Federal learning-based medical procurement parameter privacy protection method and related equipment
Through layered privacy protection and federated learning technology, the data privacy and cross-institutional collaboration issues in medical procurement parameter evaluation are solved, efficient and secure multi-dimensional evaluation is achieved, and a scientific evaluation report is generated.
Patent Information
- Application Number
- CN202510931598.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-07
- Publication Date
- 2025-10-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing medical procurement parameter evaluation has problems such as data privacy leakage risks, difficulty in cross-institutional data sharing, insufficient training samples for evaluation models, inaccurate evaluation results and lack of multi-dimensional evaluation. Especially under the premise of ensuring data security and privacy, it is difficult to achieve efficient cross-institutional collaborative evaluation.
Adopting a layered privacy protection strategy and federated learning technology, a federated learning model is constructed through the DeepFM neural network and SecureBoost+ protocol to preprocess, train, and perform multi-dimensional analysis on medical procurement parameters to generate a privacy assessment report.
It ensures data privacy and security, improves the accuracy and efficiency of assessment, realizes efficient collaborative assessment across institutions, and generates comprehensive, scientific, and multi-dimensional assessment results.
Smart Images
Figure CN120744976A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of privacy protection technology, and in particular to a medical procurement parameter privacy protection method based on federated learning and related equipment. Background Art
[0002] As the digitalization of the healthcare industry continues to advance, data security and privacy protection issues are becoming increasingly prominent in the procurement process, a crucial component of the industry. Traditional parameter-based evaluation models for medical procurement typically require extensive collection of past procurement data from various medical institutions, centralized storage of this massive data, and in-depth analysis to serve as the basis for evaluation. However, this centralized data collection and analysis approach carries significant risks. On the one hand, data leaks are highly susceptible to the slightest care in data collection, transmission, and storage. Once leaked, sensitive information such as patients' personal privacy and medical institutions' trade secrets will be exposed, resulting in immeasurable losses for the relevant parties. On the other hand, this data collection approach also carries the risk of privacy violations, failing to fully consider the privacy rights of data subjects, and can easily lead to legal disputes and a crisis of trust.
[0003] At the same time, significant differences exist in data management across medical institutions, with diverse data standards and formats lacking unified norms. This makes cross-institutional data sharing challenging, making it difficult to effectively integrate and utilize data. These data sharing barriers not only limit the amount of data available to evaluation models, resulting in insufficient training samples and thus affecting their accuracy, but also weaken the models' generalization capabilities, making them difficult to adapt to the diverse procurement scenarios and needs of different medical institutions.
[0004] Traditional manual evaluation methods are inefficient during the evaluation process. Medical procurement parameter reviews often require professionals to conduct a thorough, one-by-one comparison based on complex regulations and past clinical experience. This process is time-consuming and labor-intensive, and due to the limitations of manual operation, the parameter integrity miss rate remains high, making it prone to oversights and jeopardizing the scientific nature and accuracy of procurement decisions.
[0005] Furthermore, existing medical procurement parameter evaluation systems suffer from significant functional flaws. Most of them only verify parameter compliance from a single dimension, overlooking key factors such as compatibility with actual clinical needs and the product's cost-effectiveness. In medical procurement, simply pursuing parameter compliance is insufficient; comprehensive considerations are also required, including whether the product truly meets clinical treatment needs and offers a high cost-effectiveness. Existing systems lack these comprehensive evaluation dimensions, making it difficult for evaluation results to fully reflect the value of the purchased products.
[0006] Crucially, in the field of medical procurement parameter assessment, there is currently no publicly available solution that effectively integrates the three key elements of multi-dimensional evaluation, cross-institutional collaboration, and privacy protection. Achieving efficient cross-institutional collaborative evaluation while ensuring data security and privacy, and building a comprehensive, scientific, multi-dimensional assessment system, has become a technical bottleneck that urgently needs to be overcome in the current medical procurement field. This gap needs to be filled to promote the digital transformation and healthy development of the medical procurement industry. Summary of the Invention
[0007] Based on the problems raised by the above background technology, the purpose of the present invention is to provide a medical procurement parameter privacy protection method and related equipment based on federated learning, which solves the current problem in the medical procurement field of lacking efficient cross-institutional collaborative evaluation while ensuring data security and privacy.
