Simplified method of self-test table and method of judging physical constitution
By constructing a graph network model and selecting attention scores, the problem of self-test table of traditional Chinese medicine physical fitness is simplified, and the problem of low judgment efficiency in the existing technology is solved, and the efficiency and accuracy of physical fitness judgment is improved.
Patent Information
- Application Number
- CN202411469692.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-01-14
- Estimated Expiration
- 2044-10-21
AI Technical Summary
There are many questions in the existing TCM physical fitness self-test table, which leads to low judgment efficiency and long time for individuals to answer, which can easily lead to decreased attention and inaccurate answers.
By constructing a graph network model, the questions in the self-test table in the Chinese medicine physical fitness classification and judgment based on the attention score are deleted and selected, and the self-test table is generated simplified, the number of questions is reduced, and the answer efficiency and accuracy are improved.
The number of questions in the self-test table is simplified, the answer time is shortened, the efficiency and accuracy of physical fitness judgment are improved, and attention reduction and inaccurate answers are avoided due to long-term filling.
Smart Images
Figure CN119005148B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of medical auxiliary algorithms, and particularly relates to a self-test table simplification method and a physical constitution judgment method. Background Art
[0002] The existing self-testing method for judging TCM constitution generally judges the intensity of the symptoms corresponding to the questions in the TCM constitution classification and determination self-test form based on the individual's answer to the questions (none, rarely, sometimes, often, always) to judge the TCM constitution of the individual. However, the existing TCM constitution classification and determination self-test form has 61 questions, which is a large number of questions. It takes a long time for individuals to answer the questions in the TCM constitution classification and determination self-test form, resulting in low efficiency in constitution judgment.
[0003] Therefore, the existing technology needs to be improved and developed. Summary of the invention
[0004] The purpose of this application is to provide a method for simplifying a self-test table and a method for determining constitution, aiming to solve the problem of low efficiency in determining TCM constitution based on existing self-test methods.
[0005] In a first aspect, the present application provides a method for simplifying a self-test table, which is used to simplify a self-test table for TCM constitution classification and determination, and comprises the following steps:
[0006] S1. Obtain the first individual information and the completed TCM constitution classification and determination self-test form and constitution label corresponding to the first individual information to establish a first data set;
[0007] S2. Generate a first individual number and a first symptom vector according to the first individual information in the first data set and the TCM constitution classification and determination self-test table;
[0008] S3. forming a first graph network with the first individual number and the first symptom vector as nodes, constructing a first model, and training the first model based on the first graph network and the corresponding constitution label;
[0009] S4. Outputting the attention score of each symptom node in the first graph network based on the first model, wherein each symptom node in the first graph network is each first symptom vector as a node of the first graph network;
[0010] S5. Perform threshold screening on the questions in the TCM constitution classification and determination self-test form according to the attention score to obtain a simplified self-test form.
[0011] The self-test form simplification method of the present application selects questions in the self-test form for TCM constitution classification and determination according to the attention score, and can reduce the questions in the self-test form to simplify the self-test form without greatly affecting the accuracy of subsequent constitution determination, and can make the time required for an individual to answer questions in the simplified self-test form shorter than the time required to answer questions in the self-test form, thereby improving the efficiency of constitution determination; and the present application reduces the questions in the self-test form, and can avoid inaccurate answers to questions due to a decrease in individual attention or resistance during the long filling out or questioning process, thereby improving the accuracy of constitution determination.
[0012] Furthermore, in step S2, the process of generating a first individual number and a first symptom vector according to the first individual information in the first data set and the TCM constitution classification and determination self-test table includes:
[0013] S21. Perform one-hot encoding on the first individual information and the first symptom information in the TCM constitution classification and determination self-test table to generate a first individual vector and a first symptom vector, and use the first individual vector as the first individual number.
[0014] In a second aspect, the present application provides a method for determining physical condition, comprising the following steps:
[0015] A1. Obtain a simplified self-test form filled out by a new individual, where the simplified self-test form is obtained based on any of the above self-test form simplification methods;
[0016] A2. Generate a physical fitness judgment result based on the filled-in simplified self-test form based on the pre-trained second model.
[0017] The constitution judgment method of the present application selects questions in the self-test form for classification and determination of TCM constitution according to the attention score, and can reduce the questions in the self-test form to simplify the self-test form without greatly affecting the accuracy of subsequent constitution judgment, and can make the time required for individuals to answer questions in the simplified self-test form shorter than the time required to answer questions in the self-test form, thereby improving the efficiency of constitution judgment; and the present application reduces the questions in the self-test form, and can avoid the individual's attention being reduced or the resistance being generated during the long filling or inquiry process, resulting in inaccurate answers to the questions, thereby improving the accuracy of constitution judgment; and the present application judges the TCM constitution of an individual based on the second model, which can avoid the process of calculating the score, avoid applying an inappropriate score calculation method to the simplified self-test form, and can make an accurate judgment of the TCM constitution of the individual based on the simplified self-test form.
[0018] Furthermore, in step A2, the second model is a GraphSAGE model, and the process of generating a physical fitness judgment result according to a simplified self-test table based on the pre-trained second model includes:
[0019] A21. Obtaining second individual information and a simplified self-test table corresponding to the second individual information, generating a second individual number and a second symptom vector according to the second individual information and the simplified self-test table, and generating a second graph network according to the second individual number and the second symptom vector;
[0020] A22. Input the second graph network into the GraphSAGE model, and the GraphSAGE model performs feature aggregation based on the second graph network until a physical constitution judgment result is generated.
