A knowledge base management method, system, and medium based on data mining technology

Abstract

This application discloses a knowledge base management method, system, and medium based on data mining technology, mainly relating to the field of knowledge base management technology, to address the problems of difficulty in balancing privacy protection and data utility in existing solutions, and the risk of privacy data leakage during computation. The method includes: migrating encrypted numerical data from a privacy knowledge database to a visual database; obtaining the specific data and formulas involved in the computation through the corresponding display interface of the visual database; when encrypted numerical data exists in the specific data, inputting the encrypted numerical data and calculation formulas into a pre-trained preset neural learning algorithm to obtain first result data, and inputting unencrypted numerical data into a conventional formula to obtain second result data; and inputting the first result data and second result data into a merging result formula to obtain the final calculation result.

Description

Technical Field

[0001] This application relates to the field of knowledge base management technology, and in particular to a knowledge base management method, system and medium based on data mining technology. Background Technology

[0002] In existing technologies, knowledge base management systems typically employ traditional relational databases or NoSQL databases for data storage and management. For privacy-related data processing, the industry commonly uses techniques such as static data masking, field-level encryption, or role-based access control. For example, differentiated encryption is achieved through data classification and categorization, or homomorphic encryption is used to protect private values ​​during computation. In entity relationship mining, existing solutions largely rely on rule engines or BERT-based pre-trained models for entity recognition, such as constructing entity association networks using knowledge graphs. For numerical computation, current systems typically use predefined computation rules or fixed algorithm libraries to process associated data; some solutions introduce differential privacy techniques to reduce the risk of data leakage.

[0003] However, the existing technology has the following drawbacks: (1) It is difficult to balance privacy protection and data utility. Static encryption leads to the break of the computing link, while dynamic encryption has performance bottlenecks; (2) Privacy data needs to be directly called during the computing process, which poses a risk of leakage. Summary of the Invention

[0004] This application provides a knowledge base management method, system, and medium based on data mining technology to solve the problems of difficulty in balancing privacy protection and data utility in existing solutions, and the risk of privacy data leakage during the calculation process.

[0005] Firstly, this application provides a knowledge base management method based on data mining technology, the method comprising:

[0006] Privacy data is acquired by pre-setting privacy data sources and stored in a privacy knowledge database. The privacy data consists of entity data and privacy numerical data. The bidirectional long short-term memory network model in the natural language processing technology stack is used to obtain the relationships between entity data in the privacy knowledge database.

[0007] From the pre-defined relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data; use several types of encryption algorithms to randomly encrypt the privacy value data to obtain the encrypted value data; use the encrypted value data, calculation formula, and result data as training samples to train the pre-defined neural learning algorithm to obtain the trained pre-defined neural learning algorithm.

[0008] The encrypted numerical data in the privacy knowledge database is migrated to a visual database. The specific data and formulas used in the calculation are obtained through the corresponding interface of the visual database. When encrypted numerical data exists within the specific data, the specific formula is broken down into the calculation formula corresponding to the encrypted numerical data, the regular formula corresponding to the unencrypted numerical data, and the combined result formula of the calculation formula and the regular formula. The encrypted numerical data and calculation formula are input into a pre-trained neural learning algorithm to obtain the first result data. The unencrypted numerical data is input into the regular formula to obtain the second result data. The first result data and the second result data are input into the combined result formula to obtain the final calculation result.

[0009] In one implementation of this application, when performing calculations involving privacy data, after inputting the encrypted numerical data and the relevant calculation formula into a pre-trained preset neural learning algorithm to obtain the output data, the method further includes:

[0010] Real-time monitoring of the privacy knowledge database to check for newly added privacy data;

[0011] Obtain the relationship between entity data in newly added privacy data and other entity data in the privacy knowledge database;

[0012] From the preset relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data;

[0013] The newly added encrypted numerical data, calculation formulas, and newly added result data are used as training samples to update the preset neural learning algorithm and obtain the trained preset neural learning algorithm.

