Method and system for independent application of fused and archived multi-modal data

Abstract

The invention discloses a method and system for independent application of multi-modal data after fusion and archiving, and relates to the technical field of archive management. The technical key points are as follows: establishing a service management platform, including a user group and an archive group, managing the user group and the archive group, defining an archive as a node by adopting a chain network structure, starting archiving management, generating an archiving package, and performing independent application between nodes, so that a user can quickly retrieve and position the archive; effective interaction between data is realized to a certain extent; state data of an archiving packet is processed through a target decision network, an archiving score is calculated, a corresponding archiving interval is matched according to the archiving score, different calling levels are correspondingly set, different users are endowed with corresponding permissions for management, and the whole system judges the management quality by designing evaluation factors, so that the management efficiency is improved. And the management efficiency is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of archive management, and in particular to a method and system for independently applying multimodal data after fusion and archiving. Background Art

[0002] With the development of social economy, socialized production methods have begun to develop towards digitalization and intelligence, and are widely used in digital application systems in various industries. Electronic archives are usually used to manage them. In business systems, different types of modal data (for example, text, images, audio, etc.) will appear. However, when multimodal data is archived and used independently, it is usually impossible to fully utilize the advantages of the fused data, resulting in the functions and values ​​of some modal data not being effectively utilized; when multimodal data is fused, some modal data may form "islands" and cannot effectively interact and analyze with other data sets. Maintaining the permission management and data independent application functions of the multimodal fusion system may require higher technical support and costs. For example, users may need multiple authentications or different access paths when using independent applications, which increases the complexity of management. Summary of the invention

[0003] 1. Technical issues to be resolved In view of the deficiencies in the prior art, the present invention provides a method and system for independent application of multimodal data after fusion and archiving. The system includes: a business management platform, a data acquisition and fusion unit, an archiving management unit, a retrieval and analysis unit, and a storage and calling unit. By defining the archive as a node and adopting a chain network structure, the multimodal data is processed into an archive package, thereby achieving effective interaction between data. The status data of the archive package is processed through a target decision network, the archive score is calculated, and the corresponding archive interval is matched according to the archive score. At the same time, different calling levels are set to manage users, which improves the management efficiency to a certain extent. The quality of management is determined by designing evaluation factors, thereby solving the problems raised in the background technology.

[0004] (II) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: In a first aspect, the present application proposes a method for independent application of multimodal data fusion and archiving, the method comprising the following steps: S1, obtaining input multimodal data, and fusing the multimodal data to generate a target data set; wherein the input multimodal data type includes structured data, unstructured data, and Internet of Things data; S2, transfer the target data set to the business management platform, design a chain network structure to start archiving management, generate an archive package, and build a target decision network based on the deep learning network, calculate the archiving score, match the archive partition according to the archiving score and provide business management services; the business management platform includes archive groups and user groups, and the target decision network is obtained by reinforcement learning exploration of the sample environment, which is composed of corresponding information in at least two historical archive packages; S3, enter the business management platform for search and service analysis, perform data query on the business chain, generate a business chain set, and generate evaluation factors based on the business chain set to evaluate the business management service; generate statistical reports based on the node information of the business chain; among them, service analysis includes statistical analysis, viewing details and generating reports; S4, file update, modification and storage.

[0005] Furthermore, the process of fusing multimodal data includes: Data preparation: Collect input multimodal data; Data processing: Processing of multimodal data, including cleaning, conversion, loading and unification; Feature extraction: Extract features from the processed multimodal data. The extracted information includes target requirements, retention period, file confidentiality, open form, and environment. Splicing features: merge multimodal data at different feature levels and splice the features; Data normalization: Convert the concatenated features into a one-dimensional tensor to generate the target dataset.

[0006] Furthermore, the steps of archive management include: Each archive is defined as a node, and a chain network structure is adopted to traverse the archives through a linked list, and the archives and the corresponding target data sets are integrated into an archive package, and an archive group is formed according to a number of archive packages, and a chain network structure information set is generated according to a number of nodes; wherein the archive package at least includes: archive ID, archive name, archive list, archive content, storage requirements, storage path and creation date; the archive score of the corresponding information in the archive package is obtained, and the archive score is obtained by processing the state data of the archive package through the target decision network, and the state data includes information level data and integrity data; Information-level data: the identifier of the corresponding information, the data type of the corresponding information, the filing status of the corresponding information, and the filing requirements of the corresponding information; wherein the filing status includes an unfiled status and a filed status; Overall-level data: the importance of the corresponding information and the sensitivity of the corresponding information.

