Traditional Chinese medicine prescription recommendation method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of traditional Chinese medicine prescription recommendation, in particular to a traditional Chinese medicine prescription recommendation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining first data; determining an association relationship between any two traditional Chinese medicine names in the traditional Chinese medicine prescription set; processing all the traditional Chinese medicine prescriptions in the traditional Chinese medicine prescription set according to the association relationship to obtain traditional Chinese medicine prescription data; fusing the traditional Chinese medicine prescription data into the first data to obtain second data, and performing model training on a traditional Chinese medicine prescription recommendation model by adopting the second data to obtain an optimal traditional Chinese medicine prescription recommendation model; and acquiring real-time input data, and inputting the real-time input data into the optimal traditional Chinese medicine prescription recommendation model to obtain the optimal traditional Chinese medicine prescription, thereby achieving the effect of improving the efficiency and individuation degree of traditional Chinese medicine prescription recommendation.

Description

Technical Field

[0001] The present invention relates to the field of traditional Chinese medicine prescription recommendation, and in particular to a method, device, electronic device and storage medium for traditional Chinese medicine prescription recommendation. Background Art

[0002] Traditional Chinese Medicine (TCM) is a medical theory system originating in ancient China and honed through thousands of years of practice and summary. It utilizes a variety of methods, including observation, auscultation, questioning, and palpation, to obtain information about symptoms and then implement treatment based on syndrome differentiation. In TCM clinical practice, appropriate prescriptions are selected based on the patient's individual characteristics and the principles of combining herbs to achieve the desired balance of yin and yang in the human body and ultimately cure illness.

[0003] At present, TCM diagnosis and treatment cannot distinguish between the main symptoms and secondary symptoms, prescriptions are aimless, drug combinations are chaotic, and TCM knowledge is distributed in many ancient books and documents. The loose and unstructured characteristics make the utilization rate of TCM knowledge low, resulting in low efficiency and low personalization of TCM prescription recommendations. Summary of the Invention

[0004] The embodiments of the present invention provide a method, device, electronic device and storage medium for recommending traditional Chinese medicine prescriptions, which are used to solve the problems of low efficiency and low personalization of traditional Chinese medicine prescription recommendations in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides a method for recommending a traditional Chinese medicine prescription, comprising:

[0006] Acquiring first data, where the first data includes personal information, disease information, and a set of traditional Chinese medicine prescriptions;

[0007] Determine the association relationship between any two Chinese medicine names in the Chinese medicine prescription set;

[0008] Processing all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data;

[0009] Integrating the traditional Chinese medicine prescription data into the first data to obtain second data, and using the second data to perform model training on a traditional Chinese medicine prescription recommendation model to obtain an optimal traditional Chinese medicine prescription recommendation model;

[0010] Real-time input data is acquired and input into the optimal Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription, wherein the real-time input data includes real-time personal information and real-time disease information.

[0011] In one possible implementation, determining the association relationship between any two Chinese medicine names in the Chinese medicine prescription set includes:

[0012] Obtain the names of the Chinese medicines in all the Chinese medicine prescriptions in the Chinese medicine prescription set;

[0013] Group any two Chinese medicine names from all Chinese medicine names into one group, and determine the number of times the two Chinese medicine names in each group appear simultaneously in all Chinese medicine prescriptions;

[0014] The distance between the two Chinese medicine names in each group is determined according to the number of times.

[0015] In a possible implementation, determining the distance between two Chinese medicine names in each group according to the number of times includes:

[0016] According to the total number of all TCM prescriptions and the number of times, the distance between two TCM names in each group was determined.

[0017] In a possible implementation, processing all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data includes:

[0018] Convert all determined distances into a distance matrix;

[0019] Performing cluster analysis on all the Chinese medicines in all the Chinese medicine prescriptions according to the distance matrix to obtain Chinese medicine category groups;

[0020] The name of each Chinese medicine in the Chinese medicine category group is replaced with the name of the Chinese medicine category group to which it belongs, so as to obtain abstract prescription data in groups.

[0021] In a possible implementation, the disease information includes TCM syndrome type information, which includes syndrome type, syndrome type category, syndrome type characteristics, syndrome type cause, and the relationship between various entities in the TCM syndrome type information;

[0022] After obtaining the first data, the method further includes:

[0023] Constructing a knowledge graph based on the syndrome type, the syndrome type category, the syndrome type characteristics, the syndrome type cause, and the relationship between each entity in the TCM syndrome type information;

[0024] Establishing a traditional Chinese medicine knowledge base based on the knowledge graph;

[0025] Integrate the traditional Chinese medicine knowledge base into the second data.

[0026] In one possible implementation, constructing a knowledge graph based on the syndrome type, the syndrome type category, the syndrome type characteristics, the syndrome type cause, and the relationship between entities in the TCM syndrome type information includes:

[0027] The syndrome type, syndrome category, syndrome characteristics, and syndrome cause are respectively taken as nodes, and any two nodes are extracted as entity pairs;

[0028] Obtaining the relationship between the entity pairs, and using the entity pairs and the relationship between the entity pairs as the basic unit of the knowledge graph, where the basic unit is a triple;

[0029] All triples are combined to construct the knowledge graph.