[0014] In one implementation of this application, several types of encryption algorithms are used to randomly encrypt private numerical data to obtain encrypted numerical data, specifically including:

[0015] Randomly input the privacy-preserving numerical data into any encryption algorithm to obtain the corresponding encrypted numerical data;

[0016] After obtaining the encrypted numerical data of all privacy data, a deduplication check is performed; if a duplicate is found, the duplicate encrypted numerical data is input into any encryption algorithm again to update the corresponding encrypted numerical data.

[0017] Until no duplicate encrypted numerical data exists.

[0018] In one implementation of this application, encrypted numerical data, calculation formulas, and result data are used as training samples to train a preset neural learning algorithm, thereby obtaining a trained preset neural learning algorithm, specifically including:

[0019] The encrypted numerical data, calculation formulas, and result data are used as training samples to train the preset neural learning algorithm.

[0020] Randomly select several sets of encrypted numerical data and calculation formulas, input them into a preset neural learning algorithm, and when the output data is correct, the training of the preset neural learning algorithm is considered complete.

[0021] In one implementation of this application, the specific data and formulas involved in the calculation are obtained through a display interface corresponding to the visualization database, specifically including:

[0022] The visualization interface displays visualized data from a visualization database; this visualization database is connected to several databases and is used to obtain data for database migration.

[0023] Select a specific formula using the built-in preset formula selection box in the display interface; the preset formula selection box is configured with several formulas.

[0024] Secondly, this application provides a knowledge base management system based on data mining technology, the system comprising:

[0025] The acquisition module is used to obtain privacy data through a preset privacy data source and store the privacy data in a privacy knowledge database. The privacy data consists of entity data and privacy numerical data. The module uses a bidirectional long short-term memory network model in the natural language processing technology stack to obtain the relationships between entity data in the privacy knowledge database.

[0026] The training module is used to determine the calculation formula for privacy numerical data corresponding to entity data with related relationships from the preset relationship-calculation formula mapping database; input the privacy numerical data into the calculation formula to obtain the result data; randomly encrypt the privacy numerical data using several types of encryption algorithms to obtain encrypted numerical data; and use the encrypted numerical data, calculation formula, and result data as training samples to train the preset neural learning algorithm to obtain the trained preset neural learning algorithm.

[0027] The output module is used to migrate encrypted numerical data from the privacy knowledge database to the visualization database. Through the display interface corresponding to the visualization database, the specific data and formulas involved in the calculation are obtained. When there is encrypted numerical data in the specific data, the specific formula is split into the calculation formula corresponding to the encrypted numerical data, the regular formula corresponding to the unencrypted numerical data, and the combined result formula of the calculation formula and the regular formula. The encrypted numerical data and the calculation formula are input into the pre-trained preset neural learning algorithm to obtain the first result data, and the unencrypted numerical data is input into the regular formula to obtain the second result data. The first result data and the second result data are input into the combined result formula to obtain the final calculation result.

[0028] In one implementation of this application, the training module includes a first training unit.

[0029] Used to monitor in real time whether new privacy data appears in the privacy knowledge database;

[0030] Obtain the relationship between entity data in newly added privacy data and other entity data in the privacy knowledge database;

[0031] From the preset relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data;

[0032] The newly added encrypted numerical data, calculation formulas, and newly added result data are used as training samples to update the preset neural learning algorithm and obtain the trained preset neural learning algorithm.

[0033] In one implementation of this application, the training module includes an encryption unit.

[0034] It is used to randomly input private numerical data into any encryption algorithm to obtain the corresponding encrypted numerical data;

[0035] After obtaining the encrypted numerical data of all privacy data, a deduplication check is performed; if a duplicate is found, the duplicate encrypted numerical data is input into any encryption algorithm again to update the corresponding encrypted numerical data.

[0036] Until no duplicate encrypted numerical data exists.

[0037] In one implementation of this application, the training module includes a second training unit.

[0038] It is used to train a preset neural learning algorithm by using encrypted numerical data, calculation formulas, and result data as training samples.

