Application of privacy matrix factorization on large language model-based medical recommendation services

Abstract

The invention discloses application of privacy matrix decomposition on medical recommendation service based on a large language model. The system comprises a data acquisition module; a privacy matrix decomposition module; a large language model processing module; a recommendation decision module; the user interface module breaks through the restrictions of regions, time and professional resources through personalized online medical recommendation services, can obtain high-quality medical consultation and health recommendation services for social vulnerable groups such as remote areas, rural areas, old people and patients with mobility difficulties, promotes fair distribution of medical resources, enhances user trust, and improves the user experience. Particularly, due to the application of a privacy matrix decomposition technology, worries of a user on personal health data safety are eliminated, the trust degree of the user on a medical platform is improved, health consultation service popularization is promoted, the overall health management consciousness and the self-health-care ability of the society are improved, medical knowledge and health management concepts are popularized by means of a convenient online platform, and the health management system has a wide application prospect. Misdiagnosis and treatment caused by asymmetry of medical information are reduced, and the health quality of the whole society is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical recommendation services based on large language models. Specifically, it involves the application of privacy matrix factorization in medical recommendation services based on large language models. Background Art

[0002] In the field of medical recommendation services, with the rapid development of information technology, it has become a trend to achieve personalized recommendations using big data analysis and artificial intelligence algorithms. However, there are still many problems to be solved in this field.

[0003] In terms of privacy protection, although existing technologies such as data encryption, de-identification, and anonymization protect users' personal information and sensitive data to a certain extent, in the context of medical recommendation scenarios, how to ensure data privacy and security while providing accurate personalized services remains a difficult problem. As an emerging data protection method, privacy matrix factorization technology has been relatively maturely applied in fields such as e-commerce and social networks, but its application in medical recommendation services is still in its infancy. Existing privacy protection schemes rarely effectively combine data privacy protection with personalized recommendation systems, making it difficult to balance the dual requirements of privacy protection and accurate recommendation.

[0004] Traditional matrix factorization methods have technical bottlenecks in ensuring the security and privacy of sensitive data when dealing with large-scale data sets. In comprehensive medical recommendation systems, the contradiction between personalized recommendation and data privacy protection is prominent. Most systems use machine learning and big data analysis technologies to provide personalized services, but when dealing with sensitive medical information, they cannot effectively protect user privacy and do not fully utilize privacy matrix factorization technology to balance recommendation accuracy and privacy protection, resulting in difficulties in improving the accuracy and data security of personalized recommendation services. Summary of the Invention

[0005] In view of the problems in the related art, the present invention proposes to apply privacy matrix factorization in medical recommendation services based on large language models. By combining privacy matrix factorization technology, on the premise of ensuring user privacy and security, it provides users with efficient and accurate personalized medical recommendation services, solving the deficiencies of existing medical recommendation systems in terms of privacy protection and recommendation accuracy.

[0006] System Architecture: The medical recommendation service system of the present invention mainly consists of a data collection module, a privacy matrix factorization module, a large language model processing module, a recommendation decision module, and a user interface module.

[0007] Data Collection Module: Responsible for collecting health-related data from the user side, covering symptom descriptions, medical histories, living habits, etc.; at the same time, collecting medical resource-related data, such as hospital department settings, expert information, surgical features, medical insurance policies, etc. This module provides a comprehensive data basis for subsequent recommendation services.

[0008] Privacy Matrix Decomposition Module: Matrix-represents the user's health data and medical resource data, and uses privacy matrix decomposition technology to encrypt the data. During the decomposition process, sensitive information in the matrix is hidden, and encryption algorithms are used to prevent the leakage of user data, ensuring data privacy and providing secure and reliable data support for subsequent modules.

[0009] Large Language Model Processing Module: Adopts a large language model based on natural language processing (NLP) to parse the text information input by the user, extract potential medical needs, and combine the data output by the privacy matrix decomposition module for personalized recommendation. The powerful natural language understanding ability of the large language model can accurately grasp the user's intention and lay a foundation for personalized recommendation.

[0010] Recommendation Decision Module: Based on the output of the large language model and the results of privacy matrix decomposition, comprehensively consider multi-dimensional information such as the user's economic ability, medical insurance reimbursement policy, and social welfare, and generate personalized medical plan recommendations for the user. The recommended content includes suitable hospitals, departments, experts, and corresponding treatment plans.

[0011] User Interface Module: Through the interface design of the mobile and PC sides, realizes the convenient interaction between the user and the system. The user can obtain medical recommendations by inputting information through simple text descriptions. The system adjusts and optimizes the recommendation results and algorithms according to the user's feedback to improve the user experience.

