A method, system and storage medium for intelligent identification of traditional Chinese medicine syndromes

Abstract

The present invention relates to the field of medical care information technology, and in particular to a method, system and storage medium for intelligent identification of traditional Chinese medicine syndromes. The method includes obtaining training sample data and symptom manifestation categories of the four-diagnosis medical records of the diagnosed person; determining frequency adjustment parameters according to other syndrome types in the sample data and the frequency of occurrence of other syndromes; and then determining the target correlation coefficient between symptoms and syndromes; constructing a decision tree according to the target correlation coefficient and the information gain index; determining the combined correlation coefficient according to the frequency and target correlation coefficient of any two symptom manifestation categories; determining the decision weight based on the combined correlation coefficient of any two symptoms and the target correlation coefficient of all symptoms; weighting each decision tree according to the decision weight, constructing a syndrome identification model, and performing intelligent syndrome identification of the four-diagnosis medical records according to the syndrome identification model. The present invention can improve the accuracy and reliability of intelligent syndrome identification.

Description

Technical Field

[0001] The present invention relates to the field of medical care information technology, and in particular to a method, system and storage medium for intelligent identification of traditional Chinese medicine syndromes. Background Art

[0002] Syndrome in TCM theory is also called syndrome, which is a summary of the pathology of a certain stage or type of disease process in the body, and a comprehensive description of the disease state in TCM. The identification of syndrome is to collect symptoms and signs from the four diagnostic methods of TCM (observation, auscultation, questioning and palpation).

[0003] In the related technology, decision trees are constructed based on the random forest algorithm, and intelligent syndrome identification effects are achieved. However, the principle of the random forest algorithm is to train multiple decision trees through a number of randomly sampled samples and a number of randomly combined features, and multiple decision trees are combined to form a random forest. The weight of the influence of each decision tree on the output result in the final random forest is the same. However, the correlation between different symptoms and syndromes is different, and the symptoms of each decision tree are randomly selected when constructing a random forest. Therefore, the accuracy and reliability of directly constructing a decision tree and identifying syndromes are low. Summary of the invention

[0004] In order to solve the technical problem of low accuracy and reliability in building a decision tree and performing syndrome identification in the related art, the present invention provides a method, system and storage medium for intelligent identification of TCM syndromes. The technical solutions adopted are as follows:

[0005] The present invention proposes a method for intelligent identification of TCM syndromes, the method comprising:

[0006] Obtain training sample data of the four-diagnosis medical records of the diagnosed person, and determine that the training sample data belongs to the target sample data of the target syndrome; obtain the symptom manifestation category in the training sample data, and select one symptom to be analyzed from all the symptom manifestation categories;

[0007] For any target sample data, according to other syndrome types in the target sample data and the occurrence frequencies of the symptoms to be analyzed belonging to other syndromes, the frequency adjustment parameters of the symptoms to be analyzed in the target sample data are determined; according to the occurrence frequencies of the symptoms to be analyzed in the target sample data and the frequency adjustment parameters of the symptoms to be analyzed in all the target sample data, the target correlation coefficient of the symptoms to be analyzed is determined; the information gain index of the symptoms to be analyzed in constructing a decision tree is determined; according to the target correlation coefficient and the information gain index, a decision tree is constructed for the symptoms to be analyzed;

[0008] According to the frequency of any two symptom manifestation categories appearing in the training sample data and the target sample data, and the target correlation coefficients corresponding to the two symptom manifestation categories, determine the combined correlation coefficient of the two symptom manifestation categories; based on the combined correlation coefficient of any two symptom manifestation categories in any decision tree, and the target correlation coefficients of all symptoms for the target syndrome, determine the decision weight of the decision tree;

[0009] Each decision tree is weighted according to the decision weight, a syndrome recognition model is constructed, and intelligent syndrome recognition of the four-diagnosis medical records is performed according to the syndrome recognition model.

[0010] Further, determining the frequency adjustment parameter of the symptom to be analyzed in the target sample data includes:

[0011] Determine the frequency and value of the symptoms to be analyzed in other syndromes;

[0012] The product of the number of syndrome types and the frequency sum value in the target sample data is normalized to the inverse of the product, which is used as the frequency adjustment parameter.

[0013] Further, according to the frequency of occurrence of the symptom to be analyzed in the target sample data and the frequency adjustment parameter of the symptom to be analyzed in all the target sample data, the target correlation coefficient of the symptom to be analyzed is determined, including:

[0014] The mean of the frequency adjustment parameters of the symptoms to be analyzed in all the target sample data is used as the overall adjustment parameter of the symptoms to be analyzed;

[0015] The frequency of occurrence of the symptom to be analyzed in the target sample data and the product of the overall adjustment parameter are normalized to obtain the target correlation coefficient of the symptom to be analyzed.

