A Method and System for Predicting the Outcomes of Traditional Chinese Medicine Consultation in Urology Using Large Language Models

Abstract

This invention discloses a method and system for predicting the results of traditional Chinese medicine (TCM) consultations in urology using a large language model, belonging to the technical field of large language modeling. This invention preprocesses disease description data and obtains self-correlation parameters between the disease description data; then, based on the large language model, it constructs a prediction model; outputting core keyword data describing lesions enables the prediction of specific causes in consultations, improving accuracy; by recording feedback data related to the core keyword data in real time; inputting the feedback data into a self-correcting model and outputting adjustment values; and adjusting the parameters within the large language model in real time based on the adjustment values, through continuous use and adjustment based on feedback data, the prediction model for a specific knowledge domain in urology is further refined, thereby further improving the accuracy of predicting specific causes in consultations.

Description

Technical Field

[0001] This invention relates to the field of large language model technology, and in particular to a method and system for predicting the results of traditional Chinese medicine consultations in urology using large language models. Background Technology

[0002] Existing medical big language models, such as Bian Que and Ling Xin, often use the natural language understanding (NLP) and reasoning capabilities of big language models (LLM) and combine them with private knowledge bases for data analysis and fine-tuning to accurately understand the text content related to TCM consultation reports in urology and develop systems that recommend food and medicine homology prescriptions based on big language models.

[0003] However, in practical use, existing artificial intelligence technology, based on TCM consultation reports, can only provide prescriptions for diseases that are also derived from food and medicine. Moreover, most recommendations cannot output the specific etiological parameters influencing the diagnosis, and therefore cannot provide corresponding prescriptions for food and medicine based on the specific etiological factors in the consultation. Summary of the Invention

[0004] To address the problems of existing technologies, embodiments of the present invention provide a method and system for predicting the results of traditional Chinese medicine consultations in urology using a large language model, comprising:

[0005] On the one hand, a method for predicting the results of traditional Chinese medicine consultations in urology using a large language model is provided, the method comprising:

[0006] Acquire disease description data; the disease description data includes at least one of text data and image data;

[0007] The disease description data is preprocessed, and the self-correlation parameters between the disease description data are obtained;

[0008] Construct a prediction model based on the large language model;

[0009] The prediction model is trained based on the preprocessed disease description data and the self-correlation parameters to obtain the trained prediction model; the output of the trained prediction model is core keyword data, which is used as reference data for consultation and drug recommendation.

[0010] A self-correction model is established, which is used to correct the parameters within the large language model;

[0011] Real-time recording of feedback data related to the core keyword data;

[0012] The feedback data is input into the self-correction model, and the adjusted value is output.

[0013] The parameters within the large language model are adjusted in real time based on the adjustment value.

[0014] Optionally, the disease description data is preprocessed, and the self-correlation parameters between the disease description data are obtained, including:

[0015] Extract keywords from the disease description data;

[0016] The keywords are logically categorized and structured.

[0017] Based on the preset recognition model, the self-correlation parameters between the keywords are obtained.

[0018] Optionally, constructing the prediction model based on the large language model includes:

[0019] Construct a basic large language model, where the input of the basic large language model is keywords and the output is core keywords;

[0020] A basic probabilistic neural network is constructed, wherein the input of the basic probabilistic neural network is the core keyword, and the output is the core keyword data and the probability of the corresponding lesion.

[0021] Optionally, training the prediction model based on the preprocessed disease description data and the autocorrelation parameters to obtain the trained prediction model includes:

[0022] Set a first evaluation model corresponding to the basic large language model, and a second evaluation model corresponding to the basic probabilistic neural network;

[0023] The prediction model is trained based on the preprocessed disease description data and the autocorrelation parameters.

[0024] Based on the first evaluation model and the second evaluation model, the completion conditions of the prediction model are set.