[0007] Furthermore, the specific process of obtaining the archiving score is as follows: Target requirements, retention period, document confidentiality, and openness are extracted from the extracted feature information and defined as key factors. Each key factor is assigned a corresponding weight and score, and the weight and score are kept synchronized. Target requirements: core files are assigned a weight of ε3 > important files are assigned a weight of ε2 > ordinary files are assigned a weight of ε1; Preservation period: long-term, weighted ε3> medium-term, weighted ε2> short-term, weighted ε1; Based on the target requirements and preservation period, multiply the score of each factor by its weight, and then add up all the results to get the importance level; Document confidentiality: Business secrets, assigned a weight of ε3 > personal privacy, assigned a weight of ε2 > other sensitive information, assigned a weight of ε1; Open form: secret approval, weighted ε3> internal disclosure, weighted ε2> public, weighted ε1; Based on the confidentiality and openness of the document, the score of each factor is multiplied by its weight, and then all the results are added together to get the sensitivity level; Then an archive score is generated and the archive partition is matched according to the archive score, based on the formula: ; In the formula, de represents the archiving score, zy represents the importance score, mg represents the sensitivity score, β1 and β2 are weight coefficients, and both β1 and β2 are greater than 0.

[0008] Furthermore, the steps of setting the archive interval include: The database is divided into M grid intervals and defined as archiving intervals. The storage information in the database is allocated to the M archiving intervals in sequence according to the archiving scores. The corresponding calling permissions are set for any archiving interval. N calling levels are set, and there are n calling ports under any calling level Ni. Different permission ranges are set according to the levels.

[0009] Furthermore, the steps of performing business management services include: The user registers or logs in, and the user has a unique user ID; After the user logs in or registers successfully, the corresponding subject attributes are loaded according to the user, including employee, position and responsibilities; the permission description is configured according to the subject attributes, and the user ID is added to the user's permission description. The role permission is configured according to the permission description, and the files of the file group are managed. If the file group does not exist, a new file is added through the business management platform, and the administrator of the user group is set to the user ID in the current permission description.

[0010] Furthermore, the steps of S3 include: Verify the identity and permissions of user groups and filter based on RABC permissions; The menu of the interface is displayed according to the user's permission information; the user group uses the archive search and analysis services on the corresponding interface to perform global search, advanced search and viewing of the archive group; By querying the chain network structure information set that meets the conditions through the database, data query is performed on the chain network structure that meets the requirements to obtain a business chain set, which includes a query vector, a key vector and a value vector; by constructing a three-dimensional coordinate system, and taking the query vector as the X-axis, the key vector as the Y-axis, and the value vector as the Z-axis, a vector feature space is established, a three-dimensional contour map is drawn on the vector feature space, and the three-dimensional contour map is superimposed with the preset standard three-dimensional contour map, and its horizontal and vertical screenshots are obtained and compared to generate the total overlapping area and the total non-overlapping area, wherein the total non-overlapping area is the offset during the search response; the evaluation factors in the search process are obtained according to the total overlapping area and the total non-overlapping area, and the quality of the business management service is determined; The formula on which the evaluation factor is based is: ; In the formula, BP represents the evaluation factor, k0 represents the correction weight coefficient, cmj represents the overlapping area, and fmj represents the non-overlapping area; Obtain the node information of the business chain, which includes operation logs, archive status, and related archive packages, identify and collect the user ID corresponding to the operation node information, count the user's operation behavior, and generate statistical reports.

[0011] In the second aspect, the present application proposes a system for independent application after fusion and archiving of multimodal data, including: a business management platform, a data acquisition and fusion unit, an archiving management unit, a retrieval and analysis unit, and a storage and calling unit; A data acquisition and fusion unit acquires input multimodal data and fuses the multimodal data to generate a target data set; wherein the input multimodal data type includes structured data, unstructured data, and IoT data; The archiving management unit transfers the target data set to the business management platform, designs a chain network structure to start archiving management, generates an archiving package, and builds a target decision network based on a deep learning network, calculates the archiving score, matches the archiving partition according to the archiving score, and performs business management services; wherein the target decision network is obtained by reinforcement learning exploration of the sample environment, and the sample environment is composed of corresponding information in at least two historical archive packages; The retrieval and analysis unit enters the business management platform to perform retrieval and service analysis, performs data query on the business chain, generates a business chain set, and generates evaluation factors based on the business chain set to evaluate the business management service; generates statistical reports based on the node information of the business chain; among them, service analysis includes statistical analysis, viewing details, and generating reports; The storage and call unit is used to manage and control the user's access rights to archive storage and archives. Based on user identity and role, different permission levels and access control policies are defined to ensure that only authorized users can access and operate archives; The business management platform includes archive groups and user groups and is used to manage various operations and activities of archive groups and user groups.