[0039] Randomly select several sets of encrypted numerical data and calculation formulas, input them into a preset neural learning algorithm, and when the output data is correct, the training of the preset neural learning algorithm is considered complete.

[0040] Thirdly, this application provides a non-volatile computer storage medium storing computer instructions thereon, which, when executed, implement a knowledge base management method based on data mining technology as described above.

[0041] As can be seen from the above technical solutions, this application has the following advantages:

[0042] 1. Optimizing the balance between privacy protection and data utility:

[0043] A multi-type encryption algorithm random encryption mechanism is employed to avoid computational interruptions caused by single static encryption. By dynamically selecting the encryption method, computational continuity is maintained while ensuring data security.

[0044] By establishing a mapping relationship between encrypted data and calculation results through neural learning algorithms, the trained model can directly output calculation results based on encrypted data, achieving accurate calculation under encrypted conditions and solving the problem of "encryption is invalid".

[0045] 2. Prevention and control of privacy data leakage risks:

[0046] A hierarchical data storage architecture is constructed, with raw privacy data permanently isolated in a privacy knowledge database, and the computation process only calling encrypted numerical data. A neural agent computation module is developed to transform computations involving privacy data into a black-box operation of "encrypted input - result output". A mapping library of relationships and computation formulas is established to ensure the controllability of the neural algorithm's output results through predefined computation rules. Attached Figure Description

[0047] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a knowledge base management method based on data mining technology provided in an embodiment of this application.

[0049] Figure 2 This is a schematic diagram of the internal structure of a knowledge base management system based on data mining technology provided in an embodiment of this application. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.

[0052] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0053] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0054] The embodiment provides a knowledge base management method based on data mining technology, such as Figure 1 As shown in the embodiments of this application, the method mainly includes the following steps:

[0055] Step 110: Obtain privacy data through a preset privacy data source and store the privacy data in a privacy knowledge database; wherein, the privacy data consists of entity data and privacy numerical data; use the bidirectional long short-term memory network model in the natural language processing technology stack to obtain the relationship between entity data in the privacy knowledge database.

[0056] The privacy data consists of entity data and privacy numerical data, and there is a corresponding relationship between the entity data and the privacy numerical data.

[0057] In some embodiments, after performing entity identification on non-privacy data containing entity data in a privacy knowledge database to obtain the relationships between entities, the method further includes:

[0058] Generate a graph showing the relationships between entities;

[0059] The entity relationship graph is imported into the Neo4j graph database for persistent storage. An inverted index is built in Elasticsearch to support full-text search, and a timestamp is attached to each entity node to enable version tracking.

[0060] As will be understood by those skilled in the art, secure data isolation storage achieves physical isolation between private and non-private data by acquiring data from a pre-set privacy data source and storing it in an independent privacy knowledge database, thus cutting off unauthorized access paths at the source.

[0061] Intelligent Relationship Recognition: This step can use the BiLSTM model to perform deep semantic analysis on non-privacy entity data, which can automatically construct an entity association graph and establish a relationship foundation for subsequent calculations.

[0062] Step 120: From the preset association-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with an association relationship; input the privacy value data into the calculation formula to obtain the result data; use several types of encryption algorithms to randomly encrypt the privacy value data to obtain the encrypted value data; use the encrypted value data, calculation formula, and result data as training samples to train the preset neural learning algorithm to obtain the trained preset neural learning algorithm.

[0063] It should be noted that the preset relationship-calculation formula mapping database is a formulaic calculation framework. Through the predefined relationship-calculation formula mapping library, it ensures that each calculation request strictly follows preset rules. The entire process fully preserves the data calculation utility while protecting the original privacy data.

[0064] In the calculation involving privacy data, after inputting the encrypted numerical data and calculation formulas into a pre-trained neural learning algorithm to obtain the output data, the method further includes:

[0065] Real-time monitoring of the privacy knowledge database to check for newly added privacy data;

[0066] Obtain the relationship between entity data in newly added privacy data and other entity data in the privacy knowledge database;

[0067] From the preset relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data;

[0068] The newly added encrypted numerical data, calculation formulas, and newly added result data are used as training samples to update the preset neural learning algorithm and obtain the trained preset neural learning algorithm.