[0025] Optionally, establishing the self-correcting model includes:

[0026] A correction function is set, which is used to indicate the correlation between the feedback data and the parameters in the large language model;

[0027] A feedback data acquisition model is set up using a large language model. The feedback data acquisition model is used to acquire secondary descriptions and operations from doctors based on the core keyword data.

[0028] Based on the correction function and the feedback data acquisition model, a self-correction model is established. The input of the self-correction model is the correction keyword, and the output is the adjustment value.

[0029] Optionally, the real-time recording of feedback data related to the core keyword data includes:

[0030] Obtain secondary descriptions and operations from doctors based on the core keyword data;

[0031] The secondary description and operation are preprocessed to obtain the correction keywords.

[0032] Optionally, adjusting the parameters within the large language model in real time based on the adjustment value includes:

[0033] The secondary descriptions and operations are then transformed into disease description data;

[0034] The trained prediction model is then retrained based on the disease description data.

[0035] Optionally, the method further includes:

[0036] The image data is identified and converted into text data.

[0037] On the other hand, a urological TCM consultation result prediction system applying a large language model is provided. The system includes a client and a processing device, wherein:

[0038] The client is used to obtain disease description data; the disease description data includes at least one of text data and image data.

[0039] The processing device is used to preprocess the disease description data and obtain the self-correlation parameters between the disease description data;

[0040] The processing device is used to construct a prediction model based on a large language model;

[0041] The processing device is used to train the prediction model based on the preprocessed disease description data and the autocorrelation parameters to obtain the trained prediction model; the output of the trained prediction model is core keyword data, which is used as reference data for consultation and drug recommendation.

[0042] The processing device is used to establish a self-correction model, which is used to correct the parameters within the large language model;

[0043] The client is used to record feedback data related to the core keyword data in real time.

[0044] The processing device is used to input the feedback data into the self-correction model and output the adjustment value;

[0045] The processing device is used to adjust the parameters within the large language model in real time according to the adjustment value.

[0046] Optionally, the processing device is specifically used for:

[0047] Extract keywords from the disease description data;

[0048] The keywords are logically categorized and structured.

[0049] Based on the preset recognition model, the self-correlation parameters between the keywords are obtained.

[0050] Optionally, the processing device specifically uses:

[0051] Construct a basic large language model, where the input of the basic large language model is keywords and the output is core keywords;

[0052] A basic probabilistic neural network is constructed, wherein the input of the basic probabilistic neural network is the core keyword, and the output is the core keyword data and the probability of the corresponding lesion.

[0053] Optionally, the processing device specifically uses:

[0054] Set a first evaluation model corresponding to the basic large language model, and a second evaluation model corresponding to the basic probabilistic neural network;

[0055] The prediction model is trained based on the preprocessed disease description data and the autocorrelation parameters.

[0056] Based on the first evaluation model and the second evaluation model, the completion conditions of the prediction model are set.

[0057] Optionally, the processing device specifically uses:

[0058] A correction function is set, which is used to indicate the correlation between the feedback data and the parameters in the large language model;

[0059] A feedback data acquisition model is set up using a large language model. The feedback data acquisition model is used to acquire secondary descriptions and operations from doctors based on the core keyword data.

[0060] Based on the correction function and the feedback data acquisition model, a self-correction model is established. The input of the self-correction model is the correction keyword, and the output is the adjustment value.

[0061] Optionally, the processing device specifically uses:

[0062] Obtain secondary descriptions and operations from doctors based on the core keyword data;

[0063] The secondary description and operation are preprocessed to obtain the correction keywords.

[0064] Optionally, the processing device specifically uses:

[0065] The secondary descriptions and operations are then transformed into disease description data;

[0066] The trained prediction model is then retrained based on the disease description data.

[0067] Optionally, the processing device specifically uses:

[0068] The image data is identified and converted into text data.

[0069] The present invention has at least the following beneficial effects:

[0070] 1. By preprocessing disease description data and obtaining self-correlation parameters between disease description data, and constructing a prediction model based on a large language model, the system outputs core keyword data to describe lesions, thereby enabling the prediction of specific causes of disease in the consultation and improving accuracy.