[0012] In a third aspect, the present application proposes a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method proposed in the first aspect when executing the computer program.

[0013] In a fourth aspect, the present application proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method proposed in the first aspect.

[0014] (III) Beneficial effects The present invention provides a method and system for independent application of multimodal data after fusion and archiving, which has the following beneficial effects: The present invention builds a business management platform, sets up archive groups and user groups, performs distributed management on the archive groups and user groups, and designs a chain network structure with archives as nodes. Each node can process data independently and is regarded as an independent module without relying on a central server. At the same time, the chain structure allows new nodes to be added easily, supports network expansion, and is convenient for users to quickly retrieve and locate archives, so as to realize information storage and management more flexibly. The dependency relationship between archives can be expressed more accurately, reducing unnecessary coupling. The present invention builds a target decision network based on a deep learning network, focuses on the status data of the archive package, especially the overall level data, calculates the archive score based on the sensitivity score and the importance score, and divides the archive interval according to the size of the archive score; and improves the resource utilization of the network by training the network using reinforcement learning; The present invention searches the archive group through the user group to obtain a business chain set, including a query vector, a key vector and a value vector. Through vectorized representation, the system can more accurately understand the user's search intention, reduce irrelevant results, and improve the accuracy of the search. By constructing a feature space, the evaluation factor of the search process is calculated, resources are reasonably allocated, work efficiency is improved, and response speed is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flowchart of a method for independent application of multimodal data fusion and archiving according to the present invention; Figure 2 The module diagram of a system for independent application after fusion and archiving of multimodal data according to the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 This embodiment provides a method for independently applying multimodal data after fusion archiving, the method comprising the following steps: S1, obtain input multimodal data, and fuse the multimodal data to generate a target data set; The process of fusing multimodal data includes: Data preparation: Collect input multimodal data; the input multimodal data types include structured data, unstructured data, and IoT data; Structured data refers to all kinds of information stored in digital form, such as database tables, spreadsheets, etc. Unstructured data refers to information that cannot be directly calculated by computers, including text data, image data, audio data, and video data; IoT data refers to information obtained by various sensors stored in digital form, such as: the location of the server where the monitoring file is located; Data processing: Processing of multimodal data, including cleaning, conversion, loading and unification; It should be noted that cleaning: deduplication, processing of missing values, format standardization and outlier detection; conversion: format conversion and encoding; for example, converting categorical data into numerical form; loading: selecting appropriate data storage formats according to needs, batch loading of large-scale data and incremental updates to ensure that the processed data is the latest; unification: semantic alignment; for example, associating text descriptions with images to ensure semantic correspondence; Feature extraction: Extract features from the processed multimodal data. The extracted information includes target requirements, retention period, file confidentiality, open form, and environment. It should be noted that, by combining NLP technology and named entity recognition (NER) technology, features are extracted from the processed multimodal data. The meaning of each extracted feature is: Target requirement: identifying the purpose or target of the data; Preservation period: identifying the preservation time requirement of the data; File confidentiality: identifying the confidentiality level of the data; Open form: understanding the degree of openness of the data and whether it is provided in a public form; Environment: identifying the application environment of the data; Specifically, NLP technology is used for environment and storage period; named entity recognition (NER) technology is used to extract file confidentiality, open form and target requirements; to fully extract target requirements, storage period, file confidentiality, open form and environment from data, multiple technologies must be combined to achieve it. A single technology is difficult to meet all needs because different features have different characteristics: Regarding environment and shelf life: It is necessary to understand the semantics of the environment and the preservation period, such as "server room", "preserve for 5 years", "preserve until 2025". Natural language processing (NLP) technology can be used to identify target requirements and preservation periods, such as converting "preserve until 2025" into a specific time and "server room" into a specific address. If necessary, the positioning navigation system can be used to obtain address information. Regarding confidentiality, openness and target requirements: These features are often professional terms that require domain knowledge. Named Entity Recognition (NER) technology can be used to understand the deep meaning and context of the text in combination with the archives domain knowledge base, such as the archives category library and the archives protection level library, to identify and classify them; Splicing features: merge multimodal data at different feature levels and splice the features; Select appropriate features, namely target requirements, retention period, document confidentiality, open form, and environment, and combine features based on domain knowledge and experience, for example, "target requirements" + "retention period", "document confidentiality" + "environment", to form new features; Data standardization: convert the concatenated features into a one-dimensional tensor to generate the target data set; If you need to concatenate several features, you can use numpy's concatenate or hstack methods. You can use numpy's reshape method to convert the features of each sample into a one-dimensional tensor. You can use these features as input to generate labels and integrate them into a data set. S2, transfer the target data set to the business management platform, design a chain network structure to start archiving management, generate an archive package, and build a target decision network based on the deep learning network, calculate the archiving score, match the archive partition according to the archiving score and provide business management services; the business