[0069] Those skilled in the art will understand that this step can monitor newly added data in the database in real time, automatically identify the relationships between newly added entities, and obtain the corresponding calculation formulas through a preset mapping library. The newly added data is then incorporated into the training samples to continuously optimize the neural learning algorithm, maintaining the model's ability to process new data.

[0070] This involves employing several types of encryption algorithms to randomly encrypt private numerical data, resulting in encrypted numerical data, specifically including:

[0071] Randomly input the privacy-preserving numerical data into any encryption algorithm to obtain the corresponding encrypted numerical data;

[0072] After obtaining the encrypted numerical data of all privacy data, a deduplication check is performed; if a duplicate is found, the duplicate encrypted numerical data is input into any encryption algorithm again to update the corresponding encrypted numerical data.

[0073] Until no duplicate encrypted numerical data exists.

[0074] Specifically, encrypted numerical data, calculation formulas, and result data are used as training samples to train a pre-defined neural learning algorithm, resulting in a well-trained pre-defined neural learning algorithm. This process includes:

[0075] The encrypted numerical data, calculation formulas, and result data are used as training samples to train the preset neural learning algorithm.

[0076] Randomly select several sets of encrypted numerical data and calculation formulas, input them into a preset neural learning algorithm, and when the output data is correct, the training of the preset neural learning algorithm is considered complete.

[0077] Those skilled in the art will understand that this step can encrypt the privacy-sensitive numerical data by randomly selecting multiple encryption algorithms, ensuring that the same original data will yield different encryption results at different times. This method ensures the uniqueness of the encrypted numerical data through a deduplication mechanism, effectively preventing the possibility of retrieving the original data from the encrypted data.

[0078] This step utilizes encrypted numerical data, calculation formulas, and calculation results as training samples to construct a neural learning algorithm capable of directly processing encrypted data. The algorithm undergoes validation testing to ensure the accuracy of the output, achieving reliable computation without exposing the original data.

[0079] This step involves only encrypted data and calculation formulas throughout the entire calculation process; the original privacy data remains encrypted at all times. The calculation result is directly output through a neural learning algorithm, completely avoiding the plaintext exposure of privacy data during the calculation process.

[0080] Step 130: Migrate the encrypted numerical data from the privacy knowledge database to the visualization database; obtain the specific data and formulas involved in the calculation through the display interface of the visualization database; when there is encrypted numerical data in the specific data, split the specific formula into the calculation formula corresponding to the encrypted numerical data, the regular formula corresponding to the unencrypted numerical data, and the combined result formula of the calculation formula and the regular formula; input the encrypted numerical data and the calculation formula into the pre-trained preset neural learning algorithm to obtain the first result data, input the unencrypted numerical data into the regular formula to obtain the second result data; input the first result data and the second result data into the combined result formula to obtain the final calculation result.

[0081] Specifically, the specific data and formulas used in the calculation are obtained through the display interface corresponding to the visualization database. This can be achieved by: displaying the visualization data in the visualization database through the display interface; the visualization database being connected to several databases to obtain data for database migration; and selecting a specific formula through the preset formula selection box built into the display interface; the preset formula selection box is configured with several formulas.

[0082] Those skilled in the art will understand that this step achieves physical isolation of sensitive information by migrating the encrypted numerical data to a visualization database. The original privacy data remains in the privacy knowledge database, while only the encrypted numerical data enters the visualization environment, effectively preventing data leakage.

[0083] This step allows computational tasks involving privacy-sensitive data to be performed within a visualized environment. By inputting encrypted numerical data and calculation formulas into a trained neural learning algorithm, the computational results are directly obtained, ensuring the integrity of the computational function while maintaining the data encryption state.

[0084] Furthermore, the method of breaking down a specific formula into a calculation formula corresponding to encrypted numerical data, a regular formula corresponding to unencrypted numerical data, and a formula resulting from the combination of the calculation formula and the regular formula can be implemented by existing AI intelligent algorithms, which will not be elaborated upon in this application.