[0071] 2. By recording feedback data related to core keyword data in real time; inputting the feedback data into the self-correction model and outputting adjustment values; adjusting the parameters within the large language model in real time based on the adjustment values; through continuous use and adjustment based on feedback data, the prediction model for specific knowledge domains in the field of urology has been further refined, thereby further enabling the prediction of specific causes of diseases in consultation and improving accuracy.

[0072] 3. This invention outputs core keyword data through a predictive model and uses this core keyword data as reference data for application in consultation and symptom-related drug and food recommendations; thereby establishing a more accurate, faster, and higher-quality drug and food homology prescription recommendation system, which helps urology patients quickly find suitable drug and food homology prescriptions, saving costs for doctors and patients. Attached Figure Description

[0073] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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.

[0074] Figure 1 A schematic flowchart illustrating a method for constructing perception capabilities for long-range robot tasks, provided by an embodiment of the present invention;

[0075] Figure 2A schematic diagram of a perception capability construction system for long-range robot tasks provided in an embodiment of the present invention;

[0076] Figure 3 This is a schematic diagram of a sensing capability building device provided in an embodiment of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0078] The method and system for predicting the results of traditional Chinese medicine consultations in urology using a large language model, as described in this invention, outputs core keyword data and the correlations between these core keyword data after training the large language model (i.e., the prediction model in this paper). This core keyword data and the correlations between it serve as reference data for consultations and symptom-related food and medicine recommendations. This establishes a more accurate, faster, and higher-quality system for recommending food and medicine formulas with the same origin, helping urology patients quickly find suitable formulas and saving costs for both doctors and patients.

[0079] The above training process is achieved by fine-tuning the large language model based on urological clinical case data, classic Chinese medicine textbooks, and Chinese medicine consultation reports. This embodiment of the invention uses urological clinical case data as an example to illustrate the technical effects that the method described in this embodiment of the invention can achieve.

[0080] It should be noted that the urology TCM consultation result prediction system using a large language model described in this embodiment of the invention, in addition to executing the urology TCM consultation result prediction method using a large language model described in this embodiment of the invention, also includes a series of functions such as data preprocessing, knowledge base management, continuous iterative optimization, large language model API access, large language model fine-tuning, testing and optimization, interface design and development, system testing and optimization, and daily maintenance.

[0081] Reference Figure 1 As shown, a method for predicting the results of traditional Chinese medicine consultations in urology using a large language model is provided. The method includes:

[0082] 101. Obtain disease description data;

[0083] The disease description data includes at least one of text data and image data;

[0084] The sources of the aforementioned disease description data include diagnostic reports, consultation data entered by doctors, and disease description data entered by patients (such as those entered through online consultation systems).

[0085] Some of the disease description data is image data, such as images of lesions that affect diagnostic reports, images of lesions uploaded by doctors, and images of lesions uploaded by patients.

[0086] 102. Preprocess the disease description data and obtain the self-correlation parameters between the disease description data;

[0087] 103. Construct a prediction model based on the large language model;

[0088] 104. Based on the preprocessed disease description data and autocorrelation parameters, train the prediction model to obtain the trained prediction model.

[0089] The output of the trained prediction model is core keyword data and the correlation between core keyword data. The core keyword data and the correlation between core keyword data are used as reference data for consultation and drug recommendation.

[0090] 105. Establish a self-correction model, which is used to correct parameters within the large language model;

[0091] 106. Record feedback data related to core keyword data in real time;

[0092] 107. Input the feedback data into the self-correction model and output the adjustment value;

[0093] 108. Adjust the parameters within the large language model in real time based on the adjustment values.

[0094] Optional, refer to Figure 2 As shown, the disease description data undergoes preprocessing, and the self-correlation parameters between disease description data are obtained, including:

[0095] 201. Extract keywords from disease description data;

[0096] Specifically, for text data, the above process is as follows:

[0097] Keywords in text data are identified through a large language model. This embodiment of the invention does not limit the training process of the large language model.