[0012] Data Processing Method: Step 1: User Data Input and Preprocessing: The user inputs health-related data such as symptoms and medical history into the system, and the system preprocesses the input data, including operations such as data cleaning and format conversion, to make it meet the requirements of subsequent processing.

[0013] Step 2: Privacy Matrix Decomposition and Encryption: Matrix-process the preprocessed data and encrypt it using privacy matrix decomposition technology. This technology decomposes the user data matrix through a specific algorithm, while protecting the user's sensitive information, mining the user's potential needs and preferences.

[0014] Step 3: Large Language Model Processing and Matching: Use the large language model to perform natural language understanding on the user's input symptom description and health status, match the extracted information with the processed medical resource data, and generate preliminary medical recommendations.

[0015] Step 4: Personalized Recommendation Generation: Combine multi-dimensional information such as the user's economic situation and medical insurance policy to optimize the preliminary recommendation and generate personalized medical recommendations that meet the actual needs of the user.

[0016] Step 5: Recommendation result output and feedback optimization: The medical recommendation results are output through a simple user interface, and users operate according to the recommendations and provide feedback. The system optimizes the recommendation algorithm and privacy matrix decomposition technology based on the feedback to improve the accuracy and personalization of the recommendation service.

[0017] Privacy matrix processing steps: Define the scope of data collection: clarify all data types that need to be collected for the project and how to process them to ensure the comprehensiveness and standardization of data collection.

[0018] Design privacy protection measures: Configure corresponding privacy protection strategies for each data type, and use encryption, anonymization and other technical means to ensure data privacy.

[0019] System implementation and monitoring: Implement technical solutions for data collection, storage, use and sharing, and conduct regular security audits and privacy monitoring to ensure the security and compliance of system operations.

[0020] Technology fusion innovation: pioneeringly combining large language models with privacy matrix decomposition technology. Existing medical recommendation systems are difficult to balance privacy protection and accurate recommendations. This invention uses the natural language processing capabilities of large language models and the data protection advantages of privacy matrix decomposition technology to achieve accurate recommendations without leaking user privacy. Beneficial effects of the present invention Social impact: Improve the fairness of medical services: Through personalized online medical recommendation services, we can break through geographical, time and professional resource limitations. Socially vulnerable groups such as remote areas, rural areas, the elderly, and patients with limited mobility can obtain high-quality medical consultation and health recommendation services, thereby promoting fair distribution of medical resources.

[0021] Enhance user trust: Privacy protection mechanisms, especially the application of privacy matrix decomposition technology, eliminate users' concerns about the security of personal health data, enhance users' trust in medical platforms, promote the popularization of health consulting services, and improve the overall health management awareness and self-care ability of the society.

[0022] Popularize the concept of health management: With the help of convenient online platforms, popularize medical knowledge and health management concepts, reduce misdiagnosis and wrong treatment caused by asymmetric medical information, and improve the health literacy of the whole society.

[0023] Technical effect: Improve recommendation accuracy: Privacy matrix decomposition technology enables medical recommendation services to provide accurate and personalized recommendations based on user health data. Encrypted matrix processing of user health records and other data enables differentiated recommendations, improving recommendation accuracy and user satisfaction.

[0024] Enhanced Privacy Protection: Compared with traditional data analysis techniques, privacy matrix factorization significantly enhances the privacy protection capabilities of medical data. The encrypted data is used for calculations and recommendations. Even if the system is attacked, the actual health information of users cannot be restored, ensuring the privacy and security of users and the compliance of the platform.

[0025] Improved Data Processing Efficiency: The privacy matrix factorization technology enables the recommendation system to operate efficiently with large-scale user data and massive health records. While ensuring privacy protection, it maintains high computational performance and fast response speed, providing high-quality personalized medical services for a large number of users.

[0026] Innovative Data Processing Method: Privacy matrix factorization combines data encryption and analysis methods to perform machine learning and data mining while protecting data privacy. It processes large-scale user data and executes complex analysis tasks such as disease prediction and personalized health guidance, providing a reference for data processing in other industries.

[0027] Promote In-depth Technology Application: The application of privacy matrix factorization technology in the medical field promotes the in-depth integration of big data analysis and artificial intelligence technologies in medical services. From traditional manual diagnosis to intelligent medical recommendations, and from single data point analysis to multi-dimensional data comprehensive analysis, it provides new application scenarios and practical platforms for big data and artificial intelligence technologies. Description of the Drawings

[0028] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a schematic flow diagram of the application of privacy matrix factorization in the medical recommendation service based on the large language model of the present invention; Figure 2 is a tabular diagram of the application of privacy matrix factorization in the medical recommendation service based on the large language model of the present invention; Detailed Embodiments

[0029] Please refer to Figure 1-2 as shown, which is an embodiment of the present invention.