[0030] In one possible implementation, after establishing a TCM knowledge base based on the knowledge graph, the method further includes:

[0031] Extract the nodes corresponding to the entities in each triple as prompt words;

[0032] Connecting the prompt words into a first character string according to a preset rule;

[0033] constructing a second character string corresponding to the input data based on the personal information in the first data, the disease description information in the disease information, and the first character string;

[0034] Predefined control markers are used to explicitly demarcate the boundaries of system instructions, input data, and model responses to form closed context blocks.

[0035] In a second aspect, an embodiment of the present invention provides a device for recommending a traditional Chinese medicine prescription, which is used to implement the method for recommending a traditional Chinese medicine prescription as described in the first aspect or any possible implementation of the first aspect, wherein the device for recommending a traditional Chinese medicine prescription includes:

[0036] An acquisition module, configured to acquire first data, wherein the first data includes personal information, disease information, and a set of traditional Chinese medicine prescriptions;

[0037] A determination module, configured to determine an association relationship between any two Chinese medicine names in the Chinese medicine prescription set;

[0038] A processing module, configured to process all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data;

[0039] The processing module is further configured to integrate the traditional Chinese medicine prescription data into the first data to obtain second data;

[0040] A training module, configured to perform model training on a traditional Chinese medicine prescription recommendation model using the second data to obtain an optimal traditional Chinese medicine prescription recommendation model;

[0041] The acquisition module is further used to acquire real-time input data;

[0042] The processing module is further used to input the real-time input data into the optimal Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription, and the real-time input data includes real-time personal information and real-time disease information.

[0043] In a third aspect, an embodiment of the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the steps of the method for recommending traditional Chinese medicine prescriptions as described in the first aspect or any possible implementation of the first aspect.

[0044] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the method for recommending traditional Chinese medicine prescriptions as described in the first aspect or any possible implementation of the first aspect.

[0045] This embodiment provides a method, device, electronic device and storage medium for recommending traditional Chinese medicine prescriptions, by obtaining first data, which includes personal information, disease information and a set of traditional Chinese medicine prescriptions; processing the set of traditional Chinese medicine prescriptions in the first data, determining the association relationship between any two traditional Chinese medicine names in the set of traditional Chinese medicine prescriptions, and then processing all traditional Chinese medicine prescriptions in the set of traditional Chinese medicine prescriptions according to the association relationship to obtain traditional Chinese medicine prescription data; integrating the traditional Chinese medicine prescription data into the first data to obtain second data, and using the second data to train a traditional Chinese medicine prescription recommendation model to obtain an optimal traditional Chinese medicine prescription recommendation model; obtaining real-time input data, and inputting the real-time input data into the optimal traditional Chinese medicine prescription recommendation model to obtain an optimal traditional Chinese medicine prescription, the real-time input data including real-time personal information and real-time disease information, thereby realizing personalized traditional Chinese medicine prescription recommendation through personal information and disease information. In addition, through the analysis of traditional Chinese medicine prescriptions, the potential association characteristics between traditional Chinese medicine names are explored, the rules of traditional Chinese medicine compatibility are revealed, and the personalization level of traditional Chinese medicine prescription recommendation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0047] Figure 1 is a flow chart of a method for recommending a traditional Chinese medicine prescription provided by an embodiment of the present invention;

[0048] Figure 2is a flow chart of a method for recommending a traditional Chinese medicine prescription provided by another embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram of an example graph of entities and relationships provided by an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of entity-relationship-entity semantic relationship provided by an embodiment of the present invention;

[0051] Figure 5 This is a functional block diagram of a device for recommending traditional Chinese medicine prescriptions provided by an embodiment of the present invention;

[0052] Figure 6 This is a functional block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In the following description, specific details such as particular system structures and techniques are provided for purposes of illustration, not limitation, to facilitate a thorough understanding of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced in alternative embodiments without these specific details. In other instances, detailed descriptions of well-known systems, devices, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.

[0054] In order to make the objectives, technical solutions and advantages of the present invention more clear, the following will be described through specific implementation methods in conjunction with the accompanying drawings.

[0055] The following is a detailed description of an embodiment of the present invention. This example is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiment.

[0056] Figure 1 Flowchart of a method for recommending a traditional Chinese medicine prescription provided in an embodiment of the present invention.

[0057] like Figure 1 As shown, it shows a flow chart of the implementation of the method for recommending traditional Chinese medicine prescriptions provided by an embodiment of the present invention, which is described in detail as follows:

[0058] In step 101, first data is obtained, where the first data includes personal information, disease information, and a set of traditional Chinese medicine prescriptions.

[0059] Personal information can include gender, age, and other information, while disease information can include disease descriptions and Traditional Chinese Medicine (TCM) syndrome types. TCM syndrome types are general descriptions of the pathological attributes of a disease at a specific stage in Traditional Chinese Medicine (TCM) theory. These descriptions are formed through a comprehensive assessment of factors such as the cause, location, nature, and course of the disease, and can be used to guide clinical treatment. A Traditional Chinese Medicine (TCM) prescription collection includes information about each individual TCM prescription.

[0060] The first data can be represented by a data set, for example, the first data is H train ={(u1,u2,u3,…u m ,{v1,v2,…,v n}) i}, where (u1,u2,u3,…u m ) represent the sex, age, disease and TCM syndrome type of the i-th patient and their personal and disease information, respectively, and m represents the total number of personal and disease information. {v1,v2,…,v n} represents the TCM prescription corresponding to the i-th patient. n represents the number of TCM names in the current TCM prescription. The value range of i is greater than or equal to 1.