[0021] Furthermore, in step A21, the process of generating a second graph network according to the second individual node and the second symptom vector includes:
[0022] A211. The second individual number and the second symptom vector are used as nodes to form a second graph network;
[0023] A212. Embedding the symptom node of the second graph network, wherein the symptom node of the second graph network is a second symptom vector as a node of the second graph network;
[0024] In step A22, the process of the GraphSAGE model performing feature aggregation based on the second graph network until a physical fitness judgment result is generated includes:
[0025] A221. The GraphSAGE model performs feature aggregation based on the symptom nodes after embedding in the second graph network until a constitution judgment result is generated.
[0026] In this embodiment, the physical constitution judgment method of the present application embeds the symptom nodes of the second graph network, and enables the GraphSAGE model to perform feature aggregation based on the symptom nodes after the embedding processing in the second graph network. During the feature aggregation process, the features of each node after the information is updated can be aggregated to the corresponding nodes whose information has not been updated, thereby further improving the accuracy of physical constitution judgment based on the GraphSAGE model.
[0027] Furthermore, in step A221, the process in which the GraphSAGE model performs feature aggregation based on the symptom nodes after embedding processing in the second graph network until a constitution judgment result is generated includes:
[0028] A2211. An individual node in the second graph network is taken as the central node, the symptom node corresponding thereto is taken as the one-hop neighbor node, and the individual nodes other than the central node corresponding to the one-hop neighbor node are taken as the two-hop neighbor nodes;
[0029] A2212. Aggregate the features of two-hop neighbor nodes to one-hop neighbor nodes;
[0030] A2213. Aggregate the features of one-hop neighbor nodes to the central node;
[0031] A2214. An individual node in the second graph network other than the central node is used as a new central node, the symptom node corresponding thereto is used as a new one-hop neighbor node, and the individual nodes other than the central node and the new central node corresponding to the new one-hop neighbor node are used as new two-hop neighbor nodes;
[0032] A2215. Repeat steps A2212-A2214 until feature aggregation is completed for all individual nodes;
[0033] A2216. Treat all individual nodes as if no feature aggregation has been performed and repeat steps A2211-A2215 until a physical constitution judgment result is generated.
[0034] In this embodiment, the physical constitution judgment method of the present application aggregates the features of two-hop neighbor nodes to one-hop neighbor nodes, and aggregates the features of one-hop neighbor nodes to the central node. During the feature aggregation process, the features of the symptom nodes after the information is updated can be aggregated to the corresponding central nodes whose information has not been updated, and the features of the individual nodes after the information is updated can be aggregated to the individual nodes corresponding to them having at least one identical symptom vector, thereby further improving the accuracy of physical constitution judgment based on the GraphSAGE model.
[0035] Furthermore, the training process of the second model includes:
[0036] B1. Filter the completed TCM constitution classification and determination self-test form according to the simplified self-test form, obtain the completed simplified self-test form, establish a second data set according to the completed simplified self-test form and the first body information and constitution label corresponding to the completed simplified self-test form, use the first body number and the first symptom vector corresponding to each first body information in the second data set as input constitution samples, use each constitution label in the second data set as output constitution sample, divide the constitution sample set consisting of multiple corresponding input constitution samples and output constitution samples into a training set and a test set, and the training set and the test set each contain multiple corresponding input constitution samples and output constitution samples;
[0037] B2. Train the second model based on the training set;
[0038] B3. Test the second model based on the test set.
[0039] As can be seen from the above, the present application provides a method for simplifying a self-test form and a method for determining a physical constitution, wherein the method for simplifying the self-test form provided by the present application selects the questions in the self-test form for classification and determination of TCM physical constitution according to the attention score, and can reduce the questions in the self-test form to simplify the self-test form without greatly affecting the accuracy of subsequent physical constitution determination, and can make the time required for an individual to answer the questions in the simplified self-test form shorter than the time required to answer the questions in the self-test form, thereby improving the efficiency of physical constitution determination; and the present application reduces the questions in the self-test form, and can avoid the individual's attention being reduced or resistance arising during the long filling out or questioning process, resulting in inaccurate answers to the questions, thereby improving the accuracy of physical constitution determination.
[0040] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or be understood by practicing the embodiments of the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the written description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 A flowchart of a method for simplifying a self-test table provided in an embodiment of the present application.
[0042] Figure 2 This is a flow chart of the method for determining physical condition provided in an embodiment of the present application. DETAILED DESCRIPTION
[0043] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0044] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.
[0045] First, as Figure 1 As shown, the present application provides a method for simplifying a self-test table, which is used to simplify a self-test table for TCM constitution classification and determination, and comprises the following steps:
[0046] S1. Obtain the first individual information and the completed TCM constitution classification and determination self-test form and constitution label corresponding to the first individual information to establish a first data set;
[0047] S2. Generate a first individual number and a first symptom vector according to the first individual information in the first data set and the TCM constitution classification and determination self-test table;
[0048] S3. forming a first graph network with the first individual number and the first symptom vector as nodes, constructing a first model, and training the first model based on the first graph network and the corresponding constitution label;
[0049] S4. Outputting the attention score of each symptom node in the first graph network based on the first model, wherein each symptom node in the first graph network is each first symptom vector as a node of the first graph network;
[0050] S5. Perform threshold screening on the questions in the TCM constitution classification and determination self-test form according to the attention score to obtain a simplified self-test form.