[0008] The present invention is achieved through the following technical solutions: The first aspect of the present invention provides a method for protecting medical procurement parameter privacy based on federated learning, comprising the following steps: Step S1: obtaining medical procurement parameters, and preprocessing the medical procurement parameters using a layered privacy protection strategy to obtain private medical procurement parameters; Step S2: constructing a federated learning model using the DeepFM neural network and the SecureBoost+ protocol, and processing the private medical procurement parameters using the federated learning model to obtain a private medical procurement result; Step S3: Perform a multi-dimensional analysis on the privacy-based medical procurement results, and generate an analysis report based on the results of the multi-dimensional analysis.
[0009] In this technical solution, a layered privacy protection strategy is employed to preprocess medical procurement parameters to obtain private medical procurement parameters. A federated learning model is then constructed using the DeepFM neural network and the SecureBoost+ protocol to process these private medical procurement parameters to obtain private medical procurement results. Finally, a multi-dimensional analysis of these results is performed to generate an analysis report. This ensures the privacy and security of medical procurement data, mitigates the risk of data leakage, and leverages the federated learning model to fully leverage the value of data, improving the scientific nature and accuracy of medical procurement decisions. This technical solution, combining layered privacy protection with federated learning, provides a novel solution for the medical procurement sector that balances data privacy and utilization efficiency.
[0010] In an optional embodiment, step S1 includes the following steps: Generalizing the medical procurement parameters based on the k-anonymity principle to obtain generalized medical procurement parameters; A correlation feature perturbation engine is used to detect strongly correlated feature combinations in the generalized medical procurement parameters, and Laplace noise is injected into the strongly correlated feature combinations to obtain private medical procurement parameters.
[0011] In an optional embodiment, step S1 further includes: If the medical procurement parameter is an unstructured text, performing NLP parsing on the medical procurement parameter to obtain text data; A clinical term confusion layer is used to replace clinical terms in the text data to generate obfuscated text data.
[0012] In an optional embodiment, the federated learning model includes: a local DeepFM model, a global DeepFM model, and a SecureBoost+ protocol aggregation model; The local DeepFM model is used to train the hospital's local privacy-preserving medical procurement parameters to generate DeepFM model parameters. The SecureBoost+ protocol aggregation model is used to encrypt and aggregate the DeepFM model parameters to generate aggregated model parameters; The global DeepFM model is used to train the aggregation model parameters to generate privacy-sensitive medical procurement results.
[0013] In an optional embodiment, step 2 includes the following steps: The first hospital uses the local DeepFM model to train the privacy-preserving medical procurement parameters to obtain first DeepFM model parameters; The second hospital uses the local DeepFM model to train the privacy-preserving medical procurement parameters to obtain second DeepFM model parameters; Performing encryption aggregation processing on the first DeepFM model parameters and the second DeepFM model parameters using the SecureBoost+ protocol aggregation model to obtain aggregated model parameters; The aggregation model parameters are iteratively trained using a global DeepFM model, and the iteratively trained global DeepFM model is used to generate privacy-preserving medical procurement results.
[0014] In an optional embodiment, the encryption aggregation process includes the following steps: Performing public key encryption on the SecureBoost+ protocol aggregation model; Based on the SecureBoost+ protocol aggregation model after public key encryption, the first DeepFM model parameters and the second DeepFM model parameters are searched for secure splitting points under vertical federation using obfuscation circuit technology, and the model parameters obtained by the secure splitting point search are aggregated to generate aggregated model parameters.
[0015] In an optional embodiment, the encryption aggregation process further includes: Encrypting the first DeepFM model parameter and the second DeepFM model parameter during transmission using a through-state encryption algorithm; Gaussian noise is injected in the process of finding safe splitting points to obtain model parameter aggregation. A second aspect of the present invention provides a medical procurement parameter privacy protection system based on federated learning, comprising: a privacy processing module, configured to obtain medical procurement parameters and pre-process the medical procurement parameters using a layered privacy protection strategy to obtain privacy-sensitive medical procurement parameters; A federated learning module, configured to construct a federated learning model using the DeepFM neural network and the SecureBoost+ protocol, and to process the private medical procurement parameters using the federated learning model to obtain private medical procurement results; The multi-dimensional analysis module is used to perform multi-dimensional analysis on the privacy-based medical procurement results and generate an analysis report based on the results of the multi-dimensional analysis.