[0016] Furthermore, the information gain index of the symptom to be analyzed in constructing the decision tree is determined, including:

[0017] Based on the ID3 algorithm, the difference between the entropy of the training set of the decision tree and the entropy of the training set of the decision tree under the given symptom to be analyzed is used as the information gain indicator.

[0018] Further, according to the target correlation coefficient and the information gain index, a decision tree is constructed for the symptoms to be analyzed, including:

[0019] Calculating the product of the target correlation coefficient and the information gain index, and normalizing the result to be used as the division weight of the symptoms to be analyzed for training in the decision tree;

[0020] Obtain the division weights of all symptoms in the decision tree to be trained, take the symptom with the largest division weight as the first layer to build the decision tree, and then calculate the division weights of all remaining symptoms except those that have already participated in the construction of the decision tree, take the largest one as the second layer, and so on to complete the construction of the decision tree.

[0021] Furthermore, determining the combined correlation coefficient of the two symptom manifestation categories includes:

[0022] The sum of the target correlation coefficients of the two symptom manifestation categories was used as the first related influence indicator;

[0023] Based on the frequency ratio of the two symptoms occurring together and the frequency ratio of the two symptoms occurring separately, the second relevant impact index of the combination of the two symptoms is determined;

[0024] The product of the first relevant impact index and the second relevant impact index is calculated, and maximum and minimum value normalization processing is performed to obtain a combined correlation coefficient.

[0025] Furthermore, the calculation formula corresponding to the second relevant impact indicator is:

[0026] In the formula, It represents the second related impact index between the symptom of category i and the symptom of category v, It indicates the frequency of the i-th symptom and the v-th symptom appearing in all target sample data at the same time. represents the frequency of the i-th symptom and the v-th symptom appearing in all training sample data at the same time, Indicates the frequency of occurrence of the ith symptom in all target sample data, Indicates the frequency of occurrence of the vth symptom in all target sample data, represents the frequency of occurrence of the i-th symptom in all training sample data, Represents the frequency of occurrence of the vth type of symptoms in all training sample data.

[0027] Furthermore, the decision weight of the decision tree is determined based on the combined correlation coefficient of any two symptom manifestation categories in any decision tree and the target correlation coefficient of all symptoms, including:

[0028] The mean of the target correlation coefficients of all symptoms in the same decision tree is used as the first weight indicator;

[0029] The mean of the combined correlation coefficients corresponding to all combinations of any two symptoms in the same decision tree is used as the second weight indicator;

[0030] The product of the first weight index and the second weight index is calculated, and mapped to [-1, 1] based on the Premnmx function to obtain the decision weight.

[0031] On the other hand, the present invention also provides a system for intelligent identification of TCM syndromes, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the aforementioned method for intelligent identification of TCM syndromes when executing the computer program.

[0032] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the steps of the aforementioned method for intelligent identification of TCM syndromes.

[0033] The present invention has the following beneficial effects:

[0034] In an embodiment of the present invention, in order to address the problem that the influence weight of each decision tree in the random forest constructed in the related technology on the output result is the same, which leads to low accuracy and reliability of syndrome identification, this solution determines the target correlation coefficient of each symptom to each syndrome through frequency analysis, and then realizes the construction of the decision tree; then, through the combination of two symptoms and the target correlation coefficient of all symptoms in the decision tree to the syndrome, the decision weight of the decision tree is determined, and the decision weight is the weight information of the corresponding decision tree itself when performing decision analysis. The decision weight is used to weight the decision of each decision tree, and a syndrome identification model is constructed, so that different decision trees can be given different weights, and more accurate random forest output results can be obtained, so as to improve the accuracy and reliability of intelligent syndrome identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. 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 creative work.

[0036] Figure 1 A flow chart of a method for intelligent identification of TCM syndromes provided by one embodiment of the present invention;

[0037] Figure 2 A schematic diagram of the random forest model structure provided by one embodiment of the present invention;

[0038] Figure 3 A structural diagram of a TCM syndrome intelligent identification system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a method, system and storage medium for intelligent identification of TCM syndromes proposed by the present invention, its specific implementation, structure, features and effects in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0040] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0041] Explanation of some terms that appear in this plan:

[0042] The four diagnostic methods in traditional Chinese medicine are inspection, auscultation, questioning and palpation.

[0043] Syndrome is a special term in traditional Chinese medicine, which is a general term for a series of interrelated symptoms, namely the body's reaction state and its movement and changes at the overall level during the disease process obtained through the four examinations of observation, auscultation, listening, and palpation. It is abbreviated as syndrome or symptom, which is the reaction state of the essence of organic connection between the location, cause, nature, course of the disease and the strength of the body's disease resistance at a certain stage in the disease process, manifested as clinically observable symptoms, etc. Specifically, it can include eight types of syndromes: yin, yang, exterior, interior, cold, heat, deficiency and excess.