[0051] Table 1 Correspondence between questions and symptoms in the existing TCM constitution classification and self-assessment table
[0052] Problems in the self-test table for TCM constitution classification and determination Symptoms Problems in the self-test table of TCM constitution classification and determination Symptoms Do you have cold hands and feet? Cold hands and feet Do you have an oily or shiny face or nose? Greasy or shiny face or nose Do you feel cold in your stomach, back, waist and knees? Coldness in the epigastric region, back, waist, or knees Are you prone to acne or boils? Prone to acne or boils Do you feel cold and wear more clothes than others? Wear more clothes than others because you are afraid of cold weather Do you have a bitter taste or bad odor in your mouth? Bitter taste or bad breath Are you less able to tolerate cold (winter cold, summer cold air conditioners, fans, etc.) than the average person? Less able to tolerate cold than the average person (winter cold, summer cold air conditioners, fans, etc.) Do you have sticky and uncomfortable stools and a feeling of incomplete defecation? Sticky and uncomfortable stools with a feeling of being unable to defecate Do you catch colds more easily than other people? Susceptible to colds Do you feel a warm sensation in your urethra when you urinate, or is your urine dark in color? Feeling of heat in the urethra when urinating Dark urine Do you feel uncomfortable or afraid of eating cold food? Feeling uncomfortable when eating or drinking cold things or being afraid of eating or drinking cold things Is your discharge yellow? Yellow vaginal discharge (female) Do you tend to get diarrhea after catching a cold or eating cold food? Diarrhea after catching a cold or eating or drinking something cold Is your scrotal area moist? Wet scrotum area (male) Do you feel heat in your hands and feet? Hot palms and soles Do you have purple bruises on your skin without noticing? Purple bruises may appear on the skin without your knowledge Do you feel heat in your body or face? Feeling of heat on the body and face Do you have tiny red lines on your cheekbones? There are slight red lines on both cheekbones Do you have dry skin or lips? Dry skin or lips Do you have any pain anywhere in your body? Body pain Are your lips redder than average? The color of lips is redder than that of ordinary people Do you have a dull complexion or are you prone to brown spots? Dull complexion or prone to brown spots Do you tend to be constipated or have dry stools? Constipation or dry stools Are you prone to dark circles? Dark circles under eyes Do your cheeks feel flushed or reddish? Flushing or reddish cheeks Are you forgetful? Easy to forget things Do you experience dry eyes? Dry eyes Do you have dark lip color? Dark lip color Do you feel dry mouth and throat and always want to drink water? Dry mouth and throat, want to drink water Do you sneeze even when you don't have a cold? Sneezing even without a cold Do you tire easily? Easy fatigue Do you have a stuffy or runny nose even when you don’t have a cold? I have a stuffy and runny nose even without a cold. Are you prone to shortness of breath (shortness of breath, difficulty catching your breath? Shortness of breath Do you have coughing and wheezing due to seasonal changes, temperature changes, or odors? Coughing and wheezing due to seasonal changes, temperature changes, or odors Do you get easily flustered? Easy to panic Are you prone to allergies? Prone to allergies Do you get dizzy easily or feel faint when standing up? Easy dizziness or lightheadedness when standing up Does your skin break out easily? Skin prone to urticaria, wheals, hives, rashes, wind bumps Do you suffer from insomnia? Insomnia Have you ever had purpura due to allergies? Purpura due to allergies Do you like to be quiet and lazy to talk? I like quietness and I don't want to talk Does your skin turn red and get scratched when you scratch it? The skin turns red and scratches when scratched Is your voice weak? Weak voice Are you feeling depressed? Depressed Do you sweat easily when you are a little active? Sweating easily after a little bit of activity Are you prone to stress and anxiety? Tension and anxiety Do you feel chest tightness or abdominal distension? Chest tightness or abdominal distension Are you sentimental and emotionally fragile? Sentimental and fragile Do you feel physically uncomfortable or down? Feeling uncomfortable or down Are you easily frightened or startled? Easily frightened or startled Do you have a fat and flabby belly? Belly fat and soft Do you have pain in your flank or breast? Pain in the flank or breast Do you have excessive oil secretion on your forehead? Excessive oil secretion in the forehead Do you sigh for no reason? Sighing for no reason Are your upper eyelids swollen than others? Upper eyelid is swollen than others Do you feel like there is something foreign in your throat that you can’t spit out or swallow? There is a foreign body sensation in the throat and it is difficult to spit out or swallow. Do you have a sticky feeling in your mouth? Sticky mouth Are you energetic? Full of energy Do you usually have a lot of phlegm, especially in your throat? Do you always feel that there is phlegm blocking it? Excessive phlegm and a feeling of phlegm blocking the throat Can you adapt to changes in the external natural and social environment? Adapt to changes in the external natural and social environment Do you have a thick, greasy tongue or do you feel like your tongue is coated with a thick coating? Thick and greasy tongue coating
[0053] Specifically, the existing self-test form for classification and determination of TCM constitution (hereinafter referred to as the self-test form) contains multiple questions, which are used to judge the intensity of the symptoms corresponding to the questions in the individual's body according to the individual's answers to the questions in the self-test form (none, rarely, sometimes, often, always), so as to judge which TCM constitution the individual belongs to. The questions in the self-test form and the corresponding symptoms are shown in Table 1. The types of TCM constitution of an individual include balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-damp constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution and special constitution.
[0054] More specifically, in step S1, after the individual completes the self-test form and the individual's constitution is determined based on the self-test form, the first individual information and the completed TCM constitution classification and determination self-test form and constitution label corresponding to the first individual information are used as data to establish a first data set, wherein the first individual information can be the name of the individual filling out the self-test form, or other information that can distinguish different individuals filling out the self-test form; the completed TCM constitution classification and determination self-test form corresponding to the first individual information is a collection of questions in the self-test form and the individual's questions and answers to the questions in the self-test form; and the constitution label is the TCM constitution of the individual corresponding to the first individual information.
[0055] More specifically, in step S2, the first individual number is the first individual information after the numbering, and the single first symptom vector is the vector representation of the intensity of the single symptom on the individual body corresponding to the answer in the self-test form. Since the self-test form has 61 questions, one self-test form has 61 answers, and one individual number corresponds to 61 first symptom vectors.
