090] A scene matching analyzer training unit is used to train a scene matching analyzer based on the label data processing records when processing intelligence data according to the intelligence data scenario, wherein multiple scene existing label data sets and multiple scene completed label data sets are collected, and the scene label matching degree sets are annotated to train the scene matching analyzer.

[0091] The first scene matching degree analysis unit is used to use the scene matching analyzer to perform scene matching degree analysis on the existing label data set and the first completed label data set to obtain a first scene matching degree.

[0092] The completion adaptation degree calculation unit is used to perform weighted calculation on the first tag matching degree and the first scene matching degree to obtain a first completion adaptation degree.

[0093] Furthermore, the optimal completion label data set optimization unit in the label optimization and completion module 13 includes:

[0094] The second completed label data set generating unit is used to continue to randomly select and generate a second completed label data set in the label data completion space.

[0095] The complementary label data exchange and update unit is used to randomly exchange and update multiple first complementary label data and multiple second complementary label data of the multiple missing element labels in the first complementary label data set and the second complementary label data set to obtain a third complementary label data set.

[0096] The complement fitness analysis and calculation unit is used to analyze and calculate to obtain the second complement fitness and the third complement fitness, and retain the complement label data set corresponding to the largest complement fitness among the first complement fitness, the second complement fitness and the third complement fitness.

[0097] The iterative optimization unit is used to continue to randomly select and generate a new completed label data set in the label data completion space, randomly exchange and update it with the retained completed label data set, and perform iterative optimization until the convergence optimization times are reached, output the completed label data set with the largest completion fitness, and obtain the optimal completed label data set.

[0098] Furthermore, the optimization and standardization module 14 includes:

[0099] A complete label data set generating unit is used to generate a complete label data set of the intelligence data according to the existing label data set and the optimal completed label data set.

[0100] The standardized processing solution library acquisition unit is used to acquire the standardized processing solution library for standardized processing of intelligence data.

[0101] The standardized processing scheme optimization unit is used to optimize the standardized processing scheme in the standardized processing scheme library according to the complete label data set to obtain the optimal standardized processing scheme.

[0102] Furthermore, the standardized processing scheme optimization unit in the optimization standardization module 14 includes:

[0103] The sample processing data set collection and annotation unit is used to collect multiple sample complete label data sets and sample standardization processing solution sets based on the historical data of intelligence data standardization processing, and annotate to obtain the sample processing fitness set.

[0104] The standardized processing analyzer training unit is used to train the standardized processing analyzer by using the multiple sample complete label data sets, the sample standardized processing scheme set and the sample processing fitness set.

[0105] The first standardized processing scheme selection unit is used to randomly select a first standardized processing scheme in the standardized processing scheme library.

[0106] The first processing fitness analysis unit is used to use the standardized processing analyzer to perform standardized processing analysis on the first standardized processing scheme and the complete label data set to obtain a first processing fitness.

[0107] The optimal standardized processing scheme optimization unit is used to continue to optimize the standardized processing scheme in the standardized processing scheme library until convergence, output the standardized processing scheme with the largest processing adaptability, and obtain the optimal standardized processing scheme.