management platform includes archive groups and user groups, and the target decision network is obtained by reinforcement learning exploration of the sample environment, which is composed of corresponding information in at least two historical archive packages; It should be noted that the target decision network is a reinforcement learning-based model used to make decisions from sample environments; it learns and optimizes by exploring sample environments and using historical data and experience; The steps of archive management include: Each archive is defined as a node, and a chain network structure is adopted to traverse the archives through a linked list, and the archives and the corresponding target data sets are integrated into archive packages, and an archive group is formed according to a number of archive packages, and a chain network structure information set is generated according to a number of nodes; wherein the archive package at least includes: archive ID, archive name, archive list, archive content, storage requirements, storage path and creation date, and each archive package also includes data collected within a specific time period, which is process information generated by using a target decision network, including action information, reward signal and timestamp; It should be noted that the action information represents the action or decision taken under a specific state; the reward signal represents the reward or punishment based on the feedback of the action taken, which is usually used to evaluate the effect of the decision; the timestamp is the time information of each record, which is used to analyze the changes in the time series; in this embodiment, the sample environment is strengthened by a sample environment composed of information-level data and holistic data, and different actions are selected to interact with the environment, and state changes and reward feedback are observed. By exploring the obtained data, the decision strategy is continuously adjusted to maximize the expected reward, and the data in the historical archive package is used for playback learning to improve the learning efficiency and stability of the network; subsequently, the state data of the archive package is processed according to the target decision network to obtain the archive score, including the importance score and the sensitivity score; the reward function value can be obtained by weighting the scores of each item, and the target decision network is updated according to the reward function value; Obtaining an archive score of corresponding information in the archive package, wherein the archive score is obtained by processing the state data of the archive package through the target decision network, and the state data includes information level data and integrity data; Information-level data: the identification of the corresponding information, the data type of the corresponding information, the filing status of the corresponding information, and the filing requirements of the corresponding information; the filing status includes the unfiled status and the filed status; Overall-level data: the importance of the corresponding information, the sensitivity of the corresponding information; The specific process of obtaining the archiving score is as follows: Target requirements, retention period, document confidentiality, and openness are extracted from the extracted feature information and defined as key factors. Each key factor is assigned a corresponding weight and score, and the weight and score are kept synchronized. Target requirements: core files are assigned a weight of ε3 > important files are assigned a weight of ε2 > ordinary files are assigned a weight of ε1; Preservation period: long-term, weighted ε3> medium-term, weighted ε2> short-term, weighted ε1; Based on the target requirements and preservation period, multiply the score of each factor by its weight, and then add up all the results to get the importance score; Document confidentiality: Business secrets, assigned a weight of ε3 > personal privacy, assigned a weight of ε2 > other sensitive information, assigned a weight of ε1; Open form: secret approval, weighted ε3> internal disclosure, weighted ε2> public, weighted ε1; Based on the confidentiality and openness of the file, the score of each factor is multiplied by its weight, and then all the results are added together to get the sensitivity score; Assume that weight ε3, weight ε2 and weight ε1 are 3, 2 and 1 respectively; Data 1: Target requirement: core archives (weight ε3=3) = 9 points; preservation period: mid-term (weight ε2=2) = 4 points; the importance score is: 9+4=13; Document confidentiality: business secret (weight ε3=3) = 9 points; open form: secret approval: (weight ε3=3) = 9 points; then the sensitivity score is: 9+9=18; Data 2: Target requirement: important archives (weight ε2=2) = 4 points; preservation period: medium term (weight ε2=2) = 4 points; then the importance score is: 4+4=8; Document confidentiality: trade secret (weight ε3=3) = 9 points; Open form: internal disclosure: (weight ε2=2) = 4 points; then the sensitivity score is: 9+4=13; Through simple calculations, we can see that the importance score of data 1 is greater than that of data 2, and the sensitivity score of data 1 is greater than that of data 2; It should be noted that for general business systems: Core files: files that have a critical impact on the operation, management, and legal compliance of an organization or institution, such as a company's business license; important files are files that have a significant impact on the operation and decision-making of an organization and may be used more frequently; ordinary files: files that have a significant impact on the operation and decision-making of an organization, such as project plans and progress reports; ordinary files: files that have a minor impact on the daily operation of an organization, such as issued management documents, etc. Long-term: refers to archives that are kept for a long time, usually for more than 10 years, and sometimes even permanently; Medium-term: refers to archives that are kept for a period of 3 to 10 years; Short-term: refers to archives that are kept for 3 years or less; Trade secrets: refers to information with economic value that is not known to the public, such as research and development data, customer lists, marketing strategies, etc. Personal privacy: involves personal identity and contact information; Secret approval: refers to the approval process of certain sensitive or confidential files that are strictly controlled, usually only certain personnel (such as senior management or security department) can access them; internal disclosure: refers to information sharing within the organization, where files are available for review by specific employees or departments but not to the outside world; public: files are open to the outside world and can be accessed and reviewed by anyone; Generate an archive score based on the importance score and the sensitivity score, and match the archive partition according to the archive score, based on the formula: ; In the formula, de represents the filing score, zy represents the importance score, mg represents the sensitivity score, β1 and β2 are weight coefficients, and β1+β2=1; it should be noted that the larger the importance score and the sensitivity score, the