[0061] Optionally, the TCM prescription collection includes a variety of prescriptions, such as those found in ancient books and literature, prescriptions given by ordinary TCM practitioners, and even TCM prescriptions based on the diagnosis of famous TCM experts. These precious experiences of TCM prescriptions based on the diagnosis of famous TCM experts are shared, learned, and passed down by TCM practitioners and the entire society, and also provide doctors with preliminary diagnostic suggestions, which can greatly improve the efficiency of diagnosis and treatment.

[0062] In some embodiments, the first data can be divided into non-essential items and essential items for TCM prescription recommendations. Non-essential items can include information such as the patient's gender and age. While this information has some reference value for TCM prescription recommendations, it is not a decisive factor. Essential items can include the patient's disease, TCM syndrome type, and a closely corresponding TCM prescription set. This information is the core basis for recommending TCM prescriptions.

[0063] In step 102, the association relationship between any two Chinese medicine names in the Chinese medicine prescription set is determined.

[0064] Correlations exist between the herbs in traditional Chinese medicine (TCM) prescriptions, including the sovereign, minister, assistant, and envoy functions, as well as the eighteen counteractions and nineteen precautions. We explore these relationships, delve into their underlying correlations, and uncover the principles governing their compatibility. This allows us to formalize these complex relationships, standardize TCM prescriptions, and meet the data requirements for large-scale model training and fine-tuning, thereby improving the efficiency and personalization of TCM prescription recommendations. Personalization here refers to the individualized nature of TCM prescriptions.

[0065] In this embodiment, monarch, minister, assistant and envoy are terms in traditional Chinese medicine, which are the basic principles of prescription compatibility. They originally refer to the four types of people, monarch, minister, assistant and envoy, who play different roles respectively. In this embodiment, they refer to the different effects of each medicine in a traditional Chinese medicine prescription.

[0066] The Eighteen Antidotes and Nineteen Fears refer to the names of various drugs that can interact with each other. It can also be understood as meaning that certain drugs should not be used together, as their combined use will result in various side effects.

[0067] In some embodiments, all TCM names from all prescriptions are extracted and co-occurrence analysis techniques using association rule mining are used to further explore potential associations between TCMs. Step 102 includes: obtaining TCM names from all TCM prescriptions in a collection of TCM prescriptions; grouping any two TCM names from all TCM names into groups, determining the number of times the two TCM names in each group co-occur across all TCM prescriptions; and determining the distance between the two TCM names in each group based on the number of co-occurrences.

[0068] Let q={q1,q2,…q a} is the set of all Chinese medicine names, a is the number of all types of Chinese medicine names in all prescriptions. Now all Chinese medicine names are set in pairs, and p is the set of all pairwise combinations of Chinese medicine names, where p = {(q j =q k )|q j ,q k ∈q,j=(1,2,…J),k=(1,2,…K),J=K}. Let f(q j ,q k ,H train ) is the name of the Chinese medicine q j and q k The number of times it appears simultaneously in all prescriptions.

[0069] In some embodiments, determining the distance between two Chinese medicine names in each group according to the number of times includes: determining the distance between two Chinese medicine names in each group according to the total number and number of all Chinese medicine prescriptions.

[0070] For example, according to d(q j ,q k )=wf(q j ,q k ,H train ) determines the distance between two Chinese medicine names in each group, where d(q j ,q k ) is a distance function, which is the name of the Chinese medicine q j With q k The distance between them is , and w is the total number of all Chinese medicine prescriptions.

[0071] In step 103, all the Chinese medicine prescriptions in the Chinese medicine prescription set are processed according to the association relationship to obtain Chinese medicine prescription data.

[0072] In order to achieve effective grouping and standardized processing of traditional Chinese medicine prescriptions and to adapt to the data requirements of large-scale model training and fine-tuning, a cluster analysis method is used in this embodiment to systematically classify and group traditional Chinese medicines.

[0073] In some embodiments, step 103 includes: converting all determined distances into a distance matrix; performing cluster analysis on all TCM prescriptions based on the distance matrix to obtain TCM category groups; and replacing the name of each TCM in the TCM category group with the name of the TCM category group to obtain abstract prescription data in groups. It should be noted that the optimal TCM prescription subsequently output by the optimal TCM prescription recommendation model is the name of the TCM category group.

[0074] The distance is converted into an n×n distance matrix D:

[0075]

[0076] Where D[j,k]=d(q j ,q k ), D[j,k] represents the element in the jth row and kth column of the matrix D. d(q j ,q k ) is used to calculate the j-th point q j and the kth point q k The distance between them.

[0077] In one embodiment, all the names of Chinese medicines in the Chinese medicine prescriptions are analyzed and clustered based on the above-mentioned distance matrix to obtain abstract prescription data, so that the Chinese medicine prescriptions are structured on the basis of compliance with the compatibility rules and converted into a format that meets the input requirements of the large model, including data cleaning, feature encoding, serialization processing and other steps to ensure data quality and improve model training efficiency and fine-tuning effects, thereby improving the recommendation efficiency and personalization level of Chinese medicine prescriptions.

[0078] See also Figure 2 As shown, step 106 and step 107 may be further included after step 101.

[0079] In step 106, a TCM knowledge base is constructed based on the TCM syndrome type information in the disease information.