[0056] More specifically, in step S3, the first model is a model based on the result generated by the graph attention mechanism, wherein the result is a vector representation of the corresponding node in the graph network, and the graph attention mechanism is used to adjust and determine the attention scores of each neighbor node of the node, and the attention scores of each neighbor node can characterize the importance of each neighbor node of the node to the result generated by the first model. In the present application, the first model is used to generate a physical judgment result, that is, a vector representation of an individual node in the first graph network, wherein the individual node in the first graph network is the first body number as a node of the first graph network. The attention mechanism is used to learn and determine the attention scores of each symptom node connected to the individual node, and the attention scores of each symptom node can characterize the importance of each symptom node for generating a physical judgment result. In the first graph network, each individual node corresponds to 61 symptom nodes, and the first model can be trained based on the existing model training method according to the first body number, the first symptom vector and the physical label corresponding to each other.
[0057] More specifically, in step S5, a score threshold is set as needed, a symptom node with an attention score less than the score threshold is determined, the questions in the self-test table corresponding to the node are removed from the self-test table, the remaining questions are retained, and the self-test table with only the remaining questions is used as a simplified self-test table. Since the importance of the physical fitness judgment corresponding to the symptom node with a low attention score is low, the questions in the self-test table are deleted according to the attention score, and the questions in the self-test table can be reduced to simplify the self-test table under the premise of having little impact on the accuracy of subsequent physical fitness judgment, so that the time required for individuals to answer the questions in the simplified self-test table is shorter than the time required to answer the questions in the self-test table, thereby improving the efficiency of physical fitness judgment.
[0058] Table 2 Symptoms corresponding to each symptom node in the first network and the attention score of each symptom node in the embodiment
[0059] Symptom Name Attention score Symptom Name Attention score Cold hands and feet 0.0153 There are slight red lines on both cheekbones 0.013 Coldness in the epigastric region, back, waist, or knees 0.0139 Body pain 0.0053 Wear more clothes than others because you are afraid of cold weather 0.0149 Dull complexion or prone to brown spots 0.0101 Less able to tolerate cold than the average person (winter cold, summer cold air conditioners, fans, etc.) 0.017 Dark circles under eyes 0.0131 Susceptible to colds 0.0172 Easy to forget things 0.011 Feeling uncomfortable when eating or drinking cold things or being afraid of eating or drinking cold things 0.0044 Dark lip color 0.0068 Diarrhea after catching a cold or eating or drinking something cold 0.0148 Sneezing even without a cold 0.0111 Hot palms and soles 0.0097 I have a stuffy and runny nose even without a cold. 0.0071 Feeling of heat on the body and face 0.0155 Coughing and wheezing due to seasonal changes, temperature changes, or odors 0.0119 Dry skin or lips 0.0101 Prone to allergies 0.0139 The color of lips is redder than that of ordinary people 0 Skin prone to urticaria, wheals, hives, rashes, wind bumps 0.0123 Constipation or dry stools 0.0106 Purpura due to allergies 0.0137 Flushing or reddish cheeks 0.0004 The skin turns red and scratches when scratched 0.0134 Dry eyes 0.0146 Full of energy 0.0077 Dry mouth and throat, want to drink water 0.0109 Depressed 0.0005 Sweating easily after a little bit of activity 0.0141 Adapt to changes in the external natural and social environment 0.0091 Easy fatigue 0.0174 Insomnia 0 Shortness of breath 0.0005 Tension and anxiety 0.0121 Easy to panic 0.0105 Sentimental and fragile 0.0121 Easy dizziness or lightheadedness when standing up 0.0132 Easily frightened or startled 0.0121 I like quietness and I don't want to talk 0.0103 Pain in the flank or breast 0.0111 Weak voice 0.0005 Sighing for no reason 0.0123 Chest tightness or abdominal distension 0.0086 There is a foreign body sensation in the throat and it is difficult to spit out or swallow. 0.0085 Feeling uncomfortable or down 0.0119 Greasy or shiny face or nose 0.0004 Belly fat and soft 0.0095 Prone to acne or boils 0.0114 Excessive oil secretion in the forehead 0.0136 Bitter taste or bad breath 0.0146 Upper eyelid is swollen than others 0.0085 Sticky and uncomfortable stools with a feeling of being unable to defecate 0.0161 Sticky mouth 0.0018 Feeling of heat in the urethra when urinating Dark urine 0.0151 Excessive phlegm and a feeling of phlegm blocking the throat 0.0101 Yellow vaginal discharge (female) 0 Thick and greasy tongue coating 0.012 Wet scrotum area (male) 0 Purple bruises may appear on the skin without your knowledge 0.0121
[0060] Table 3 Symptoms corresponding to symptom nodes with attention scores lower than 0.01 in the first network of the embodiment
[0061] The color of lips is redder than that of ordinary people Feeling uncomfortable when eating or drinking cold things or being afraid of eating or drinking cold things Flushing or reddish cheeks Hot palms and soles Shortness of breath Chest tightness or abdominal distension Weak voice Upper eyelid is swollen than others Sticky mouth Body pain Depressed Dark lip color Insomnia I have a stuffy and runny nose even without a cold. Greasy or shiny face or nose Full of energy Yellow vaginal discharge (female) Adapt to changes in the external natural and social environment Wet scrotum area (male) There is a foreign body sensation in the throat and it is difficult to spit out or swallow. Belly fat and soft