[0030] Apply privacy matrix factorization in the medical recommendation service of the large language model.

[0031] System Setup and Deployment: According to the system architecture design, develop and deploy a data collection module, a privacy matrix factorization module, a large language model processing module, a recommendation decision module, and a user interface module. Ensure that each module operates stably on a hardware server or a cloud computing platform, and configure the corresponding database to store user data and medical resource data.

[0032] Data collection and preprocessing: The data collection module continuously collects user health data and medical resource data, and preprocesses the collected data regularly. Data cleaning removes duplicate and incorrect data, and format conversion unifies the data into a format convenient for processing, providing high-quality data for subsequent analysis and recommendation.

[0033] Privacy matrix factorization and encryption processing: The privacy matrix factorization module performs matrix representation and encryption processing on the preprocessed data according to the defined data collection scope and privacy protection measures. The encryption algorithm and privacy protection strategy are updated regularly to ensure data security.

[0034] Large language model training and optimization: The large language model processing module regularly collects a large amount of medical text data, including medical literature, medical records, etc., and trains and optimizes the large language model. It improves the model's ability to understand and process natural language in the medical field, and enhances the accuracy of personalized recommendation.

[0035] Recommendation decision-making and service provision: The recommendation decision-making module generates personalized medical recommendations based on the output of the large language model and the result of privacy matrix factorization, combined with multi-dimensional information such as the user's economic ability and medical insurance policy obtained in real time. It is pushed to the user in a timely manner through the user interface module, and the user usage and feedback data are recorded.

[0036] System optimization and iteration: According to user feedback and system operation data, the recommendation algorithm and privacy matrix factorization technology are optimized regularly. The model parameters are adjusted, and the data processing process is improved to continuously enhance the accuracy, personalization level, and privacy protection level of the medical recommendation service.

[0037] Personalized recommendation optimization: In the present invention, the privacy matrix factorization technology performs matrix representation and encryption processing on the user's medical needs and information such as hospitals, doctors, and departments, providing secure data input for the large language model to achieve personalized recommendation. Compared with the traditional matrix factorization method, a privacy protection module is added to make up for the defects of the traditional method in privacy protection.

[0038] Multi-dimensional recommendation mode: The medical recommendation service comprehensively considers multiple dimensions such as the user's basic condition, economic ability, medical insurance reimbursement policy, and social welfare. With the natural language processing ability of the large language model, the system customizes the optimal medical plan according to the multi-faceted information input by the user, which is not available in the prior art.

[0039] Multi-level privacy protection: The privacy matrix factorization technology is used to protect the privacy of user health data. At the same time, combined with technical means such as data encryption and de-identification, multi-level protection of user data is carried out. When sharing data with medical institutions, insurance companies, etc., the user information security is strictly ensured, in line with the national personal privacy protection laws and regulations.

[0040] 1. User Input Layer: Health data provided by users, such as symptoms, signs, past medical history, lifestyle, etc., which will be used for subsequent recommendation generation.

[0041] 2. Data Encryption and Privacy Protection Module: Adopting privacy matrix factorization technology, encrypt and decompose user data. Privacy matrix factorization can transform user health data into an encrypted matrix to ensure that sensitive information is not leaked. While protecting user privacy in this step, it also enables the data to enter the next step of calculation and analysis.

[0042] 3. Data Processing and Recommendation Algorithm Module: Using the encrypted data after privacy matrix factorization, analyze the data through technologies such as deep learning and machine learning, and generate personalized medical recommendations. Including health prediction, disease diagnosis, drug recommendation, etc. This module will provide accurate medical advice through algorithms.

[0043] 4. Medical Recommendation Service Layer (Output Layer): According to the encrypted calculation results and the output of the AI model, the system generates health advice, disease prevention or treatment plans suitable for users, or provides necessary drug advice, etc.

[0044] User Feedback and Data Update Layer: Users provide feedback based on the received medical recommendations, and the platform can optimize the recommendation algorithm and privacy matrix factorization technology according to the feedback to further improve the accuracy and personalization of the recommendation service. Alternative Solutions: Recommendation System Based on Traditional Matrix Factorization: Traditional matrix factorization technology can be used to build a recommendation system, but there is a risk of user data privacy leakage and it cannot effectively protect user sensitive information. The privacy matrix factorization technology of the present invention improves the data privacy protection ability.

[0045] Recommendation System Based on Encryption Algorithm: Using encryption algorithms to encrypt user data can protect data privacy, but it may lead to slow data processing speed and affect the system response efficiency. The privacy matrix factorization technology of the present invention optimizes the processing speed while ensuring data encryption, providing an efficient personalized recommendation service.