higher the filing score value, which means the higher the authority to call the information; The steps for setting the archiving interval include: The database is divided into M grid intervals and defined as archive intervals. The stored information in the database is allocated to the M archive intervals in sequence according to the archive scores. The corresponding call permissions are set for any archive interval. N call levels are set, and there are n call ports under any call level Ni. Different permission ranges are set according to the levels to achieve authorized information management between multiple levels and ensure information privacy. Among them, several levels are vertically integrated, and the corresponding level of authority is set according to the level to manage the corresponding range of information; several ports of the same level are chain-integrated, and the port username is set. A signing mechanism is set for any data call and transmission link; for example, there are 4 levels of call levels, from top to bottom: A-level system administrator, B-level archive auditor, C-level archive creator and D-level archive queryer, and the storage information of the 4 levels is set with 4 levels of authority, which are set in sequence according to the archiving score. No query, call or other operations are allowed if they are not within the authority. The steps for providing business management services include: The user registers or logs in, and the user has a unique user ID; After the user logs in or registers successfully, the corresponding subject attributes are loaded according to the user, including employee, position and responsibilities; The permission description is configured according to the subject attributes, and the user ID is added to the user's permission description. The role permission is configured according to the permission description, and the files of the file group are managed. If the file group does not exist, a new file is added through the business management platform, and the administrator of the user group is set to the user ID in the current permission description; S3, enter the business management platform for search and service analysis, perform data query on the business chain, generate a business chain set, and generate evaluation factors based on the business chain set to evaluate the business management service; generate statistical reports based on the node information of the business chain; among them, service analysis includes statistical analysis, viewing details and generating reports; The specific process is: Verify the identity and permissions of user groups and filter based on RABC permissions; It should be noted that RABC is a role-based permission management mechanism. In this embodiment, dynamic permission control can be achieved by combining the storage of the chain network structure. For example, in combination with the above four call levels, each call level includes several users, and the trust value is set according to the user's subject attribute. The setting range of the trust value is a real number [0, 1], where 1 represents complete trust and 0 represents untrustworthiness; and through the set trust evaluation model, four levels of several call levels are defined. Usually, the initial trust value of a newly registered user is set to 0.5. With the relevant behavior operations (such as retrieval behavior, increase, decrease) of the registered user in the system, the trust value is changed (such as increase or decrease); Assuming that the behavior parameters of the corresponding user are obtained within the time period T, the trust value of the user is calculated based on the behavior parameters, and the formula is as follows: ; In the formula, R(T) is the trust value at the current moment, X(t) represents the user's behavior parameter, α is the weight coefficient, and α is greater than 0; The updated trust value is based on the formula: ; Where H(t) is the updated trust value, H(tT) is the trust value of the user at time t-Ts, ω1 and ω2 are weight coefficients, and ω1+ω2=1; Determine the permission level of the registered user based on the current trust value; generally speaking, users with high trust values ​​will obtain higher-level permission operations (high trust thresholds), while users with low trust values ​​will not be able to perform permission operations with trust thresholds higher than their trust values; it should be noted that the system administrator is defined as having the highest trust value, and the A-level system administrator, B-level file auditor, C-level file creator, and D-level file queryer are defined by calculating the user's trust value, and the trust value corresponding to A-level>the trust value corresponding to B-level>the trust value corresponding to C-level>the trust value corresponding to D-level; The menu of the interface is displayed according to the user's permission information; the user group uses the archive search and analysis services on the corresponding interface to perform global search, advanced search and viewing of the archive group; The chain network structure information set that meets the conditions is queried through the database, and the data query is performed on the chain network structure that meets the requirements to obtain a business chain set, which includes a query vector, a key vector and a value vector; a three-dimensional coordinate system is constructed, and the query vector is used as the X-axis, the key vector is used as the Y-axis, and the value vector is used as the Z-axis to establish a vector feature space, and a three-dimensional contour map is drawn on the vector feature space. The three-dimensional contour map is also superimposed with the preset standard three-dimensional contour map, and its horizontal and vertical screenshots are obtained and compared to generate the total overlapping area and the total non-overlapping area, where the total non-overlapping area is the offset during the retrieval response; the evaluation factor in the retrieval process is obtained according to the total overlapping area and the total non-overlapping area, and the formula based on it is: ; In the formula, BP represents the evaluation factor, k0 represents the correction weight coefficient, cmj represents the overlapping area, and fmj represents the non-overlapping area; The evaluation factor BP in the retrieval process is obtained and its standard deviation is calculated to generate the standard efficiency value CP in the retrieval process. The formula is as follows: ; In the formula, g represents a multiple, which is usually 2 to 3, and σ represents the standard deviation. The evaluation factor BP is also compared with the standard efficiency value CP. When BP≤CP, the first evaluation character is generated, when CP<BP≤CP, the second evaluation character is generated, and when BP>CP, the third evaluation character is generated. The first evaluation character, the second evaluation character, and the third evaluation character represent "excellent", "good", and "poor". Obtain node information of the business chain, including operation logs, archive status, and related archive packages, identify and collect user IDs corresponding to the operation node information, count the user's operation behavior, and generate statistical reports; S4, file update, modification and storage.