[0062] Problems in the self-test table for TCM constitution classification and determination Symptoms Problems in the self-test table of TCM constitution classification and determination Symptoms Do you have cold hands and feet? Cold hands and feet Are you prone to acne or boils? Prone to acne or boils Do you feel cold in your stomach, back, waist and knees? Coldness in the epigastric region, back, waist, or knees Do you have a bitter taste or bad odor in your mouth? Bitter taste or bad breath Do you feel cold and wear more clothes than others? Wear more clothes than others because you are afraid of cold weather Do you have sticky and uncomfortable stools and a feeling of incomplete defecation? Sticky and uncomfortable stools with a feeling of being unable to defecate Are you less able to tolerate cold (winter cold, summer cold air conditioners, fans, etc.) than the average person? Less able to tolerate cold than the average person (winter cold, summer cold air conditioners, fans, etc.) Do you feel a warm sensation in your urethra when you urinate, or is your urine dark in color? Feeling of heat in the urethra when urinating Dark urine Do you catch colds more easily than other people? Susceptible to colds Do you have purple bruises on your skin without noticing? Purple bruises may appear on the skin without your knowledge Do you tend to get diarrhea after catching a cold or eating cold food? Diarrhea after catching a cold or eating or drinking something cold Do you have tiny red lines on your cheekbones? There are slight red lines on both cheekbones Do you feel heat in your body or face? Feeling of heat on the body and face Do you have a dull complexion or are you prone to brown spots? Dull complexion or prone to brown spots Do you have dry skin or lips? Dry skin or lips Are you prone to dark circles? Dark circles under eyes Do you tend to be constipated or have dry stools? Constipation or dry stools Are you forgetful? Easy to forget things Do you experience dry eyes? Dry eyes Do you have a stuffy or runny nose even when you don’t have a cold? I have a stuffy and runny nose even without a cold. Do you feel dry mouth and throat and always want to drink water? Dry mouth and throat, want to drink water Do you have coughing and wheezing due to seasonal changes, temperature changes, or odors? Coughing and wheezing due to seasonal changes, temperature changes, or odors Do you tire easily? Easy fatigue Are you prone to allergies? Prone to allergies Do you get easily flustered? Easy to panic Does your skin break out easily? Skin prone to urticaria, wheals, hives, rashes, wind bumps Do you get dizzy easily or feel faint when standing up? Easy dizziness or lightheadedness when standing up Have you ever had purpura due to allergies? Purpura due to allergies Do you like to be quiet and lazy to talk? I like quietness and I don't want to talk Does your skin turn red and get scratched when you scratch it? The skin turns red and scratches when scratched Do you sweat easily when you are a little active? Sweating easily after a little bit of activity Are you prone to stress and anxiety? Tension and anxiety Do you feel physically uncomfortable or down? Feeling uncomfortable or down Are you sentimental and emotionally fragile? Sentimental and fragile Do you have excessive oil secretion on your forehead? Excessive oil secretion in the forehead Are you easily frightened or startled? Easily frightened or startled Do you usually have a lot of phlegm, especially in your throat? Do you always feel that there is phlegm blocking it? Excessive phlegm and a feeling of phlegm blocking the throat Do you have pain in your flank or breast? Pain in the flank or breast Do you have a thick, greasy tongue or do you feel like your tongue is coated with a thick coating? Thick and greasy tongue coating Do you sigh for no reason? Sighing for no reason
[0063] In one embodiment, after executing steps S1, S2, S3 and S4, the symptoms corresponding to each symptom node and the attention score of each symptom node in the first graph network output based on the first model are as shown in Table 2. Taking 0.01 as the score threshold, the symptoms corresponding to 21 symptom nodes with attention scores less than 0.01 are determined as shown in Table 3, and the corresponding 21 questions in the self-test table are removed from the self-test table. As shown in Table 4, the simplified self-test table only has 40 questions corresponding to 40 symptoms. According to the simplified self-test table, the questions and corresponding answers in the self-test table filled in step S1 are deleted, and the questions and corresponding answers in the deleted self-test table are used as the filled-in simplified self-test table. Table, the filled-in simplified self-test form and the first individual information and physical label corresponding to the filled-in simplified self-test form are used as data to establish a second data set different from the first data set in step S1; the first symptom information in the filled-in simplified self-test form in the second data set and the first individual information corresponding to the filled-in simplified self-test form are uniquely encoded to generate a first individual vector and a first symptom vector, and the first individual vector is used as the first individual number; a graph network is formed with the first individual number and the first symptom vector as nodes and the symptom vector is embedded, and the graph network and its corresponding physical label are input into the GraphSAGE model to train the GraphSAGE model to obtain a simplified self-test form. The simplified TCM constitution classification and determination self-test form-second model parameters are parameters of a GraphSAGE model that can perform constitution judgment based on the filled-in simplified self-test form; when a new individual who needs to perform constitution judgment appears, obtain the second individual information corresponding to the new individual and the filled-in simplified self-test form, wherein the second individual information can be the name of the new individual who fills in the simplified self-test form, or other information that can distinguish different new individuals who fill in the simplified self-test form; perform unique-hot encoding on the second symptom information in the simplified self-test form and the second individual information corresponding to the simplified self-test form to generate a second individual vector and a second symptom vector, The second individual vector is used as the second individual number, wherein the second symptom information is the intensity of a single symptom on the new individual's body corresponding to the answer in the simplified self-test form, the second individual vector is a vector representation of the second individual information, a single second symptom vector is a vector representation of the intensity of a single symptom on the new individual's body corresponding to the answer in the simplified self-test form, and the second individual number is the second individual information after numbering; a second graph network is formed with the second individual number and the second symptom vector as nodes and the second symptom vector is embedded, the second graph network is input into the GraphSAGE model, and the GraphSAGE model generates a constitution judgment result based on the simplified TCM constitution classification self-test form-second model parameters.Among them, individuals who need to have their physical fitness judged only need to answer 40 questions when filling out a simplified self-test form. Therefore, this embodiment reduces a large number of questions in the self-test form to simplify the self-test form, which can greatly shorten the time for individuals to fill out the self-test form or to question individuals based on the self-test form, thereby significantly improving the efficiency of physical fitness judgment.