[0098] For image data, depending on its source, the above process can be specifically described as follows:

[0099] Set standard keywords corresponding to the recognition results; the above standard keywords correspond to the feature images.

[0100] Based on the depth image recognition algorithm, the feature image is identified;

[0101] Based on neural networks (such as probabilistic neural networks), feature images are classified to obtain the standard keywords corresponding to the feature images;

[0102] Obtain the descriptive text of the feature image, such as a diagnostic report or a description of the lesion entered by the patient or doctor;

[0103] By using a large language model, keywords in the descriptive text are identified;

[0104] Set the identified keywords and standard keywords as keywords for the image data;

[0105] 202. Categorize keywords logically and structure them;

[0106] 203. Based on the preset recognition model, obtain the self-correlation parameters between keywords.

[0107] The above process can be specifically described as follows:

[0108] Set up a causal model (such as a structural equation model) whose input data is any two keywords and whose output is the autocorrelation parameter (i.e., correlation) between the two keywords;

[0109] Optionally, based on the large language model, constructing a prediction model includes:

[0110] Construct a basic large language model. The input of the basic large language model is keywords, and the output is core keywords.

[0111] Construct a basic probabilistic neural network. The input of the basic probabilistic neural network is the core keyword, and the output is the core keyword data and the probability of the corresponding lesion.

[0112] Optionally, the prediction model is trained based on the preprocessed disease description data and autocorrelation parameters to obtain the trained prediction model, including:

[0113] Set up the first evaluation model corresponding to the basic large language model, and the second evaluation model corresponding to the basic probabilistic neural network;

[0114] The prediction model is trained based on the preprocessed disease description data and the autocorrelation parameters.

[0115] Based on the first and second evaluation models, the completion conditions of the prediction model are set.

[0116] Optionally, establishing a self-correcting model includes:

[0117] Set a correction function, which is used to indicate the correlation between the feedback data and the parameters in the large language model;

[0118] By using a large language model, a feedback data acquisition model is set up. This model is used to obtain secondary descriptions and operations from doctors based on core keyword data.

[0119] Based on the correction function and feedback data, a self-correction model is established. The input of the self-correction model is the correction keyword, and the output is the adjustment value.

[0120] Optional, real-time recording of feedback data related to core keyword data includes:

[0121] Obtain secondary descriptions and operations from doctors based on core keyword data;

[0122] Preprocess the secondary descriptions and operations to obtain corrective keywords.

[0123] Optionally, real-time adjustments to parameters within the large language model can be made based on the adjustment values, including:

[0124] Transform secondary descriptions and manipulations into disease description data;

[0125] Based on the disease description data, the trained prediction model is trained a second time.

[0126] Optionally, the method also includes:

[0127] Image data is recognized and converted into text data. The method described in this embodiment of the invention can also be further specified in practical applications as follows:

[0128] Step 1: Obtain consultation reports and prescription data samples from the Traditional Chinese Medicine hospital, clean and organize them into symptom and prescription data samples;

[0129] Step 2: Combine data samples with traditional Chinese medicine formulas to construct a dataset, which includes a training set and a validation set;

[0130] Step 3: Vectorize each data sample in the dataset to obtain a data sample vector;

[0131] Step 4: Classify the data sample vectors, build a sample vector library for multiple categories, and create an index;

[0132] Step 5: Select training data samples for each dataset and vectorize the symptoms in the training data samples;

[0133] Step 6: Retrieve data sample vectors similar to the symptom vectors from the sample vector library;

[0134] Step 7: Use the vector transformation model to convert the retrieved similar data sample vectors and prescription vectors back into data samples and symptoms. Use the prompt concatenation method to fuse the retrieved data samples with the symptoms as additional contextual information to construct an enhanced input.