[0016] A third aspect of the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a medical procurement parameter privacy protection method based on federated learning when executing the computer program.
[0017] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a medical procurement parameter privacy protection method based on federated learning.
[0018] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. Combining a layered privacy protection strategy with federated learning technology not only effectively addresses data privacy issues in medical procurement parameter evaluation, but also improves the accuracy and efficiency of evaluation. 2. By introducing multiple encryption technologies and differential privacy protection mechanisms in the federated learning process, data security is further enhanced. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the examples. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. In the drawings: Figure 1 A flowchart of the method for protecting medical procurement parameter privacy based on federated learning provided in Example 1 of the present invention; Figure 2 This is a structural diagram of an electronic device provided in Example 3 of the present invention. DETAILED DESCRIPTION
[0020] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.
[0021] Embodiment 1 of the present invention provides a medical procurement parameter privacy protection method based on federated learning, such as Figure 1 As shown in Figure 1, the medical procurement parameter privacy protection method based on federated learning includes the following steps: Step S1: obtaining medical procurement parameters, and preprocessing the medical procurement parameters using a layered privacy protection strategy to obtain private medical procurement parameters; Step S2: constructing a federated learning model using the DeepFM neural network and the SecureBoost+ protocol, and processing the private medical procurement parameters using the federated learning model to obtain a private medical procurement result; Step S3: Perform a multi-dimensional analysis on the privacy-based medical procurement results, and generate an analysis report based on the results of the multi-dimensional analysis.
[0022] It is important to note that by preprocessing medical procurement parameters using a layered privacy protection strategy to obtain private medical procurement parameters, a federated learning model is constructed using the DeepFM neural network and the SecureBoost+ protocol to process these private medical procurement parameters to obtain private medical procurement results. Finally, a multi-dimensional analysis of these private medical procurement results is performed to generate an analysis report. This ensures the privacy and security of medical procurement data, mitigates the risk of data leakage, and leverages the federated learning model to fully leverage the value of data, improving the scientific nature and accuracy of medical procurement decisions. This technical solution, combining layered privacy protection with federated learning, provides a novel solution for the medical procurement field that balances data privacy and utilization efficiency.
[0023] In this example, we use Hospital A (a tertiary hospital, teaching hospital), Hospital B (a tertiary hospital, general hospital), and Hospital C (a secondary hospital, focusing on primary healthcare) as examples to implement a medical procurement parameter privacy protection method based on federated learning. The purpose is to jointly evaluate the procurement parameters of a certain brand of anesthesia machine (model X-2024) to ensure that the parameters meet regulatory requirements, match clinical needs, and have a reasonable price-performance ratio, while protecting the privacy of each hospital's procurement data.
[0024] In an optional embodiment, step S1 includes the following steps: Generalizing the medical procurement parameters based on the k-anonymity principle to obtain generalized medical procurement parameters; A correlation feature perturbation engine is used to detect strongly correlated feature combinations in the generalized medical procurement parameters, and Laplace noise is injected into the strongly correlated feature combinations to obtain private medical procurement parameters.
[0025] It should be noted that the k-anonymity principle is a data privacy protection technology designed to protect individual privacy by ensuring that each individual in the dataset has at least k-1 other individuals sharing the same feature combination through data anonymization. The k-anonymity principle requires that when publishing or sharing data, each individual in the dataset cannot be uniquely identified, and at least k individuals share the same attribute value combination. In this step, the primary purpose of the k-anonymity principle is to prevent unique identification of medical data, thereby protecting the privacy of medical data. K-anonymity can reduce the risk of individual identification after a data leak and safeguard hospital procurement privacy. A correlation feature perturbation engine is used to detect strongly correlated feature combinations in the data, such as the unusual combination of "high-end model + ultra-low price." Laplace noise is then injected into these sensitive correlated fields. The noise intensity is positively correlated with the mutual information between features; that is, the stronger the correlation, the greater the noise intensity.
[0026] Furthermore, k is greater than or equal to 5.