[0018] In summary, the system includes a business management platform, a data acquisition and fusion unit, an archiving management unit, a retrieval and analysis unit, and a storage and calling unit. The business management platform includes user groups and archive groups, and performs distributed management on user groups and archive groups. By defining archives as a node and adopting a chain network structure, archiving management is enabled, and the fused multimodal data is processed into an archive package. Nodes are applied independently to facilitate users to quickly retrieve and locate archives. To a certain extent, effective interaction between data is achieved. The status data of the archive package is processed through the target decision network, the archive score is calculated, and the corresponding archive interval is matched according to the archive score. At the same time, different call levels are set to manage users. The entire system designs an evaluation factor to determine the quality of management, which improves the efficiency of management to a certain extent.

[0019] Example 2: This example provides a system for independent application of multimodal data fusion and archiving. Figure 2 ,The system includes: a business management platform, a data acquisition and fusion unit, an archiving management unit, a retrieval and analysis unit, and a storage and calling unit; A data acquisition and fusion unit acquires input multimodal data and fuses the multimodal data to generate a target data set; wherein the input multimodal data type includes structured data, unstructured data, and IoT data; The archiving management unit transfers the target data set to the business management platform, designs a chain network structure to start archiving management, generates an archiving package, and builds a target decision network based on a deep learning network, calculates the archiving score, matches the archiving partition according to the archiving score, and performs business management services; wherein the target decision network is obtained by reinforcement learning exploration of the sample environment, and the sample environment is composed of corresponding information in at least two historical archive packages; The retrieval and analysis unit enters the business management platform to perform retrieval and service analysis, performs data query on the business chain, generates a business chain set, and generates evaluation factors based on the business chain set to determine the efficiency of the business management service; generates statistical reports based on the node information of the business chain; among them, service analysis includes statistical analysis, viewing details, and generating reports; The storage and call unit is used to manage and control the user's access rights to archive storage and archives. Based on user identity and role, different permission levels and access control policies are defined to ensure that only authorized users can access and operate archives; The business management platform includes archive groups and user groups and is used to manage various operations and activities of archive groups and user groups.