[0080] This step 106 can be processed in parallel with step 102 and step 103, or step 102 and step 103 can be executed first and then step 106, or step 106 can be executed first and then step 102 and step 103.

[0081] Optionally, the disease information includes TCM syndrome type information, which includes syndrome type, syndrome type category, syndrome type characteristics and syndrome type causes, as well as the relationship between various entities in the TCM syndrome type information.

[0082] Step 106 may include: constructing a knowledge graph based on syndrome types, syndrome type categories, syndrome type characteristics, syndrome type causes, and the relationship between entities in traditional Chinese medicine syndrome type information; and establishing a traditional Chinese medicine knowledge base based on the knowledge graph.

[0083] Syndrome type, syndrome type category, syndrome type characteristics, and syndrome type causes are regarded as entities, each entity is regarded as a node, and the relationship between entities is regarded as an edge. Among them, syndrome type represents the general term for all syndrome types, which can be understood as a general term; syndrome type category represents the name of all syndrome type categories, for example, syndrome type categories can include dampness and turbidity blocking syndrome, liver depression and qi stagnation syndrome, blood deficiency syndrome, etc.; syndrome type characteristics represent the specific clinical manifestations and characteristics of the syndrome, for example, the manifestations of blood deficiency syndrome can include dizziness, pale complexion, etc.; syndrome type causes represent the causes that lead to the formation of the syndrome. Among them, the relationship between each entity can include: inclusion relationship, characteristic relationship, and causal relationship. From syndrome type to syndrome type category, it is represented by inclusion relationship; from syndrome type category to syndrome type characteristics, it is represented by characteristic relationship; from syndrome type category to syndrome type cause, it is represented by causal relationship. For example Figure 3 The example diagram of entities and relationships shown in the figure, n1 represents syndrome type, n2 represents syndrome type category, syndrome type category can contain multiple, respectively using n 2-1, ,n 2-2, ,n 2-3, ,…, n3 represents syndrome characteristics, n4 represents syndrome causes, each syndrome category includes corresponding syndrome characteristics and syndrome causes, syndrome characteristics are represented by n 3-1, ,n 3-2, ,n 3-3, ,… indicates that the syndrome types and causes are represented by n 4-1, ,n 4-2, ,n 4-3, ,…express.

[0084] In one embodiment, a knowledge graph is constructed based on syndrome types, syndrome type categories, syndrome type characteristics, syndrome type causes, and the relationship between entities in traditional Chinese medicine syndrome type information, including: taking syndrome types, syndrome type categories, syndrome type characteristics, and syndrome type causes as nodes, and extracting any two nodes as entity pairs; obtaining the relationship between entity pairs, and taking the entity pairs and the relationship between entity pairs as the basic units of the knowledge graph, the basic unit being a triple; and combining all triples to construct a knowledge graph.

[0085] A triple can be represented as (h, r, t), where h represents the head entity, r represents the relationship between entities, and t represents the tail entity. For example, h can be n1, t can be n2, and r can be r1, resulting in a triple represented as (n1, r1, n2). Triples can also be represented as (n2, r2, n3) or (n2, r3, n4). For example, (syndrome type, inclusion relationship, dampness and turbidity obstruction syndrome) indicates that dampness and turbidity obstruction syndrome is a syndrome type.

[0086] Optional, such as Figure 4 The entity-relationship-entity semantic relationship diagram shown can also clearly represent the relationship between entity pairs: from syndrome type to syndrome type category, it is represented by the inclusion relationship; from syndrome type category to syndrome type feature, it is represented by the feature relationship; from syndrome type category to syndrome type cause, it is represented by the causal relationship.

[0087] It should be noted that, from Figure 3 or Figure 4 It can be seen that the subscript of the letter representing the head entity is smaller than the subscript of the letter representing the tail entity. Therefore, when extracting entity pairs, they can be extracted in the order of syndrome type → syndrome type category → syndrome type feature → syndrome type cause, so that the subscript of the head entity is smaller than that of the tail entity, and there is a real relationship between the head entity and the tail entity.

[0088] Optionally, when combining all triples to construct a knowledge graph, it is necessary to delete duplicate triples or triples with no relationship between the head entity and the tail entity. The resulting knowledge graph is a clear, structured knowledge graph that can represent TCM syndrome types and their complex relationships with syndrome characteristics and syndrome causes, thereby achieving the effect of using triples to construct a knowledge graph and further realizing the construction of a TCM knowledge base. For example, when the triple is (dampness and turbidity blocking syndrome, empty, syndrome type), it can be considered that there is no relationship between the head entity and the tail entity.

[0089] In one embodiment, after establishing a TCM knowledge base based on the knowledge graph, the input data, i.e., personal information and disease information in the first data, can also be structured to enhance the ability of the TCM prescription recommendation model to follow the task objectives, help the model to more clearly identify the task type, reduce ambiguity, and guide the model to focus on key features. Figure 2 As shown, after establishing the TCM knowledge base according to the knowledge graph, steps 108 and 109 are also included.

[0090] In step 108, the nodes corresponding to the entities in each triple are extracted as prompt words; the prompt words are connected into a first character string according to preset rules; and a second character string corresponding to the input data is constructed based on the personal information in the first data, the disease description information in the disease information, and the first character string.