[0085] As described above, this embodiment employs a multi-type encryption algorithm random encryption mechanism to avoid computational interruptions caused by single static encryption. By dynamically selecting the encryption method, computational continuity is maintained while ensuring data security. A neural learning algorithm establishes a mapping relationship between encrypted data and computation results. The trained model can directly output computation results based on encrypted data, achieving accurate computation under encryption conditions and solving the problem of "encryption equals invalidation."

[0086] This embodiment constructs a hierarchical data storage architecture, permanently isolating the original privacy data in a privacy knowledge database, and only calling encrypted numerical data during the calculation process. A neural agent computation module is developed to transform calculations involving privacy data into a black-box operation of "encrypted input - result output". A mapping library of association relationships and calculation formulas is established to ensure the controllability of the neural algorithm's output results through predefined calculation rules.

[0087] In addition, this application Figure 2 This application provides a knowledge base management system based on data mining technology. For example... Figure 2 As shown in the embodiments of this application, the system mainly includes:

[0088] The module 210 is used to obtain privacy data through a preset privacy data source and store the privacy data in a privacy knowledge database. The privacy data consists of entity data and privacy numerical data. The module uses a bidirectional long short-term memory network model in the natural language processing technology stack to obtain the relationship between entity data in the privacy knowledge database.

[0089] Those skilled in the art will understand that the acquisition module 210 obtains data through a preset privacy data source and stores it in an independent privacy knowledge database, achieving physical isolation between privacy data and non-privacy data and cutting off unauthorized access paths at the source. The acquisition module 210 can use a BiLSTM model to perform deep semantic analysis on non-privacy entity data and can automatically construct an entity association graph, establishing a relational foundation for subsequent calculations.

[0090] The training module 220 is used to determine the calculation formula for the privacy numerical data corresponding to the entity data with the relationship from the preset association-calculation formula mapping database; input the privacy numerical data into the calculation formula to obtain the result data; randomly encrypt the privacy numerical data using several types of encryption algorithms to obtain the encrypted numerical data; and use the encrypted numerical data, calculation formula, and result data as training samples to train the preset neural learning algorithm to obtain the trained preset neural learning algorithm.

[0091] Training module 220 includes a first training unit.

[0092] This tool is used to monitor in real time whether new privacy data appears in the privacy knowledge database; to obtain the relationship between entity data in the new privacy data and other entity data in the privacy knowledge database; to determine the calculation formula for the privacy numerical data corresponding to the entity data with the relationship from the preset relationship-calculation formula mapping database; to input the privacy numerical data into the calculation formula and obtain the result data; and to use the new encrypted numerical data, the calculation formula, and the new result data as training samples to update and train the preset neural learning algorithm, thereby obtaining the trained preset neural learning algorithm.

[0093] Training module 220 includes an encryption unit.

[0094] It is used to randomly input private numerical data into any encryption algorithm to obtain the corresponding encrypted numerical data;

[0095] After obtaining the encrypted numerical data of all privacy data, a deduplication check is performed; if a duplicate is found, the duplicate encrypted numerical data is input into any encryption algorithm again to update the corresponding encrypted numerical data.

[0096] Until no duplicate encrypted numerical data exists.

[0097] Training module 220 includes a second training unit.

[0098] It is used to train a preset neural learning algorithm by using encrypted numerical data, calculation formulas, and result data as training samples.

[0099] Randomly select several sets of encrypted numerical data and calculation formulas, input them into a preset neural learning algorithm, and when the output data is correct, the training of the preset neural learning algorithm is considered complete.

[0100] Those skilled in the art will understand that the training module 220 can encrypt privacy-related numerical data by randomly selecting multiple encryption algorithms, ensuring that the same original data yields different encryption results at different times. This method ensures the uniqueness of encrypted numerical data through a deduplication mechanism, effectively preventing the possibility of retrieving the original data from the encrypted data.