[0020] In the application, the several formulas involved are all calculated by removing dimensions and taking their numerical values, and the formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The formula is set by technical personnel in this field according to actual conditions.

[0021] Embodiment 3: Embodiment 3 of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method of Embodiment 1 when executing the computer program.

[0022] In addition, this embodiment further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method of Embodiment 1 are implemented.

[0023] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0024] In a possible implementation manner, the present invention may also be implemented in the form of a program product, which includes a program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps of implementing the method in Example 1.

[0025] The program code for executing the present invention may be written in any combination of one or more programming languages, and may be executed entirely on a user device, partially on a user device, as an independent software package, partially on a user device and partially on a remote device, or entirely on a remote device.

[0026] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0027] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for independent application of multimodal data after fusion and archiving, characterized in that: The method comprises the following steps: S1, obtaining input multimodal data, and fusing the multimodal data to generate a target data set; wherein the input multimodal data type includes structured data, unstructured data, and Internet of Things data; S2, transfer the target data set to the business management platform, design a chain network structure to start archiving management, generate an archive package, and build a target decision network based on the deep learning network, calculate the archiving score, match the archive partition according to the archiving score and provide business management services; the business management platform includes archive groups and user groups, and the target decision network is obtained by reinforcement learning exploration of the sample environment, which is composed of corresponding information in at least two historical archive packages; S3, enter the business management platform for search and service analysis, perform data query on the business chain, generate a business chain set, and generate evaluation factors based on the business chain set to evaluate the business management service; generate statistical reports based on the node information of the business chain; among them, service analysis includes statistical analysis, viewing details and generating reports; S4, file update, modification and storage.

2. The method for independent application after fusion and archiving of multimodal data according to claim 1, characterized in that: The process of fusing multimodal data includes: Data preparation: Collect input multimodal data; Data processing: Processing of multimodal data, including cleaning, conversion, loading and unification; Feature extraction: Extract features from the processed multimodal data. The extracted information includes target requirements, retention period, file confidentiality, open form, and environment. Splicing features: merge multimodal data at different feature levels and splice the features; Data normalization: Convert the concatenated features into a one-dimensional tensor to generate the target dataset.

3. The method for independent application after fusion and archiving of multimodal data according to claim 1, characterized in that: The steps of archive management include: Each archive is defined as a node, and a chain network structure is adopted to traverse the archives through a linked list, and the archives and the corresponding target data sets are integrated into an archive package, and an archive group is formed according to a number of archive packages, and a chain network structure information set is generated according to a number of nodes; wherein the archive package at least includes: archive ID, archive name, archive list, archive content, storage requirements, storage path and creation date; the archive score of the corresponding information in the archive package is obtained, and the archive score is obtained by processing the state data of the archive package through the target decision network, and the state data includes information level data and integrity data; Information-level data: the identifier of the corresponding information, the data type of the corresponding information, the filing status of the corresponding information, and the filing requirements of the corresponding information; wherein the filing status includes an unfiled status and a filed status; Overall-level data: the importance of the corresponding information and the sensitivity of the corresponding information.

4. The method for independent application after fusion and archiving of multimodal data according to claim 3 is characterized in that: The specific process of obtaining the archiving score is as follows: Target requirements, retention period, document confidentiality, and openness are extracted from the extracted feature information and defined as key factors. Each key factor is assigned a corresponding weight and score, and the weight and score are kept synchronized. Target requirements: core files are assigned a weight of ε3 > important files are assigned a weight of ε2 > ordinary files are assigned a weight of ε1; Preservation period: long-term, weighted ε3> medium-term, weighted ε2> short-term, weighted ε1; Based on the target requirements and preservation period, multiply the score of each factor by its weight, and then add up all the results to get the importance level; Document confidentiality: Business secrets, assigned a weight of ε3 > personal privacy, assigned a weight of ε2 > other sensitive information, assigned a weight of ε1; Open form: secret approval, weighted ε3> internal disclosure, weighted ε2> public, weighted ε1; Based on the confidentiality and openness of the document, the score of each factor is multiplied by its weight, and then all the results are added together to get the sensitivity level; Then an archive score is generated and the archive partition is matched according to the archive score, based on the formula: ； In the formula, de represents the archiving score, zy represents the importance score, mg represents the sensitivity score, β1 and β2 are weight coefficients, and both β1 and β2 are greater than 0.