[0135] Step 8: Fine-tune the large language model using the constructed augmented input and intelligent batching strategies.

[0136] Step 9: Calculate the similarity between the urology symptom vector and each data sample vector in the private knowledge base;

[0137] Step 10: Based on the similarity calculation results, select N data sample vectors that are similar to the symptom vector;

[0138] Step 11: Convert these N data sample vectors back into symptom and prescription data samples;

[0139] Step 12: Finally, use the reasoning ability of the large language model to summarize the prescription data sample as a unique prescription result.

[0140] Reference Figure 3 As shown, a system for predicting the results of traditional Chinese medicine consultations in urology using a large language model is provided. The system includes a client and a processing device, wherein:

[0141] The client is used to obtain disease description data; the disease description data includes at least one of text data and image data;

[0142] The processing equipment is used to preprocess disease description data and obtain self-correlation parameters between disease description data;

[0143] The processing equipment is used to build predictive models based on large language models;

[0144] The processing equipment is used to train the prediction model based on the preprocessed disease description data and autocorrelation parameters to obtain the trained prediction model; the output of the trained prediction model is core keyword data, which is used to describe the lesions.

[0145] The processing equipment is used to build a self-correction model, which is then used to correct the parameters within the large language model.

[0146] The client is used to record feedback data related to core keyword data in real time;

[0147] The processing equipment is used to input feedback data into the self-correcting model and output adjustment values;

[0148] The processing equipment is used to adjust the parameters within the large language model in real time based on the adjustment values.

[0149] Optionally, the processing equipment is specifically used for:

[0150] Extract keywords from disease description data;

[0151] Keywords are logically categorized and structured.

[0152] Based on the preset recognition model, obtain the self-correlation parameters between keywords.

[0153] Optional, the specific processing equipment used is:

[0154] Construct a basic large language model. The input of the basic large language model is keywords, and the output is core keywords.

[0155] Construct a basic probabilistic neural network. The input of the basic probabilistic neural network is the core keyword, and the output is the core keyword data and the probability of the corresponding lesion.

[0156] Optional, the specific processing equipment used is:

[0157] Set up the first evaluation model corresponding to the basic large language model, and the second evaluation model corresponding to the basic probabilistic neural network;

[0158] The prediction model is trained based on the preprocessed disease description data and the autocorrelation parameters.

[0159] Based on the first and second evaluation models, the completion conditions of the prediction model are set.

[0160] Optional, the specific processing equipment used is:

[0161] Set a correction function, which is used to indicate the correlation between the feedback data and the parameters in the large language model;

[0162] By using a large language model, a feedback data acquisition model is set up. This model is used to obtain secondary descriptions and operations from doctors based on core keyword data.

[0163] Based on the correction function and feedback data, a self-correction model is established. The input of the self-correction model is the correction keyword, and the output is the adjustment value.

[0164] Optional, the specific processing equipment used is:

[0165] Obtain secondary descriptions and operations from doctors based on core keyword data;

[0166] Preprocess the secondary descriptions and operations to obtain corrective keywords.

[0167] Optional, the specific processing equipment used is:

[0168] Transform secondary descriptions and manipulations into disease description data;

[0169] Based on the disease description data, the trained prediction model is trained a second time.

[0170] Optional, the specific processing equipment used is:

[0171] Image data is recognized and converted into text data.

[0172] The system described in this embodiment of the invention can be specifically implemented in practical applications as follows:

[0173] The basic large language model selection, based on the characteristics of the business scenario, is the recommendation of traditional Chinese medicine prescriptions for urology that are both food and medicine, and the understanding of TCM health consultation reports to find text data reflecting symptoms. The process includes data preparation, customizing the large language model, model training, and effect evaluation.

[0174] The fine-tuning of the large language model involves organizing a specific dataset of TCM health consultation reports related to urology. By training the model using paired data of task input and output, the language model can better master the ability to solve tasks through reading text. Fine-tuning allows the large language model to produce more accurate and higher-quality answers when recommending TCM prescriptions for urology that are both food and medicine, and significantly reduces latency when answering specific questions.