[0091] For each TCM syndrome u4 in the first data set, we extract the nodes corresponding to the syndrome characteristics and syndrome causes from the knowledge graph based on the triple relationship between the entities and relationships described above, and obtain prompt words to form the prompt word set H(u4). By emphasizing core information through prompt words, the TCM prescription recommendation model can more accurately extract features, thus achieving an update and enhancement of the dataset.

[0092] The elements in the prompt word set H(u4) are connected into a first string according to a preset rule. For example, the first string can be represented by Sprompt(u4). For example, the specific format can be:

[0093] Sprompt(u4)=concat("Syndrome name is "⊕u4⊕", syndrome cause is "⊕cause(u4)⊕",

[0094] The syndrome characteristics are expressed as "⊕features(u4)⊕". ").

[0095] Where ⊕ represents a string concatenation operation. cause(u4) and features(u4) represent the syndrome cause and syndrome features of syndrome u4, respectively.

[0096] In one embodiment, in order to simplify the input representation and enhance the learning of the association between variables, we merge the key variables u1 (gender), u2 (age), u3 (disease), and u4 (syndrome type) in the first data into a long string, namely the second string. For example, the second string can be represented by Svars, and each variable is separated by a space character: Svars = u1⊕""⊕u2⊕""⊕u3⊕""⊕u4. It should be noted that there is a space between the two quotation marks in the expression of Svars.

[0097] It should be noted that the final input data I is composed of the variable associated string Svars and the first string group Sprompt(u4), for example, the specific form is: I=concat(Svars","Sprompt(u4)).

[0098] In step 109 , predefined control markers are used to explicitly demarcate the boundaries of system instructions, input data, and model responses to form closed context blocks.

[0099] To enhance the TCM prescription recommendation model's ability to adhere to task objectives and ensure controllable input formats, this embodiment uses an instruction encoding framework based on role-task-input separation. This framework explicitly demarcates the boundaries between system instructions, user input, and model responses through predefined control markers, forming closed context blocks. Predefined control markers include <|im_start|> and <|im_end|>.

[0100] For example, system instructions embed task descriptions in the format of <|im_start|>system\n{Task Description}\n<|im_end|>. For example, "Please give a prescription based on the following personal and disease information:" specifies the type of operation that the model needs to perform.

[0101] User input data is dynamically inserted into the data to be processed through <|im_start|>user\n{User Input Text}\n<|im_end|>. For example, the user can input information such as age and disease.

[0102] During the tokenization phase, the model-specific special tokens automatically inserted by the word segmenter are disabled. Instead, the structured instruction sequence is converted into a token ID sequence that can be processed by the model by manually concatenating control tags and text content.

[0103] In steps 108 and 109, the TCM syndrome types extracted from the TCM knowledge base can be integrated into the original data (e.g., the first data) as prompt words and the input data can be structured. The prompt words can help the TCM prescription recommendation model to more clearly identify the task type and reduce ambiguity by explicitly describing the task objectives, thereby guiding the TCM prescription recommendation model to focus on key features. By emphasizing the required core information through prompt words, the TCM prescription recommendation model can extract features more accurately, thereby also achieving the update and enhancement of the first data.

[0104] In step 104, the Chinese medicine prescription data is integrated into the first data to obtain second data, and the second data is used to train a Chinese medicine prescription recommendation model to obtain an optimal Chinese medicine prescription recommendation model.

[0105] In step 107, the traditional Chinese medicine knowledge base is integrated into the second data.

[0106] The initial TCM prescription data obtained in this way, i.e., the first data, includes not only the original data, but also the processed TCM prescription data and the data of the TCM knowledge base, which makes it possible to improve the accuracy of TCM prescription recommendation of the TCM prescription recommendation model. That is, when making TCM prescription recommendations, more accurate TCM prescriptions can be provided based on the real-time data input by the user, which significantly improves the efficiency and personalization level of TCM prescription recommendations, and provides strong support for the research on the inheritance of famous doctors' experience and TCM auxiliary medical care.

[0107] In one embodiment, in order to better simulate (u1, u2, u3, ... u m ) and {v1,v2,…,v n}, using the Qwen2 model and the optimized and improved AITCM-PRM model to recommend Traditional Chinese Medicine prescriptions. The AITCM-PRM model is fine-tuned using LoRA (Low-Rank Adaptation) technology with different secondary data. LoRA reduces resource requirements by fine-tuning a small number of parameters and quickly adapts the model to specific tasks or domains, reducing fine-tuning time and accelerating model optimization for processing specialized terminology and documents.

[0108] In order to make the dataset more standardized and fully adapt to the model, and to add some TCM professional knowledge to TCM syndrome types, we processed the dataset as follows. That is, for a given TCM prescription dataset Htrain, the final second data, we are committed to using the language model to maximize the estimation of conditional probability, that is Where U is (u1,u2,u3,…u m ), V is {v1,v2,…,v n E is the mathematical expectation, which is the probability of each possible outcome in the experiment multiplied by the sum of its outcomes. i is the name of the i-th Chinese medicine in the Chinese medicine prescription, and a is the number of all types of Chinese medicine names in all prescriptions. <i represents all the names of Chinese medicines with index less than i. The value range of i is i∈(1,2,…a). θ is the model parameter. P(q i |q <i ) represents the conditional probability, that is, in q <i Under the conditions that have already occurred, q i Probability of occurrence.