5. The method for independent application after fusion and archiving of multimodal data according to claim 3 is characterized in that: The steps to set up the archive interval include: The database is divided into M grid intervals and defined as archiving intervals. The storage information in the database is allocated to the M archiving intervals in sequence according to the archiving scores. The corresponding calling permissions are set for any archiving interval. N calling levels are set, and there are n calling ports under any calling level Ni. Different permission ranges are set according to the levels.

6. The method for independent application after fusion and archiving of multimodal data according to claim 1, characterized in that: The steps to provide business management services include: The user registers or logs in, and the user has a unique user ID; After the user logs in or registers successfully, the corresponding subject attributes are loaded according to the user, including employee, position and responsibilities; the permission description is configured according to the subject attributes, and the user ID is added to the user's permission description. The role permission is configured according to the permission description, and the files of the file group are managed. If the file group does not exist, a new file is added through the business management platform, and the administrator of the user group is set to the user ID in the current permission description.

7. The method for independent application after fusion and archiving of multimodal data according to claim 1, characterized in that: The steps of S3 include: Verify the identity and permissions of user groups and filter based on RABC permissions; The menu of the interface is displayed according to the user's permission information; the user group uses the archive search and analysis services on the corresponding interface to perform global search, advanced search and viewing of the archive group; By querying the chain network structure information set that meets the conditions through the database, data query is performed on the chain network structure that meets the requirements to obtain a business chain set, which includes a query vector, a key vector and a value vector; by constructing a three-dimensional coordinate system, and taking the query vector as the X-axis, the key vector as the Y-axis, and the value vector as the Z-axis, a vector feature space is established, a three-dimensional contour map is drawn on the vector feature space, and the three-dimensional contour map is superimposed with the preset standard three-dimensional contour map, and its horizontal and vertical screenshots are obtained and compared to generate the total overlapping area and the total non-overlapping area, wherein the total non-overlapping area is the offset during the search response; the evaluation factors in the search process are obtained according to the total overlapping area and the total non-overlapping area, and the quality of the business management service is determined; The formula for the evaluation factor is: ； In the formula, BP represents the evaluation factor, k0 represents the correction weight coefficient, cmj represents the overlapping area, and fmj represents the non-overlapping area; Obtain the node information of the business chain, which includes operation logs, archive status, and related archive packages, identify and collect the user ID corresponding to the operation node information, count the user's operation behavior, and generate statistical reports.

8. A system for independent application after fusion and archiving of multimodal data, using any method described in claims 1 to 7, characterized in that: The system includes: a business management platform, a data acquisition and fusion unit, an archiving management unit, a retrieval and analysis unit, and a storage and calling unit; A data acquisition and fusion unit, used to acquire input multimodal data and fuse the multimodal data to generate a target data set; wherein the input multimodal data type includes structured data, unstructured data and IoT data; The archiving management unit is used to transfer the target data set to the business management platform, design a chain network structure to start archiving management, generate an archiving package, and build a target decision network based on a deep learning network, calculate the archiving score, match the archiving partition according to the archiving score, and provide business management services; wherein the target decision network is obtained by reinforcement learning exploration of the sample environment, and the sample environment is composed of corresponding information in at least two historical archive packages; The retrieval and analysis unit is used to enter the business management platform for retrieval and service analysis, perform data query on the business chain, generate a business chain set, and generate evaluation factors based on the business chain set to evaluate the business management service; generate statistical reports based on the node information of the business chain; among them, service analysis includes statistical analysis, viewing details and generating reports; The storage and call unit is used to manage and control the user's access rights to archive storage and archives. Based on user identity and role, different permission levels and access control policies are defined to ensure that only authorized users can access and operate archives; The business management platform includes archive groups and user groups and is used to manage various operations and activities of archive groups and user groups.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images