[0027] In this example, a layered privacy protection strategy is first used to preprocess the medical procurement parameters for Hospitals A, B, and C. Sensitive fields such as "Hospital Name" and "Purchase Time" are generalized, and "Anesthesia Machine Model" and "Budget Amount" are divided into intervals. For example: a budget of 500,000 RMB is categorized as "400,000-600,000 RMB," and a model number of X-2024 is categorized as "High-End Series." The generalized procurement parameter data destroys individual identifiability and obscures specific details.
[0028] After completing k-anonymization, we perturb the associated features. We detected a strong correlation between the "high-end model" and "budget 500,000" combination, with a mutual information of 0.9. Laplace noise (scale parameter = 0.5 × mutual information = 0.45) was injected into the "budget" field, reducing the budget to 520,000 (with a ±10% fluctuation tolerance). This noise injection destroyed strongly correlated feature combinations in the procurement parameter data, masking any unusual correlations. This step enhances data privacy, preventing sensitive information from being inferred through feature combinations and reducing the potential for interference from unusual correlations in the data on subsequent analysis.
[0029] In an optional embodiment, step S1 further includes: If the medical procurement parameter is an unstructured text, performing NLP parsing on the medical procurement parameter to obtain text data; A clinical term confusion layer is used to replace clinical terms in the text data to generate obfuscated text data.
[0030] It should be noted that natural language processing (NLP) technology is first used to parse text data to extract key information and clinical terms. A clinical term obfuscation layer is then added to the parsed text. By replacing or adding similar clinical terms, this prevents privacy leaks caused by keyword uniqueness. This protects privacy information in unstructured text data, preventing the identification of individuals or sensitive information through specific keywords or terms, while preserving the semantic information of the text data for subsequent analysis.
[0031] In an optional embodiment, the federated learning model includes: a local DeepFM model, a global DeepFM model, and a SecureBoost+ protocol aggregation model; The local DeepFM model is used to train the hospital's local privacy-preserving medical procurement parameters to generate DeepFM model parameters. The SecureBoost+ protocol aggregation model is used to encrypt and aggregate the DeepFM model parameters to generate aggregated model parameters; The global DeepFM model is used to train the aggregation model parameters to generate privacy-sensitive medical procurement results.
[0032] In an optional embodiment, step 2 includes the following steps: The first hospital uses the local DeepFM model to train the privacy-preserving medical procurement parameters to obtain first DeepFM model parameters; The second hospital uses the local DeepFM model to train the privacy-preserving medical procurement parameters to obtain second DeepFM model parameters; Performing encryption aggregation processing on the first DeepFM model parameters and the second DeepFM model parameters using the SecureBoost+ protocol aggregation model to obtain aggregated model parameters; The aggregation model parameters are iteratively trained using a global DeepFM model, and the iteratively trained global DeepFM model is used to generate privacy-preserving medical procurement results.
[0033] It should be noted that Hospital A, Hospital B, and Hospital C independently train the DeepFM model using local data to generate local DeepFM models and DeepFM model parameters. Without sharing the original data, each hospital can use local data to train the model, preserving the privacy of the data while providing a basis for subsequent model aggregation.
[0034] Then, model parameters are aggregated through the SecureBoost+ protocol, and hospital A (holding clinical demand data) and hospital B (holding price data) are vertically federated. While protecting the data privacy of each hospital, the model parameters are securely aggregated to generate a global federated learning model. This model can integrate the data characteristics of each hospital to improve the model's accuracy and generalization ability; the aggregated model parameters are decrypted and integrated into a global DeepFM model to complete the training of the federated learning model. This step realizes the training of the federated learning model for generalized medical procurement parameters while protecting data privacy, and obtains private medical procurement results through the trained federated learning model.
[0035] In an optional embodiment, the encryption aggregation process includes the following steps: Performing public key encryption on the SecureBoost+ protocol aggregation model; Based on the SecureBoost+ protocol aggregation model after public key encryption, the first DeepFM model parameters and the second DeepFM model parameters are searched for secure splitting points under vertical federation using obfuscation circuit technology, and the model parameters obtained by the secure splitting point search are aggregated to generate aggregated model parameters.
[0036] In an optional embodiment, the encryption aggregation process further includes: Encrypting the first DeepFM model parameter and the second DeepFM model parameter during transmission using a through-state encryption algorithm; Gaussian noise is injected in the process of finding safe splitting points to obtain model parameter aggregation.