[0109] By using LoRA technology to fine-tune the traditional Chinese medicine prescription recommendation model in a parameter-efficient manner, the optimal traditional Chinese medicine prescription recommendation model is obtained. LoRA implicitly modifies the parameters of the original traditional Chinese medicine prescription recommendation model by introducing a low-rank matrix, fine-tuning the target modules such as query projection (q_proj), key projection (k_proj), value projection (v_proj), output projection (o_proj), gated projection (gate_proj), up projection (up_proj), and down projection (down_proj) to capture the characteristics of specific tasks without destroying the original general knowledge of the model.

[0110] LoRA adds two low-rank matrices A and B and modifies the original parameter matrix W in the following way: W'=W+A*B ∧ T / lora_alpha.

[0111] Here A and B are learnable low-rank matrices that contain far fewer parameters than the original parameter matrix W, which makes LoRA a parameter-efficient method. ∧ T is the transposed matrix of B, and lora_alpha is the scaling factor used to adjust the scale.

[0112] Low-rank matrices A and B are initialized and learned for each target module to minimize the loss function on a specific task. The target module will receive low-rank updates, which aim to improve the performance of the TCM prescription recommendation model on a specific task by introducing additional parameters and maximizing the expected conditional probability by modifying the parameters of the original TCM prescription recommendation model.

[0113] CAT{v1,v2,…,v n} is represented as y, which means {v1, v2, ..., v n} are connected in sequence as new elements. Here, the update amount of the given low-rank matrix is ​​expressed as Δθ, and our goal is to maximize the expected conditional probability: Among them, y <j represents the set of tokens in y from index 1 to index j-1, and |y| represents the number of tokens in y.

[0114] The traditional Chinese medicine prescription recommendation model is trained through LoRA technology to obtain the optimal traditional Chinese medicine prescription recommendation model.

[0115] In step 105, real-time data is acquired and input into an optimal Chinese medicine prescription recommendation model to obtain an optimal Chinese medicine prescription. The real-time data includes real-time personal information and real-time disease information.

[0116] The real-time data here refers to real-time input data.

[0117] The method for recommending Chinese medicine prescriptions provided in this embodiment obtains first data, which includes personal information, disease information and a set of Chinese medicine prescriptions; processes the set of Chinese medicine prescriptions in the first data, determines the association between any two Chinese medicine names in the set of Chinese medicine prescriptions, and then processes all Chinese medicine prescriptions in the set of Chinese medicine prescriptions according to the association to obtain Chinese medicine prescription data; integrates the Chinese medicine prescription data into the first data to obtain second data, and uses the second data to train a Chinese medicine prescription recommendation model to obtain an optimal Chinese medicine prescription recommendation model; obtains real-time input data, and inputs the real-time input data into the optimal Chinese medicine prescription recommendation model to obtain an optimal Chinese medicine prescription, wherein the real-time input data includes real-time personal information and real-time disease information, thereby realizing personalized Chinese medicine prescription recommendation through personal information and disease information. In addition, through the analysis of Chinese medicine prescriptions, the potential association characteristics between Chinese medicine names are explored, the Chinese medicine compatibility rules are revealed, and the personalization level of Chinese medicine prescription recommendations is improved.

[0118] In this embodiment, by processing the TCM syndrome type information in the first data, a knowledge graph is constructed, and a TCM knowledge base containing professional terms related to TCM syndrome types is further established. The TCM syndrome types and their corresponding related clinical manifestations and key information on the causes of the syndrome types are extracted and deeply integrated into the data set and injected into the TCM prescription recommendation model. Combined with fine-tuning technology, the TCM prescription recommendation model can better capture the complex relationship between diseases and TCM prescriptions, further improving the level of personalization of TCM prescription recommendations.

[0119] It should be understood that the size of the serial numbers of each step in the above embodiment does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0120] The following is an embodiment of the device of the present invention. For details not described in detail, please refer to the corresponding method embodiment described above.

[0121] Figure 5 This is a functional block diagram of a device for recommending Chinese medicine prescriptions provided by an embodiment of the present invention, referring to Figure 5 The device for recommending Chinese medicine prescriptions includes: an acquisition module 501, a determination module 502, a processing module 503 and a training module 504, wherein:

[0122] An acquisition module 501 is configured to acquire first data, wherein the first data includes personal information, disease information, and a set of traditional Chinese medicine prescriptions;

[0123] Determining module 502, for determining the association relationship between any two Chinese medicine names in the Chinese medicine prescription set;

[0124] The processing module 503 is used to process all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data;

[0125] The processing module 503 is further used to integrate the traditional Chinese medicine prescription data into the first data to obtain the second data;

[0126] A training module 504 is configured to perform model training on the traditional Chinese medicine prescription recommendation model using the second data to obtain an optimal traditional Chinese medicine prescription recommendation model;

[0127] The acquisition module 501 is also used to obtain real-time input data;

[0128] The processing module 503 is further used to input the real-time input data into the optimal Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription. The real-time input data includes real-time personal information and real-time disease information.

[0129] In one embodiment, when determining the association relationship between any two Chinese medicine names in the Chinese medicine prescription set, the determining module 502 is configured to:

[0130] Get the names of all Chinese medicines in the Chinese medicine prescriptions in the Chinese medicine prescription collection;

[0131] Group any two Chinese medicine names from all Chinese medicine names into one group, and determine the number of times the two Chinese medicine names in each group appear simultaneously in all Chinese medicine prescriptions;

[0132] The distance between the two Chinese medicine names in each group was determined based on the number of times.