[0064] The self-test form simplification method of the present application selects questions in the self-test form for TCM constitution classification and determination according to the attention score, and can reduce the questions in the self-test form to simplify the self-test form without greatly affecting the accuracy of subsequent constitution determination, and can make the time required for an individual to answer questions in the simplified self-test form shorter than the time required to answer questions in the self-test form, thereby improving the efficiency of constitution determination; and the present application reduces the questions in the self-test form, and can avoid inaccurate answers to questions due to a decrease in individual attention or resistance during the long filling out or questioning process, thereby improving the accuracy of constitution determination.
[0065] In some preferred embodiments, in step S2, the process of generating a first individual number and a first symptom vector according to the first individual information in the first data set and the TCM constitution classification and determination self-test table includes:
[0066] S21. Perform one-hot encoding on the individual information and the symptom information in the TCM constitution classification and determination self-test table to generate the first body vector and the first symptom vector, and use the first body vector as the first body number.
[0067] Specifically, the first symptom information is the intensity of the individual's physical symptoms corresponding to the answers in the self-test form. The method of obtaining the first individual number and the first symptom vector can be an existing method that can number the first individual information and obtain the vector representation of the symptom intensity. In this application, the first individual information and the first symptom information in the TCM constitution classification and determination self-test form are one-hot encoded to generate the first individual vector and the first symptom vector. The vector representation of the first individual information and the first symptom information can be generated to facilitate the subsequent composition of the first graph network; and this application uses the first individual vector as the first individual number. Since the first individual vector generated based on one-hot encoding has only one valid bit, different first individual vectors have different valid bits, and different first individual information can be distinguished based on the form of data.
[0068] In this embodiment, the self-test table simplification method of the present application performs one-hot encoding on the first individual information and the first symptom information in the TCM constitution classification and determination self-test table to generate a first individual vector and a first symptom vector, and can generate a vector representation of the first individual information and the first symptom information, which is convenient for the subsequent composition of the first graph network; and the present application uses the first individual vector as the first individual number, which can distinguish different first individual information based on the form of data.
[0069] Second, as Figure 2 As shown, the present application provides a method for determining physical condition, comprising the following steps:
[0070] A1. Obtain a simplified self-test form filled out by a new individual, where the simplified self-test form is obtained based on any of the above self-test form simplification methods;
[0071] A2. Generate a physical fitness judgment result based on the filled-in simplified self-test form based on the pre-trained second model.
[0072] Specifically, the second model is a model for performing physical constitution judgment.
[0073] More specifically, different from the individuals in the simplified self-test form method, the new individuals are individuals whose constitution needs to be judged based on the simplified self-test form. The existing method of judging TCM constitution based on the self-test form is generally to calculate the scores of the symptoms corresponding to each question in the self-test form to judge the TCM constitution of the individual. However, since the simplified self-test form removes some questions compared to the self-test form, the number of symptoms corresponding to the questions is reduced, and the original method of calculating the scores is no longer applicable to the simplified self-test form. Therefore, the present application judges the TCM constitution of the individual based on the second model, eliminating the process of calculating the scores, avoiding the application of the inappropriate calculation method to the simplified self-test form, and making an accurate judgment on the TCM constitution of the individual based on the simplified self-test form.
[0074] The constitution judgment method of the present application selects questions in the self-test form for classification and determination of TCM constitution according to the attention score, and can reduce the questions in the self-test form to simplify the self-test form without greatly affecting the accuracy of subsequent constitution judgment, and can make the time required for individuals to answer questions in the simplified self-test form shorter than the time required to answer questions in the self-test form, thereby improving the efficiency of constitution judgment; and the present application reduces the questions in the self-test form, and can avoid the individual's attention being reduced or the resistance being generated during the long filling or inquiry process, resulting in inaccurate answers to the questions, thereby improving the accuracy of constitution judgment; and the present application judges the TCM constitution of an individual based on the second model, which can avoid the process of calculating the score, avoid applying an inappropriate score calculation method to the simplified self-test form, and can make an accurate judgment of the TCM constitution of the individual based on the simplified self-test form.
[0075] In some preferred embodiments, in step A2, the second model is a GraphSAGE model, and the process of generating a physical fitness judgment result according to a simplified self-test table based on the pre-trained second model includes:
[0076] A21. Obtaining second individual information and a simplified self-test table corresponding to the second individual information, generating a second individual number and a second symptom vector according to the second individual information and the simplified self-test table, and generating a second graph network according to the second individual number and the second symptom vector;
[0077] A22. Input the second graph network into the GraphSAGE model, and the GraphSAGE model performs feature aggregation based on the second graph network until a physical constitution judgment result is generated.
[0078] Specifically, in step A21, the second individual information may be the name of the new individual filling out the simplified self-test form, or other information that can distinguish different new individuals filling out the self-test form. The second individual number is the second individual information after numbering, and the single second symptom vector is a vector representation of the intensity of a single symptom on the body of the new individual corresponding to the answer in the simplified self-test form. The method for generating the second symptom vector may be to perform one-hot encoding on the second symptom information in the simplified self-test form to generate the second symptom vector, or other existing methods that can generate a symptom vector based on the simplified self-test form, wherein the second symptom information is the intensity of the symptom on the body of the new individual corresponding to the answer in the simplified self-test form. The method for generating the second graph network may be to directly form a second graph network with the second symptom vector and the second individual number as nodes, or other methods that can generate the second graph network based on the second individual number and the second symptom vector.