[0101] Training module 220 can utilize encrypted numerical data, calculation formulas, and calculation results as training samples to construct a neural learning algorithm capable of directly processing encrypted data. This algorithm ensures the accuracy of its output through validation testing, achieving reliable computation without exposing the original data.

[0102] The entire computation process of training module 220 involves only encrypted data and calculation formulas, while the original privacy data remains encrypted throughout. The computation results are directly output through a neural learning algorithm, completely avoiding the plaintext exposure of privacy data during the computation process.

[0103] Output module 230 is used to migrate encrypted numerical data from the privacy knowledge database to the visualization database; through the display interface corresponding to the visualization database, it obtains the specific data and specific formulas involved in the calculation; when there is encrypted numerical data in the specific data, it splits the specific formula into the calculation formula corresponding to the encrypted numerical data, the regular formula corresponding to the unencrypted numerical data, and the combined result formula of the calculation formula and the regular formula; it inputs the encrypted numerical data and the calculation formula into a pre-trained preset neural learning algorithm to obtain the first result data, inputs the unencrypted numerical data into the regular formula to obtain the second result data; and inputs the first result data and the second result data into the combined result formula to obtain the final calculation result.

[0104] Those skilled in the art will understand that the output module 230 achieves physical isolation of sensitive information by migrating non-privacy data, entity data, and encrypted numerical data to the visualization database. The original privacy data remains in the privacy knowledge database, while only the encrypted numerical data enters the visualization environment, effectively preventing data leakage.

[0105] The output module 230 can still perform computational tasks involving privacy-sensitive data in a visual environment. By inputting encrypted numerical data and calculation formulas into a trained neural learning algorithm, it directly obtains the calculation results, ensuring the integrity of the computational function while maintaining the data encryption state.

[0106] The output module 230 separates three types of information: non-privacy data, entity data, and encrypted numerical data. This categorized storage mechanism not only meets the needs of data visualization but also ensures the security of different types of data through differentiated processing.

[0107] In addition, this application embodiment also provides a non-volatile computer storage medium storing executable instructions thereon, which, when executed, implements the knowledge base management method based on data mining technology as described above.

[0108] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily 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 invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A knowledge base management method based on data mining technology, characterized in that, The method includes: Privacy data is acquired by pre-setting privacy data sources and stored in a privacy knowledge database. The privacy data consists of entity data and privacy numerical data. The bidirectional long short-term memory network model in the natural language processing technology stack is used to obtain the relationships between entity data in the privacy knowledge database. From the pre-defined relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data; use several types of encryption algorithms to randomly encrypt the privacy value data to obtain the encrypted value data; use the encrypted value data, calculation formula, and result data as training samples to train the pre-defined neural learning algorithm to obtain the trained pre-defined neural learning algorithm. The encrypted numerical data in the privacy knowledge database is migrated to a visual database. The specific data and formulas used in the calculation are obtained through the corresponding interface of the visual database. When encrypted numerical data exists within the specific data, the formulas are broken down into calculation formulas corresponding to the encrypted numerical data, regular formulas corresponding to the unencrypted numerical data, and a combined formula of the two. The encrypted numerical data and calculation formulas are input into a pre-trained neural learning algorithm to obtain the first result data. The unencrypted numerical data is input into the regular formula to obtain the second result data. The first and second result data are then input into the combined formula to obtain the final calculation result. When performing calculations involving privacy data, after inputting the encrypted numerical data and the relevant calculation formulas into a pre-trained preset neural learning algorithm to obtain the output data, the method further includes: Real-time monitoring of the privacy knowledge database to check for newly added privacy data; Obtain the relationship between entity data in newly added privacy data and other entity data in the privacy knowledge database; From the preset relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data; The newly added encrypted numerical data, calculation formulas, and newly added result data are used as training samples to update the training preset neural learning algorithm and obtain the trained preset neural learning algorithm. Several types of encryption algorithms are used to randomly encrypt private numerical data to obtain encrypted numerical data, specifically including: Randomly input the privacy-preserving numerical data into any encryption algorithm to obtain the corresponding encrypted numerical data; After obtaining the encrypted numerical data of all privacy data, a deduplication check is performed; if a duplicate is found, the duplicate encrypted numerical data is input into any encryption algorithm again to update the corresponding encrypted numerical data. Until no duplicate encrypted numerical data exists.