[0133] In one embodiment, when determining the distance between two Chinese medicine names in each group according to the number of times, the determining module 502 is configured to:

[0134] According to the total number and frequency of all TCM prescriptions, the distance between the two TCM names in each group was determined.

[0135] In one embodiment, the processing module 503 processes all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data, including:

[0136] Convert all determined distances into a distance matrix;

[0137] Performing cluster analysis on all the Chinese medicines in all the Chinese medicine prescriptions according to the distance matrix to obtain Chinese medicine category groups;

[0138] The name of each Chinese medicine in the Chinese medicine category group is replaced with the name of the Chinese medicine category group to which it belongs, and abstract prescription data based on groups is obtained.

[0139] In one embodiment, the disease information includes TCM syndrome type information, which includes syndrome type, syndrome type category, syndrome type characteristics, syndrome type cause, and the relationship between various entities in the TCM syndrome type information;

[0140] After the acquisition module 501 acquires the first data, the processing module 503 is further configured to:

[0141] Construct a knowledge graph based on syndrome types, syndrome categories, syndrome characteristics, syndrome causes, and the relationships between entities in TCM syndrome information;

[0142] Establish a TCM knowledge base based on the knowledge graph;

[0143] Integrate the traditional Chinese medicine knowledge base into the secondary data.

[0144] In one embodiment, when constructing a knowledge graph based on syndrome types, syndrome type categories, syndrome type characteristics, syndrome type causes, and the relationships between entities in TCM syndrome type information, the processing module 503 is used to:

[0145] The syndrome type, syndrome category, syndrome characteristics, and syndrome cause are respectively regarded as nodes, and any two nodes are extracted as entity pairs;

[0146] Obtain the relationship between entity pairs and use the entity pairs and the relationship between them as the basic unit of the knowledge graph. The basic unit is a triple.

[0147] All triples are combined to construct a knowledge graph.

[0148] In one embodiment, after the processing module 503 establishes a TCM knowledge base based on the knowledge graph, it is further configured to:

[0149] Extract the nodes corresponding to the entities in each triple as prompt words;

[0150] Connect the prompt words into a first character string according to a preset rule;

[0151] Constructing a second character string corresponding to the input data based on the personal information in the first data, the disease description information in the disease information, and the first character string;

[0152] Predefined control markers are used to explicitly demarcate the boundaries of system instructions, input data, and model responses to form closed context blocks.

[0153] The device for recommending Chinese medicine prescriptions provided in this embodiment obtains first data through an acquisition module, and the first data includes personal information, disease information, and a set of Chinese medicine prescriptions; and processes the set of Chinese medicine prescriptions in the first data, and the determination module determines the association between any two Chinese medicine names in the set of Chinese medicine prescriptions, and the processing module then processes all Chinese medicine prescriptions in the set of Chinese medicine prescriptions according to the association to obtain Chinese medicine prescription data; the processing module integrates the Chinese medicine prescription data into the first data to obtain second data, and the training module uses the second data to train the Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription recommendation model; thereby achieving the purpose of using the acquired real-time input data and inputting it into the optimal Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription, and the real-time input data includes real-time personal information and real-time disease information. In addition, through the analysis of Chinese medicine prescriptions, the potential association characteristics between Chinese medicine names are explored, the rules of Chinese medicine compatibility are revealed, and the personalization level of Chinese medicine prescription recommendations is improved.

[0154] In this embodiment, the processing module processes the TCM syndrome type information in the first data to construct a knowledge graph, and further establishes a TCM knowledge base containing professional terms related to TCM syndrome types. The TCM syndrome types and their corresponding related clinical manifestations and key information on the causes of the syndrome types are extracted and deeply integrated into the data set, and injected into the TCM prescription recommendation model. Combined with fine-tuning technology, the TCM prescription recommendation model can better capture the complex relationship between diseases and TCM prescriptions, further improving the level of personalization of TCM prescription recommendations.

[0155] Figure 6 : is a functional block diagram of an electronic device provided by an embodiment of the present invention. Figure 6 As shown, the electronic device 6 of this embodiment includes: a processor 600 and a memory 601, wherein the memory 601 stores a computer program 602 that can be run on the processor 600. When the processor 600 executes the computer program 602, the steps in the above-mentioned methods and embodiments of recommending Chinese medicine prescriptions are implemented, for example Figure 1Steps 101 to 105 are shown.

[0156] Illustratively, the computer program 602 may be divided into one or more modules / units, and the one or more modules / units are stored in the memory 601 and executed by the processor 600 to implement the present invention.

[0157] The electronic device 6 can be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 6 can include, but is not limited to, a processor 600 and a memory 601. Those skilled in the art will understand that Figure 6 It is only an example of the electronic device 6 and does not constitute a limitation of the electronic device 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 6 may also include input and output devices, network access devices, buses, etc.

[0158] The processor 600 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0159] The memory 601 may be an internal storage unit of the electronic device 6, such as a hard disk or memory of the electronic device 6. The memory 601 may also be an external storage device of the electronic device 6, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 6. Furthermore, the memory 601 may include both an internal storage unit of the electronic device 6 and an external storage device. The memory 601 is used to store the computer program 602 and other programs and data required by the electronic device 6. The memory 601 may also be used to temporarily store data that has been output or is about to be output.