[0079] More specifically, since the GraphSAGE model performs feature aggregation based on the second graph network, in this embodiment, the physical constitution judgment method of the present application can, in the process of generating the physical constitution judgment results, aggregate the features of each node that has undergone information update to each node that has not undergone information update, thereby improving the accuracy of the physical constitution judgment based on the GraphSAGE model.
[0080] In some preferred embodiments, in step A21, the process of generating the second graph network according to the second individual number and the second symptom vector includes:
[0081] A211. The second individual number and the second symptom vector are used as nodes to form a second graph network;
[0082] A212. Embedding the symptom node of the second graph network, wherein the symptom node of the second graph network is a second symptom vector as a node of the second graph network;
[0083] In step A22, the process of the GraphSAGE model performing feature aggregation based on the second graph network until a physical fitness judgment result is generated includes:
[0084] A221. The GraphSAGE model performs feature aggregation based on the symptom nodes after embedding in the second graph network until a constitution judgment result is generated.
[0085] Specifically, in step A211, the symptom nodes after embedding processing are vector representations of the TCM constitution corresponding to the new individual's physical symptoms. For example, if the TCM constitution corresponding to the symptoms "cold hands and feet" and "cold sensation in the back of the stomach or waist and knees" is yang deficiency, then the symptom nodes corresponding to the symptom "cold hands and feet" and the symptom nodes corresponding to the symptom "cold sensation in the back of the stomach or waist and knees" both contain information characterizing yang deficiency. The method for embedding the symptom nodes of the second graph network can be an existing embedding processing method. Preferably, the symptom nodes of the second graph network are embedded based on the all-MiniLM-L6-v2 model in the sentence-transformers library.
[0086] More specifically, since the symptom nodes after embedding processing can represent the TCM constitution corresponding to the new individual's physical symptoms, in this embodiment, the constitution judgment method of the present application embeds the symptom nodes of the second graph network, and enables the GraphSAGE model to perform feature aggregation based on the symptom nodes after embedding processing in the second graph network. During the feature aggregation process, the features of each node after the information is updated can be aggregated to the corresponding nodes whose information has not been updated, thereby further improving the accuracy of constitution judgment based on the GraphSAGE model.
[0087] In some preferred embodiments, in step A221, the process of performing feature aggregation based on the symptom nodes after embedding in the second graph network by the GraphSAGE model until generating a constitution judgment result includes:
[0088] A2211. An individual node in the second graph network is taken as the central node, the symptom node corresponding thereto is taken as the one-hop neighbor node, and the individual nodes other than the central node corresponding to the one-hop neighbor node are taken as the two-hop neighbor nodes;
[0089] A2212. Aggregate the features of two-hop neighbor nodes to one-hop neighbor nodes;
[0090] A2213. Aggregate the features of one-hop neighbor nodes to the central node;
[0091] A2214. An individual node in the second graph network other than the central node is used as a new central node, the symptom node corresponding thereto is used as a new one-hop neighbor node, and the individual nodes other than the central node and the new central node corresponding to the new one-hop neighbor node are used as new two-hop neighbor nodes;
[0092] A2215. Repeat steps A2212-A2214 until feature aggregation is completed for all individual nodes;
[0093] A2216. Treat all individual nodes as if no feature aggregation has been performed and repeat steps A2211-A2215 until a physical constitution judgment result is generated.
[0094] Specifically, in step A2211, the central node and the two-hop neighbor node are connected by at least one neighbor node, that is, the two individual nodes serving as the central node and the two-hop neighbor node correspond to at least one identical second symptom vector. Therefore, in this embodiment, the physical constitution judgment method of the present application aggregates the features of the two-hop neighbor nodes to the one-hop neighbor nodes, and aggregates the features of the one-hop neighbor nodes to the central node. In the feature aggregation process, the features of the symptom nodes after the information is updated can be aggregated to the corresponding central nodes whose information has not been updated, and the features of the individual nodes after the information is updated can be aggregated to the individual nodes corresponding to them having at least one identical second symptom vector, thereby further improving the accuracy of physical constitution judgment based on the GraphSAGE model.
[0095] In some preferred embodiments, the training process of the second model includes:
[0096] B1. Filter the completed TCM constitution classification and determination self-test form according to the simplified self-test form, obtain the completed simplified self-test form, establish a second data set according to the completed simplified self-test form and the first body information and constitution label corresponding to the completed simplified self-test form, use the first body number and the first symptom vector corresponding to each first body information in the second data set as input constitution samples, use each constitution label in the second data set as output constitution sample, divide the constitution sample set consisting of multiple corresponding input constitution samples and output constitution samples into a training set and a test set, and the training set and the test set each contain multiple corresponding input constitution samples and output constitution samples;
[0097] B2. Train the second model based on the training set;
[0098] B3. Test the second model based on the test set.
[0099] Specifically, in step B1, since the first data set of step S1 includes the first individual information and the completed self-test form and physical constitution label corresponding to the first individual information, and the questions in the simplified self-test form are less than the questions in the self-test form, it is only necessary to select the questions and corresponding answers in the completed self-test form according to the simplified self-test form to obtain the completed simplified self-test form. Secondly, the first individual information and physical constitution label corresponding to the self-test form before the selection process also correspond to the simplified self-test form completed after the selection process. In addition, the first individual number and the first symptom vector corresponding to the self-test form in step S2 also correspond to the simplified self-test form completed after the selection process.
[0100] In this embodiment, the constitution judgment method of the present application divides a constitution sample set composed of multiple constitution samples into a training set and a test set, and can test the generalization degree of the trained GraphSAGE model to determine its performance while training the GraphSAGE model.
[0101] Preferably, the ratio of the number of physical samples in the training set and the test set is 7:3.