[0175] The project involves building a private knowledge base. Collected data is cleaned using artificial intelligence to remove redundant, erroneous, or irrelevant information. The cleaned data is then categorized according to a logical framework, and keywords are extracted and the data is structured for subsequent processing. Natural language processing and machine learning techniques are used for in-depth data processing and analysis to improve the intelligence level of the knowledge base. The goal is to build a fully functional and easy-to-use private knowledge base to provide strong support for the development of a recommendation system for urological prescriptions based on traditional Chinese medicine (TCM) that shares the same origin as food.

[0176] At the system architecture design level, the application layer demonstrates a high degree of flexibility, integrating diverse interaction channels such as websites, telephone hotlines, and WeChat official accounts, while reserving interfaces to accommodate more innovative interaction methods that may be introduced in the future.

[0177] The service scheduling layer has built a comprehensive and sophisticated management system, covering everything from intelligent problem identification and accurate classification to efficient management and maintenance of the knowledge base; from real-time monitoring and feedback of service quality to seamless integration of human assistance and intelligent transfer mechanisms; from in-depth data analysis and customized report generation to comprehensive assurance of system stability and security.

[0178] The resource layer focuses on selecting the best service resources, choosing industry-leading service resources to support the model fine-tuning and pre-training process. The aim is to improve the overall intelligence level and response efficiency of the service through high-quality data and algorithm optimization.

[0179] At the lowest level of the deployment architecture, we adopted an integration strategy, deeply integrating the private knowledge base into the system and introducing a series of top-tier large language models, including iFlytek Spark, Tencent Hunyuan, Alibaba Tongyi Qianwen, and Baidu Wenxin.

[0180] The system deployment mainly consists of two parts: the recommendation system and the knowledge base system.

[0181] The recommendation system works as follows: When a urology-related TCM health consultation report is input into the system, the system first identifies the symptoms in the report text. Then, the system quickly retrieves the most relevant answer texts from its database. Next, using artificial intelligence technology based on a large language model, the system generates a concise final answer. The recommendation system can also be integrated with websites and WeChat official accounts, forming a complete recommendation system for urology prescriptions made from TCM herbs that are both food and medicine.

[0182] The knowledge base system analyzes the frequency and timeliness of content updates to ensure that the knowledge base reflects the latest urological prescriptions related to both food and medicine, achieving optimized effects such as updating existing symptoms to correspond with new urological prescriptions related to both food and medicine, and supplementing missing content.

[0183] In terms of routine maintenance, system administrators can fully utilize the platform's functions to deeply analyze user log data. This process involves not only verifying the accuracy of question classification and ensuring the comprehensiveness of keyword extraction, but also assessing the accuracy of speech analysis technology. This serves as a basis for determining whether necessary fine-tuning and optimization of the large language model and knowledge base are required. Subsequently, through a series of systematic testing and iterative optimization measures, the aim is to significantly improve the quality of system recommendations and user experience.

[0184] The above specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0185] The technical features of the above embodiments can be combined in any way (as long as there is no contradiction in the combination of these technical features). For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written should also be considered to be within the scope of this specification.

[0186] The present invention has been described in detail above through general description and specific embodiments. It should be noted that, without departing from the concept of the present invention, various modifications and improvements can be made to these specific embodiments, all of which fall within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the appended claims.