[0046] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. Apply privacy matrix decomposition to medical recommendation service based on large language model, characterized by: include: The data collection module is used to collect health-related data from the user, including symptom descriptions, medical history, living habits, etc. It also collects data related to medical resources, such as hospital department settings, expert information, surgical features, and medical insurance policies, to provide a comprehensive data foundation for subsequent recommendation services; The privacy matrix decomposition module is used to represent the user's health data and medical resource data in a matrix form, and encrypt the data using the privacy matrix decomposition technology. During the decomposition process, sensitive information in the matrix is ​​hidden, and the encryption algorithm is used to prevent user data leakage, ensure data privacy, and provide safe and reliable data support for subsequent modules; The large language model processing module uses a large language model based on natural language processing (NLP) to parse the text information input by the user, extract potential medical needs, and combine the data output by the privacy matrix decomposition module to make personalized recommendations; The recommendation decision module generates personalized medical plan recommendations for users based on the output of the large language model and the results of the privacy matrix decomposition, taking into account the user's economic ability, medical insurance reimbursement policy, social welfare and other multi-dimensional information. The recommendations include appropriate hospitals, departments, experts and corresponding treatment plans; The user interface module enables convenient interaction between users and the system through mobile and PC interface design. Users can enter information through simple text descriptions to obtain medical recommendations. The system adjusts and optimizes the recommendation results and algorithms based on user feedback to enhance user experience.

2. A medical recommendation service method based on a large language model combined with privacy matrix decomposition, characterized in that: The following steps are involved: User data input and preprocessing: Users input health-related data into the system, such as symptoms, medical history, etc. The system preprocesses the input data, including data cleaning, format conversion and other operations, to make it meet the requirements of subsequent processing; Privacy matrix decomposition and encryption: Matrix the pre-processed data and encrypt it using privacy matrix decomposition technology. This technology decomposes the user data matrix through a specific algorithm, while protecting the user's sensitive information, it also explores the user's potential needs and preferences; Large language model processing and matching: Use the large language model to understand the natural language of the symptom description and health status entered by the user, match the extracted information with the processed medical resource data, and generate preliminary medical recommendations; Personalized recommendation generation: Based on multi-dimensional information such as the user's economic status and medical insurance policy, the preliminary recommendations are optimized to generate personalized medical recommendations that meet the user's actual needs; Recommendation result output and feedback optimization: Medical recommendation results are output through a simple user interface. Users perform operations based on the recommendations and provide feedback. The system optimizes the recommendation algorithm and privacy matrix decomposition technology based on the feedback to improve the accuracy and personalization of the recommendation service.

3. The medical recommendation service system based on a large language model combined with privacy matrix decomposition according to claim 1 is characterized in that: When the privacy matrix decomposition module encrypts the data, the privacy matrix decomposition technology encrypts the feature matrix of the user data to ensure user privacy while improving the recommendation accuracy.

4. The medical recommendation service system based on a large language model combined with privacy matrix decomposition according to claim 1, characterized in that: The large language model in the large language model processing module is used to process natural language information input by the user, thereby generating personalized medical recommendations.

5. The medical recommendation service system based on a large language model combined with privacy matrix decomposition according to claim 1 is characterized in that: The recommendation decision module combines multi-dimensional data such as medical resource information and medical insurance policies to provide more personalized and accurate medical recommendation services.

6. The medical recommendation service method based on a large language model combined with privacy matrix decomposition according to claim 2, characterized in that: In the privacy matrix decomposition and encryption step, the privacy matrix decomposition technology encrypts the feature matrix of user data to ensure user privacy while improving recommendation accuracy.

7. The medical recommendation service method based on a large language model combined with privacy matrix decomposition according to claim 2, characterized in that: In the large language model processing and matching step, the large language model is used to process the natural language information input by the user, thereby generating personalized medical recommendations.

8. The medical recommendation service method based on a large language model combined with privacy matrix decomposition according to claim 2, characterized in that: The personalized recommendation generation step combines multi-dimensional data such as medical resource information and medical insurance policies to provide more personalized and accurate medical recommendation services.

9. The system or method according to claim 1 or 2, characterized in that: The privacy matrix processing steps also include: Define the scope of data collection: clarify all the data types that need to be collected for the project and their processing methods to ensure the comprehensiveness and standardization of data collection; Design privacy protection measures: Configure corresponding privacy protection strategies for each data type, and use encryption and anonymization technologies to protect data privacy; System implementation and monitoring: Implement technical solutions for data collection, storage, use and sharing, and conduct regular security audits and privacy monitoring to ensure the security and compliance of system operations.

Images