[0108] Through the above-mentioned detailed description of a method for fusion processing of intelligence data in this specification, those skilled in the art can clearly understand a method and platform for fusion processing of intelligence data in this embodiment. For the platform disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0109] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for fusion processing of intelligence data, characterized in that: The method comprises: Pre-constructing scenario label tables for multiple intelligence data scenarios, and pre-constructing an element label database based on historical intelligence data, wherein the element label database includes label data for multiple element labels of historical intelligence data; Acquire currently input intelligence data and intelligence data scenario, as well as input label data of the intelligence data, automatically screen element labels according to the intelligence data scenario mapping scene label table, obtain existing element label sets and existing label data sets, and missing element label sets, wherein, when the scene label table is not mapped, construct a scene label table according to the intelligence data scenario; According to the missing element label set, in the element label database, optimizing and completing the label data of multiple missing element labels in the missing element label set to obtain an optimal completed label data set, wherein the completion fitness is analyzed according to the completed label data to perform optimization and completion, and the completion fitness includes scene matching and label matching; According to the existing label data set and the optimal completed label data set, a standardized processing scheme for intelligence data is optimized to obtain an optimal standardized processing scheme, and the intelligence data is standardized to obtain a fusion processing result of the intelligence data; The currently input intelligence data and intelligence data scenario, as well as the input label data of the intelligence data, are obtained, and the scene label table is mapped according to the intelligence data scenario to automatically screen the element labels, and an existing element label set and an existing label data set, as well as a missing element label set, wherein when the scene label table is not mapped, a scene label table is constructed according to the intelligence data scenario, including: Acquire currently input intelligence data and intelligence data scenario, and acquire input label data of the intelligence data, wherein the input label data includes a plurality of input element labels and a plurality of input label data; According to the intelligence data scenario, mapping is performed in multiple scene label tables. When a mapped scene label table exists, the input label data is automatically screened and matched with element labels. When a mapped scene label table does not exist, a corresponding scene label table is constructed according to the intelligence data scenario for storage, and the input label data is automatically screened and matched with element labels. The multiple input element labels are used as an existing element label set, the multiple input label data are used as an existing label data set, and a missing element label set is obtained.

2. The method for fusion processing of intelligence data according to claim 1, characterized in that: Pre-build scenario label tables for various intelligence data scenarios, and pre-build feature label databases based on historical intelligence data, including: Constructing an element label table of intelligence data, wherein the element label table includes multiple element labels; According to the intelligence sources of the label data including different element label combinations, multiple intelligence data scenarios are set up, and multiple scenario label tables are pre-built according to the element labels of the intelligence data including label data and the element labels not including label data in each intelligence data scenario; Acquire a historical intelligence data set, and extract label data of the multiple element labels in each historical intelligence data to obtain multiple historical label data sets; Based on the element label table, an index relationship between the historical intelligence data set and multiple historical label data sets is constructed to obtain an element label database.

3. The method for fusion processing of intelligence data according to claim 1, characterized in that: According to the missing element label set, in the element label database, optimizing and completing label data of multiple missing element labels in the missing element label set includes: In the element label database, index and obtain label data sets of multiple missing element labels in the missing element label set, and construct a label data completion space; In the label data completion space, randomly selecting and generating a first completed label data set, wherein the first completed label data set includes a plurality of first completed label data of the plurality of missing element labels; Analyze the first label matching degree of the first complementary label data set according to the existing label data set, analyze the first scene matching degree of the first complementary label data set according to the historical scene intelligence data of the intelligence data scene, and calculate and obtain the first complementary adaptation degree; Continue to optimize and complete the label data of multiple missing element labels in the label data completion space until convergence, output the completed label data set with the largest completion fitness, and obtain the optimal completed label data set.

4. The method for fusion processing of intelligence data according to claim 3, characterized in that: Analyzing the first label matching degree of the first complementary label data set according to the existing label data set, analyzing the first scene matching degree of the first complementary label data set according to the historical scene intelligence data of the intelligence data scene, and calculating and obtaining the first complementary adaptation degree, including: Based on the record of label data processing when the element label of intelligence data is missing the element label set, multiple sample existing label data sets and multiple sample complete label data sets are collected, and the sample label matching degree set is obtained by annotation; Using the multiple sample existing label data sets, the multiple sample complete label data sets and the sample label matching degree set, training a label matching analyzer; Using the label matching analyzer, performing label matching analysis on the existing label data set and the first completed label data set to obtain a first label matching degree; According to the label data processing record when the intelligence data is processed in the intelligence data scenario, a scene matching analyzer is trained, wherein a plurality of scene existing label data sets and a plurality of scene completed label data sets are collected, and a scene label matching degree set is obtained by annotating and acquiring, and the scene matching analyzer is trained; Using the scene matching analyzer, performing scene matching degree analysis on the existing label data set and the first completed label data set to obtain a first scene matching degree; A weighted calculation is performed on the first tag matching degree and the first scene matching degree to obtain a first completion adaptation degree.