[0037] It should be noted that the public key is used to encrypt the aggregation model parameters of the SecureBoost+ protocol to ensure the security of data during transmission; the obfuscation circuit technology is used to achieve secure split point search under vertical federation to prevent the leakage of data information of any party during the aggregation process; during the transmission process, the BFV homomorphic encryption algorithm is used to encrypt the model parameters, allowing direct calculation on the encrypted data without decryption, further enhancing the privacy protection of the data; Gaussian noise is injected during the aggregation process to achieve differential privacy protection, where the privacy parameters ε=2.0 and δ=1e -5 Differential privacy protects individual data by adding noise, making it difficult to infer information about a single data point from the aggregated results.
[0038] After the 10th round of training, ε decays to 1.1 (dynamic privacy budget).
[0039] Furthermore, a multi-dimensional analysis of the privacy-focused medical procurement results was conducted, including: Compliance assessment: Compare with the "Medical Device Classification Catalog" and conduct a compliance assessment on the privacy-focused medical procurement results to determine whether they comply with the specifications and whether there are any missing parameters. If parameters are found to be missing, such as the mandatory parameter "anesthetic gas leakage test report", a red alert will be triggered.
[0040] Clinical Matching: When Hospital A has pediatric needs, a cosine similarity calculation is performed. For example, a cosine similarity calculation is performed on the parameter "tidal volume 500-1000mL" and the label "neonatal tidal volume ≥ 600mL". A cosine similarity of 0.85 is considered qualified. When Hospital C has basic medical care, a preliminary similarity calculation is performed on the parameter "operation interface complexity" and its matching label "one-click operation required". If the similarity is 0.62, it needs to be optimized.
[0041] Cost-effectiveness analysis: If the historical average transaction price is 480,000, the current quoted price is 520,000, and the price elasticity coefficient = 1.2 (a reasonable premium of ≤15%) is allowed), the cost-effectiveness is judged to be medium (75 points).
[0042] Finally, an evaluation report is generated and privacy verification is performed.
[0043] Embodiment 2 of the present invention provides a medical procurement parameter privacy protection system based on federated learning, including: a privacy processing module, configured to obtain medical procurement parameters and pre-process the medical procurement parameters using a layered privacy protection strategy to obtain privacy-sensitive medical procurement parameters; A federated learning module, configured to construct a federated learning model using the DeepFM neural network and the SecureBoost+ protocol, and to process the private medical procurement parameters using the federated learning model to obtain private medical procurement results; The multi-dimensional analysis module is used to perform multi-dimensional analysis on the privacy-based medical procurement results and generate an analysis report based on the results of the multi-dimensional analysis.
[0044] Embodiment 3 of the present invention provides an electronic device, such as Figure 2 As shown, the electronic device includes a processor 21, a memory 22, an input device 23 and an output device 24; the number of processors 21 in the computer device can be one or more. Figure 2 In the figure, a processor 21 is taken as an example; the processor 21, memory 22, input device 23 and output device 24 in the electronic device can be connected by a bus or other means. Figure 2 The bus connection is taken as an example.
[0045] Memory 22, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules. Processor 21 executes the software programs, instructions, and modules stored in memory 22 to perform various electronic device functions and data processing, thereby implementing the federated learning-based medical procurement parameter privacy protection method of Example 1.
[0046] The memory 22 may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function; the data storage area may store data created based on the use of the terminal, etc. Furthermore, the memory 22 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state memory device. In some instances, the memory 22 may further include a memory remotely located relative to the processor 21, and these remote memories may be connected to the electronic device via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0047] The input device 23 can be used to receive the ID and password input by the user. The output device 24 is used to output the network configuration page.
[0048] Embodiment 4 of the present invention also provides a computer-readable storage medium, wherein the computer-executable instructions, when executed by a computer processor, are used to implement the medical procurement parameter privacy protection method based on federated learning as provided in embodiment 1.
[0049] An embodiment of the present invention provides a storage medium containing computer-executable instructions, and its computer-executable instructions are not limited to the method operations provided in Example 1, but can also execute related operations in the medical procurement parameter privacy protection method based on federated learning provided in any embodiment of the present invention.