2. The knowledge base management method based on data mining technology according to claim 1, characterized in that, Using encrypted numerical data, calculation formulas, and result data as training samples, a pre-defined neural learning algorithm is trained to obtain the trained pre-defined neural learning algorithm, specifically including: The encrypted numerical data, calculation formulas, and result data are used as training samples to train the preset neural learning algorithm. Randomly select several sets of encrypted numerical data and calculation formulas, input them into a preset neural learning algorithm, and when the output data is correct, the training of the preset neural learning algorithm is considered complete.

3. The knowledge base management method based on data mining technology according to claim 1, characterized in that, The specific data and formulas used in the calculations can be obtained through the visualization interface corresponding to the database, including: The visualization interface displays visualized data from a visualization database; this visualization database is connected to several databases and is used to obtain data for database migration. Select a specific formula using the built-in preset formula selection box in the display interface; the preset formula selection box contains several formulas.

4. A knowledge base management system based on data mining technology, characterized in that, The system includes: The acquisition module is used to obtain privacy data through a preset privacy data source and store the privacy data in a privacy knowledge database. The privacy data consists of entity data and privacy numerical data. The module uses a bidirectional long short-term memory network model in the natural language processing technology stack to obtain the relationships between entity data in the privacy knowledge database. The training module is used to determine the calculation formula for privacy numerical data corresponding to entity data with related relationships from the preset relationship-calculation formula mapping database; input the privacy numerical data into the calculation formula to obtain the result data; randomly encrypt the privacy numerical data using several types of encryption algorithms to obtain encrypted numerical data; and use the encrypted numerical data, calculation formula, and result data as training samples to train the preset neural learning algorithm to obtain the trained preset neural learning algorithm. The output module is used to migrate encrypted numerical data from the privacy knowledge database to the visualization database; through the display interface corresponding to the visualization database, the specific data and formulas involved in the calculation are obtained; when there is encrypted numerical data in the specific data, the specific formula is split into the calculation formula corresponding to the encrypted numerical data, the regular formula corresponding to the unencrypted numerical data, and the combined result formula of the calculation formula and the regular formula; the encrypted numerical data and the calculation formula are input into the pre-trained preset neural learning algorithm to obtain the first result data, and the unencrypted numerical data is input into the regular formula to obtain the second result data; the first result data and the second result data are input into the combined result formula to obtain the final calculation result. The training module includes the first training unit. Used to monitor in real time whether new privacy data appears in the privacy knowledge database; Obtain the relationship between entity data in newly added privacy data and other entity data in the privacy knowledge database; From the preset relationship-calculation formula mapping database, determine the calculation formula for the privacy value data corresponding to the entity data with the relationship; input the privacy value data into the calculation formula to obtain the result data; The newly added encrypted numerical data, calculation formulas, and newly added result data are used as training samples to update the training preset neural learning algorithm and obtain the trained preset neural learning algorithm. The training module includes an encryption unit. It is used to randomly input private numerical data into any encryption algorithm to obtain the corresponding encrypted numerical data; After obtaining the encrypted numerical data of all privacy data, a deduplication check is performed; if a duplicate is found, the duplicate encrypted numerical data is input into any encryption algorithm again to update the corresponding encrypted numerical data. Until no duplicate encrypted numerical data exists.

5. The knowledge base management system based on data mining technology according to claim 4, characterized in that, The training module includes a second training unit. It is used to train a preset neural learning algorithm by using encrypted numerical data, calculation formulas, and result data as training samples. Randomly select several sets of encrypted numerical data and calculation formulas, input them into a preset neural learning algorithm, and when the output data is correct, the training of the preset neural learning algorithm is considered complete.

6. A non-volatile computer storage medium, characterized in that, It stores computer instructions, which, when executed, implement a knowledge base management method based on data mining technology as described in any one of claims 1-3.

Images