5. The method for fusion processing of intelligence data according to claim 3, characterized in that: Continuing to optimize and complete the label data of multiple missing element labels in the label data completion space, including: Continue to randomly select and generate a second completed label data set in the label data completion space; Randomly exchanging and updating a plurality of first complementary label data and a plurality of second complementary label data of the plurality of missing element labels in the first complementary label data set and the second complementary label data set to obtain a third complementary label data set; Analyze and calculate to obtain a second complement fitness and a third complement fitness, and retain a complement label data set corresponding to the largest complement fitness among the first complement fitness, the second complement fitness and the third complement fitness; Continue to randomly select and generate a new completed label data set in the label data completion space, randomly exchange and update it with the retained completed label data set, and perform iterative optimization until the convergence optimization times are reached, output the completed label data set with the largest completion fitness, and obtain the optimal completed label data set.

6. The method for fusion processing of intelligence data according to claim 1, characterized in that: According to the existing label data set and the optimal completed label data set, the standardized processing scheme of the intelligence data is optimized to obtain the optimal standardized processing scheme, including: Generate a complete label data set of the intelligence data according to the existing label data set and the optimal completed label data set; Obtain a library of standardized processing solutions for standardized processing of intelligence data; According to the complete label data set, the standardized processing solution is optimized in the standardized processing solution library to obtain the optimal standardized processing solution.

7. The method for fusion processing of intelligence data according to claim 6, characterized in that: In the standardized processing solution library, the standardized processing solution is optimized to obtain the optimal standardized processing solution, including: Based on the historical data of intelligence data standardization processing, multiple sample complete label data sets and sample standardization processing solution sets are collected, and the sample processing fitness sets are annotated; Using the plurality of sample complete label data sets, the sample standardization processing scheme set and the sample processing fitness set, training a standardization processing analyzer; Randomly selecting a first standardized treatment solution in the standardized treatment solution library; Using the standardized processing analyzer, perform standardized processing analysis on the first standardized processing scheme and the complete label data set to obtain a first processing fitness; Continue to optimize the standardized processing scheme in the standardized processing scheme library until convergence, output the standardized processing scheme with the largest processing adaptability, and obtain the optimal standardized processing scheme.

8. An intelligence data fusion processing platform, characterized in that: The platform is used to execute the intelligence data fusion processing method according to any one of claims 1 to 7, and the platform includes: An element label module, the element label module is used to pre-build a scenario label table for multiple intelligence data scenarios, and pre-build an element label database based on historical intelligence data, wherein the element label database includes label data of multiple element labels of historical intelligence data; An intelligence screening module, the intelligence screening module is used to obtain currently input intelligence data and intelligence data scenarios, as well as input label data of the intelligence data, and automatically screen element labels according to the intelligence data scenario mapping scene label table, to obtain an existing element label set and an existing label data set, as well as a missing element label set, wherein when the scene label table is not mapped, a scene label table is constructed according to the intelligence data scenario; A label optimization and completion module, wherein the label optimization and completion module is used to optimize and complete the label data of multiple missing element labels in the missing element label set in the element label database according to the missing element label set, and obtain an optimal completed label data set, wherein the completion fitness is analyzed according to the completed label data, and the optimization and completion are performed, and the completion fitness includes scene matching and label matching; An optimization and standardization module, wherein the optimization and standardization module is used to optimize the standardization processing scheme of the intelligence data according to the existing label data set and the optimal completed label data set, obtain the optimal standardization processing scheme, perform standardization processing on the intelligence data, and obtain the fusion processing result of the intelligence data; The intelligence screening module includes: An intelligence data acquisition and label extraction unit, used to acquire currently input intelligence data and intelligence data scenarios, and to acquire input label data of the intelligence data, wherein the input label data includes a plurality of input element labels and a plurality of input label data; The element label screening and classification unit is used to map the intelligence data scenario in multiple scene label tables, and when a mapped scene label table exists, automatically screen and match the input label data with element labels; when a mapped scene label table does not exist, a corresponding scene label table is constructed according to the intelligence data scenario for storage, and automatically screen and match the input label data with element labels, and the multiple input element labels are used as an existing element label set, and the multiple input label data are used as an existing label data set, and a missing element label set is obtained.

Images