[0187] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A urology traditional Chinese medicine consultation result prediction method using a large language model, characterized in that, The method comprises: acquiring disease description data; the disease description data comprises at least one of text data and image data; preprocessing the disease description data and acquiring self-association parameters between the disease description data; constructing a prediction model according to a large language model; training the prediction model according to the preprocessed disease description data and the self-association parameters to obtain a trained prediction model; the output of the trained prediction model is core keyword data, which is used as reference data for inquiry and drug recommendation; establishing a self-correcting model for correcting parameters in the large language model; real-time recording of feedback data related to the core keyword data; inputting the feedback data into the self-correcting model to output an adjustment value; real-time adjustment of the parameters in the large language model according to the adjustment value; the construction of the prediction model according to the large language model comprises: constructing a basic large language model, the input of which is a keyword and the output of which is a core keyword; constructing a basic probabilistic neural network, the input of which is the core keyword and the output of which is core keyword data and the probability of the corresponding lesion; the establishment of the self-correcting model comprises: setting a correction function for indicating the correlation between the feedback data and the parameters in the large language model; setting a feedback data acquisition model through a large language model, which is used to acquire the secondary description and operation of doctors based on the core keyword data; establishing a self-correcting model according to the correction function and the feedback data acquisition model, the input of which is a correction keyword and the output of which is an adjustment value.

2. The method of claim 1, wherein, The preprocessing of the disease description data and the acquisition of the self-association parameters between the disease description data comprise: extracting keywords in the disease description data; logically classifying and structuring the keywords; acquiring self-association parameters between the keywords according to a preset recognition model.

3. The method of claim 2, wherein, The training of the prediction model according to the preprocessed disease description data and the self-association parameters to obtain a trained prediction model comprises: setting a first evaluation model corresponding to the basic large language model and a second evaluation model corresponding to the basic probabilistic neural network; training the prediction model according to the preprocessed disease description data and the self-association parameters; setting the completion condition of the prediction model according to the first evaluation model and the second evaluation model.

4. The method of claim 3, wherein, The real-time recording of feedback data related to the core keyword data comprises: acquiring the secondary description and operation of doctors based on the core keyword data; preprocessing the secondary description and operation to obtain a correction keyword.

5. The method of claim 4, wherein, The real-time adjustment of the parameters in the large language model according to the adjustment value comprises: converting the secondary description and operation into disease description data; second training of the trained prediction model according to the disease description data.

6. The method of claim 1, wherein, The method further comprises: The image data is recognized and converted into text data.

7. A urology traditional Chinese medicine diagnosis result prediction system applying a large language model, characterized in that, The system comprises a client and a processing device, wherein: The client is configured to obtain disease description data, the disease description data comprising at least one of text data and image data; The processing device is configured to preprocess the disease description data and obtain self-association parameters between the disease description data; The processing device is configured to construct a prediction model based on a large language model; The processing device is configured to train the prediction model based on the preprocessed disease description data and the self-association parameters, and obtain a trained prediction model, wherein an output of the trained prediction model is core keyword data, and the core keyword data is used as reference data for diagnosis and drug recommendation; The processing device is configured to establish a self-correction model, and the self-correction model is used to correct parameters in the large language model; The client is configured to record feedback data related to the core keyword data in real time; The processing device is configured to input the feedback data into the self-correction model and output an adjustment value; The processing device is configured to adjust the parameters in the large language model in real time according to the adjustment value; The construction of the prediction model based on the large language model comprises: Constructing a basic large language model, wherein an input of the basic large language model is a keyword, and an output of the basic large language model is a core keyword; Constructing a basic probabilistic neural network, wherein an input of the basic probabilistic neural network is the core keyword, and an output of the basic probabilistic neural network is core keyword data and a probability of a corresponding lesion; The establishment of the self-correction model comprises: Setting a correction function, wherein the correction function is used to indicate the correlation between the feedback data and the parameters in the large language model; Setting a feedback data acquisition model based on the large language model, wherein the feedback data acquisition model is used to obtain secondary descriptions and operations of doctors based on the core keyword data; According to the correction function and the feedback data acquisition model, a self-correction model is established, wherein an input of the self-correction model is a correction keyword, and an output of the self-correction model is an adjustment value.

8. The system of claim 7, wherein, The processing device is specifically configured to: Extract keywords in the disease description data; Classify and structure the keywords according to logic; According to a preset recognition model, obtain self-association parameters between the keywords.

Images