[0102] In summary, the present application provides a method for simplifying a self-test form and a method for determining a physical constitution, wherein the method for simplifying the self-test form provided by the present application selects the questions in the self-test form for classification and determination of TCM physical constitution according to the attention score, and can reduce the questions in the self-test form to simplify the self-test form without greatly affecting the accuracy of subsequent physical constitution determination, and can make the time required for an individual to answer the questions in the simplified self-test form shorter than the time required to answer the questions in the self-test form, thereby improving the efficiency of physical constitution determination; and the present application reduces the questions in the self-test form, and can avoid inaccurate answers to questions due to a decrease in individual attention or resistance during a long period of filling out or questioning, thereby improving the accuracy of physical constitution determination.
[0103] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.
[0104] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A method for simplifying a self-test table, used to simplify a self-test table for TCM constitution classification and determination, characterized in that: The following steps are involved: S1. Obtain the first individual information and the completed TCM constitution classification and determination self-test form and constitution label corresponding to the first individual information to establish a first data set; S2. Generate a first individual number and a first symptom vector according to the first individual information in the first data set and the TCM constitution classification and determination self-test table; The process of generating a first individual number and a first symptom vector according to the first individual information in the first data set and the TCM constitution classification and determination self-test table includes: S21. Perform one-hot encoding on the first individual information and the first symptom information in the TCM constitution classification and determination self-test table to generate a first body vector and a first symptom vector, and use the first individual vector as the first body number; S3. Using the first individual number and the first symptom vector as nodes to form a first graph network, construct a first model, and train the first model based on the first graph network and the corresponding physical label; S4. outputting the attention score of each symptom node in the first graph network based on the first model, wherein each symptom node in the first graph network is each first symptom vector as a node of the first graph network, and the first model is a model based on the result generated by the graph attention mechanism, which is used to generate a physical fitness judgment result, that is, a vector representation of an individual node in the first graph network, wherein the individual node in the first graph network is the first individual number as a node of the first graph network, and the attention mechanism is used to learn and determine the attention score of each symptom node connected to the individual node, and the attention score of each symptom node represents the importance of each symptom node for generating a physical fitness judgment result; S5. Perform threshold screening on the questions in the TCM constitution classification and determination self-test form according to the attention score to obtain a simplified self-test form.
2. A method for determining physical constitution, characterized in that: The following steps are involved: A1. Obtain a simplified self-test form filled out by a new individual, the simplified self-test form is obtained based on the simplified method of the self-test form according to claim 1; A2. Generate a physical fitness judgment result based on the filled-in simplified self-test form based on the pre-trained second model.
3. A method for determining physical condition according to claim 2, characterized in that: In step A2, the second model is a GraphSAGE model, and the process of generating a physical fitness judgment result according to a simplified self-test table based on the pre-trained second model includes: A21. Obtain second individual information and a simplified self-test table corresponding to the second individual information, generate a second individual number and a second symptom vector according to the second individual information and the simplified self-test table, and generate a second graph network according to the second individual number and the second symptom vector; A22. Input the second graph network into the GraphSAGE model, and the GraphSAGE model performs feature aggregation based on the second graph network until a physical constitution judgment result is generated.
4. A method for determining physical condition according to claim 3, characterized in that: In step A21, the process of generating the second graph network according to the second individual number and the second symptom vector includes: A211. The second individual number and the second symptom vector are used as nodes to form a second graph network; A212. Embedding the symptom node of the second graph network, wherein the symptom node of the second graph network is a second symptom vector as a node of the second graph network; In step A22, the process of the GraphSAGE model performing feature aggregation based on the second graph network until a physical fitness judgment result is generated includes: A221.GraphSAGE model performs feature aggregation based on the symptom nodes after embedding processing in the second graph network until a physical constitution judgment result is generated.
5. A method for determining physical condition according to claim 4, characterized in that: In step A221, the process in which the GraphSAGE model performs feature aggregation based on the symptom nodes after embedding in the second graph network until a constitution judgment result is generated includes: A2211. An individual node in the second graph network is taken as the central node, the symptom node corresponding thereto is taken as the one-hop neighbor node, and the individual nodes other than the central node corresponding to the one-hop neighbor node are taken as the two-hop neighbor nodes; A2212. Aggregate the features of the two-hop neighbor nodes to the one-hop neighbor nodes; A2213. Aggregate the features of one-hop neighbor nodes to the central node; A2214. An individual node in the second graph network other than the central node is used as a new central node, the symptom node corresponding thereto is used as a new one-hop neighbor node, and the individual nodes other than the central node and the new central node corresponding to the new one-hop neighbor node are used as new two-hop neighbor nodes; A2215. Repeat steps A2212-A2214 until feature aggregation is completed for all individual nodes; A2216. Treat all individual nodes as if feature aggregation has not been performed and repeat steps A2211-A2215 until a physical constitution judgment result is generated.
6. A method for determining physical condition according to claim 3, characterized in that: The training process of the second model includes: B1. Filter the completed TCM constitution classification and determination self-test form according to the simplified self-test form, obtain the completed simplified self-test form, establish a second data set according to the completed simplified self-test form and the first body information and constitution label corresponding to the completed simplified self-test form, use the first body number and the first symptom vector corresponding to each first body information in the second data set as input constitution samples, use each constitution label in the second data set as output constitution sample, divide the constitution sample set consisting of multiple corresponding input constitution samples and output constitution samples into a training set and a test set, wherein the training set and the test set each contain multiple corresponding input constitution samples and output constitution samples; B2. training the second model according to the training set; B3. Test the second model according to the test set.
Citation Information
Patent Citations
Physical fitness detection method
CN110570947A
Traditional Chinese medicine constitution identification method, device and equipment and storage medium
CN116364300A