[0050] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A medical procurement parameter privacy protection method based on federated learning, characterized by: The steps include: Step S1: obtaining medical procurement parameters, and preprocessing the medical procurement parameters using a layered privacy protection strategy to obtain private medical procurement parameters; Step S2: constructing a federated learning model using the DeepFM neural network and the SecureBoost+ protocol, and processing the private medical procurement parameters using the federated learning model to obtain a private medical procurement result; Step S3: Perform a multi-dimensional analysis on the privacy-based medical procurement results, and generate an analysis report based on the results of the multi-dimensional analysis.
2. The medical procurement parameter privacy protection method based on federated learning according to claim 1 is characterized in that: The step S1 includes the following steps: Generalizing the medical procurement parameters based on the k-anonymity principle to obtain generalized medical procurement parameters; A correlation feature perturbation engine is used to detect strongly correlated feature combinations in the generalized medical procurement parameters, and Laplace noise is injected into the strongly correlated feature combinations to obtain private medical procurement parameters.
3. The medical procurement parameter privacy protection method based on federated learning according to claim 2 is characterized in that: The step S1 further includes: If the medical procurement parameter is an unstructured text, performing NLP parsing on the medical procurement parameter to obtain text data; A clinical term confusion layer is used to replace clinical terms in the text data to generate obfuscated text data.
4. The medical procurement parameter privacy protection method based on federated learning according to claim 1 is characterized in that: The federated learning model includes: a local DeepFM model, a global DeepFM model, and a SecureBoost+ protocol aggregation model; The local DeepFM model is used to train the hospital's local privacy-preserving medical procurement parameters to generate DeepFM model parameters. The SecureBoost+ protocol aggregation model is used to encrypt and aggregate the DeepFM model parameters to generate aggregated model parameters; The global DeepFM model is used to train the aggregation model parameters to generate privacy-sensitive medical procurement results.
5. The method for protecting medical procurement parameter privacy based on federated learning according to claim 4 is characterized in that: The step 2 comprises the following steps: The first hospital uses the local DeepFM model to train the privacy-preserving medical procurement parameters to obtain first DeepFM model parameters; The second hospital uses the local DeepFM model to train the privacy-preserving medical procurement parameters to obtain second DeepFM model parameters; Performing encryption aggregation processing on the first DeepFM model parameters and the second DeepFM model parameters using the SecureBoost+ protocol aggregation model to obtain aggregated model parameters; The aggregation model parameters are iteratively trained using a global DeepFM model, and the iteratively trained global DeepFM model is used to generate privacy-preserving medical procurement results.
6. The medical procurement parameter privacy protection method based on federated learning according to claim 5 is characterized in that: The encryption aggregation process includes the following steps: Performing public key encryption on the SecureBoost+ protocol aggregation model; Based on the SecureBoost+ protocol aggregation model after public key encryption, the first DeepFM model parameters and the second DeepFM model parameters are searched for secure splitting points under vertical federation using obfuscation circuit technology, and the model parameters obtained by the secure splitting point search are aggregated to generate aggregated model parameters.
7. The medical procurement parameter privacy protection method based on federated learning according to claim 5 is characterized in that: The encryption aggregation process also includes: Encrypting the first DeepFM model parameter and the second DeepFM model parameter during transmission using a through-state encryption algorithm; Gaussian noise is injected in the process of finding safe splitting points to obtain model parameter aggregation.
8. A medical procurement parameter privacy protection system based on federated learning, characterized by: include: a privacy processing module, configured to obtain medical procurement parameters and pre-process the medical procurement parameters using a layered privacy protection strategy to obtain privacy-sensitive medical procurement parameters; A federated learning module, configured to construct a federated learning model using the DeepFM neural network and the SecureBoost+ protocol, and to process the private medical procurement parameters using the federated learning model to obtain private medical procurement results; The multi-dimensional analysis module is used to perform multi-dimensional analysis on the privacy-based medical procurement results and generate an analysis report based on the results of the multi-dimensional analysis.
9. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the medical procurement parameter privacy protection method based on federated learning according to any one of claims 1 to 7 is 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, it implements the medical procurement parameter privacy protection method based on federated learning as described in any one of claims 1 to 7.