[0160] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the implementation method can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method implementation method, and will not be repeated here.

[0161] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0162] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0163] In the embodiments provided by the present invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0164] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of this embodiment.

[0165] In addition, the functional units in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0166] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned implementation method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various methods and device implementation methods. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0167] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for recommending a Chinese medicine prescription, characterized in that: include: Acquiring first data, where the first data includes personal information, disease information, and a set of traditional Chinese medicine prescriptions; Determine the association relationship between any two Chinese medicine names in the Chinese medicine prescription set; Processing all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data; Integrating the traditional Chinese medicine prescription data into the first data to obtain second data, and using the second data to perform model training on a traditional Chinese medicine prescription recommendation model to obtain an optimal traditional Chinese medicine prescription recommendation model; Real-time input data is acquired and input into the optimal Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription, wherein the real-time input data includes real-time personal information and real-time disease information.

2. The method for recommending a Chinese medicine prescription according to claim 1, wherein: Determining the association relationship between any two Chinese medicine names in the Chinese medicine prescription set includes: Obtain the names of the traditional Chinese medicines in all the traditional Chinese medicine prescriptions in the traditional Chinese medicine prescription set; Group any two Chinese medicine names from all Chinese medicine names into one group, and determine the number of times the two Chinese medicine names in each group appear simultaneously in all Chinese medicine prescriptions; The distance between the two Chinese medicine names in each group is determined according to the number of times.

3. The method for recommending Chinese medicine prescriptions according to claim 2, wherein: The distance between the two Chinese medicine names in each group is determined according to the number of times, including: According to the total number of all TCM prescriptions and the number of times, the distance between two TCM names in each group was determined.

4. The method for recommending a Chinese medicine prescription according to claim 3, wherein: All the Chinese medicine prescriptions in the Chinese medicine prescription set are processed according to the association relationship to obtain Chinese medicine prescription data, including: Convert all determined distances into a distance matrix; Performing cluster analysis on all the Chinese medicines in all the Chinese medicine prescriptions according to the distance matrix to obtain Chinese medicine category groups; The name of each Chinese medicine in the Chinese medicine category group is replaced with the name of the Chinese medicine category group to which it belongs, so as to obtain abstract prescription data in groups.

5. The method for recommending a traditional Chinese medicine prescription according to any one of claims 1 to 4, characterized in that: The disease information includes TCM syndrome type information, which includes syndrome type, syndrome type category, syndrome type characteristics, syndrome type cause and the relationship between each entity in the TCM syndrome type information; After obtaining the first data, the method further includes: Constructing a knowledge graph based on the syndrome type, the syndrome type category, the syndrome type characteristics, the syndrome type cause, and the relationship between entities in the TCM syndrome type information; Establishing a traditional Chinese medicine knowledge base based on the knowledge graph; Integrate the traditional Chinese medicine knowledge base into the second data.

6. The method for recommending a traditional Chinese medicine prescription according to claim 5, characterized in that: The constructing of a knowledge graph based on the syndrome type, the syndrome type category, the syndrome type characteristics, the syndrome type cause, and the relationship between entities in the TCM syndrome type information includes: The syndrome type, syndrome category, syndrome characteristics, and syndrome cause are respectively taken as nodes, and any two nodes are extracted as entity pairs; Obtaining the relationship between the entity pairs, and using the entity pairs and the relationship between the entity pairs as the basic unit of the knowledge graph, where the basic unit is a triple; All triples are combined to construct the knowledge graph.

7. The method for recommending a traditional Chinese medicine prescription according to claim 6, characterized in that: After establishing a TCM knowledge base based on the knowledge graph, the method further includes: Extract the nodes corresponding to the entities in each triple as prompt words; Connecting the prompt words into a first character string according to a preset rule; constructing a second character string corresponding to the input data based on the personal information in the first data, the disease description information in the disease information, and the first character string; Predefined control markers are used to explicitly demarcate the boundaries of system instructions, input data, and model responses to form closed context blocks.

8. A device for recommending Chinese medicine prescriptions, characterized in that: A method for implementing a Chinese medicine prescription recommendation according to any one of claims 1 to 7, wherein the Chinese medicine prescription recommendation device comprises: An acquisition module, configured to acquire first data, wherein the first data includes personal information, disease information, and a set of traditional Chinese medicine prescriptions; A determination module, configured to determine an association relationship between any two Chinese medicine names in the Chinese medicine prescription set; A processing module, configured to process all the Chinese medicine prescriptions in the Chinese medicine prescription set according to the association relationship to obtain Chinese medicine prescription data; The processing module is further configured to integrate the traditional Chinese medicine prescription data into the first data to obtain second data; A training module, configured to perform model training on a traditional Chinese medicine prescription recommendation model using the second data to obtain an optimal traditional Chinese medicine prescription recommendation model; The acquisition module is further used to acquire real-time input data; The processing module is further used to input the real-time input data into the optimal Chinese medicine prescription recommendation model to obtain the optimal Chinese medicine prescription, and the real-time input data includes real-time personal information and real-time disease information.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the steps of the method for recommending a traditional Chinese medicine prescription as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for recommending a traditional Chinese medicine prescription as described in any one of claims 1 to 7 are implemented.

Images