[0044] Symptoms refer to the physical signs of the person being diagnosed, such as complexion, expression, breathing, voice, diet, defecation, sleep, and pulse (in fact, the classification of symptoms in the four diagnostic methods of Traditional Chinese Medicine is more numerous and detailed, and only a few types of data are selected here as examples). The possible manifestations of facial color include (1 yellow face, 2 red face, 3 blue face, 4 pale face, etc.), the possible manifestations of expression include (1 energetic, 2 less energetic, 3 absent-minded, 4 false spirit, etc.), the possible manifestations of breathing include (1 steady, 2 rapid, 3 slow, 4 disordered, etc.), the possible manifestations of voice include (1 clear voice, 2 turbid voice, 3 dry voice, 4 hoarse, etc.), the possible manifestations of diet include (1 normal, 2 eat more, 3 eat less, 4 anorexia, etc.), the possible manifestations of defecation include (1 normal, 2 frequent, 3 diarrhea, 4 constipation, etc.), the possible manifestations of sleep include (1 normal, 2 dreamy, 3 easy to wake up, 4 insomnia, etc.), the possible manifestations of pulse include (1 steady, 2 floating pulse, 3 sinking pulse, 4 rapid pulse, etc.), the different numbers only represent different symptom manifestation categories, and do not represent the specific numerical differences.

[0045] The specific scheme of the intelligent identification method of TCM syndrome provided by the present invention is described in detail below with reference to the accompanying drawings.

[0046] See also Figure 1, which shows a flow chart of a method for intelligent identification of TCM syndromes provided by an embodiment of the present invention, the method comprising:

[0047] S101: Obtain training sample data of the four-diagnosis medical records of the diagnosed person, determine whether the training sample data belongs to the target sample data of the target syndrome; obtain the symptom manifestation category in the training sample data, and select one symptom to be analyzed from all the symptom manifestation categories.

[0048] Among them, the training sample data is the data information after the four diagnostic symptoms are numerically represented. For example, the symptom of yellowish complexion is numbered 11, slow breathing is numbered 33, and constipation is numbered 64; when the symptom manifestation categories in the four diagnostic medical records of a certain patient are yellowish complexion, slow breathing and constipation, the corresponding symptom can be numerically represented, and the training sample data obtained is {11,33,64}. Therefore, the four diagnostic medical records of all the patients are processed separately to obtain the training sample data of each patient.

[0049] It should be noted that, since this solution is to perform decision tree analysis, the number of training sample data required can be specifically 1000. Of course, it can also be adjusted according to actual conditions, and there is no restriction on this.

[0050] The target syndrome may be, for example, "wind-heat". The embodiment of the present invention aims to train a syndrome recognition model based on the random forest algorithm to estimate the probability of the patient having wind-heat syndrome. The final output result label of each decision tree constructed is either "wind-heat syndrome" or "not wind-heat syndrome". The specific random forest model can be as follows: Figure 2 As shown, Figure 2 A schematic diagram of the random forest model structure provided by an embodiment of the present invention. Decisions are made through different decision trees, and the final output result is obtained according to the combiner summary.

[0051] In the embodiment of the present invention, the symptom to be analyzed can be set as diarrhea, so as to facilitate the subsequent correlation analysis between diarrhea and wind-heat.

[0052] S102: For any target sample data, determine the frequency adjustment parameter of the symptom to be analyzed in the target sample data according to other syndrome types in the target sample data and the occurrence frequency of the symptom to be analyzed belonging to other syndromes; determine the target correlation coefficient of the symptom to be analyzed according to the occurrence frequency of the symptom to be analyzed in the target sample data and the frequency adjustment parameter of the symptom to be analyzed in all target sample data; determine the information gain index of the symptom to be analyzed in constructing a decision tree; construct a decision tree for the symptom to be analyzed according to the target correlation coefficient and the information gain index.

[0053] It is understandable that a certain symptom of a diagnosed person can correspond to multiple syndromes. For example, the symptom of yellow face can correspond to wind-heat or physical weakness. Therefore, when constructing a decision tree, it is not a strongly correlated correspondence. It is necessary to adjust the frequency adjustment parameters of the symptoms to be analyzed based on the impact of other syndromes on the symptoms.

[0054] Furthermore, in some embodiments of the present invention, the frequency adjustment parameters of the symptoms to be analyzed in the target sample data are determined, including: determining the frequency and value of the symptoms to be analyzed in other syndromes; multiplying the number of syndrome types in the target sample data by the frequency and value, and normalizing the inverse of the product as the frequency adjustment parameter.

[0055] Taking the i-th symptom as an example, when the r-th target sample data is a case where other syndromes occur concurrently with the target syndrome, the frequency adjustment parameter of the i-th symptom of the r-th target sample data is The calculation method is:

[0056]

[0057] In the formula, norm represents the maximum and minimum value normalization processing, represents the number of all syndrome types in the rth target sample data, F represents the total number of syndrome types in all training sample data, and f represents the syndrome index, where f does not represent the target syndrome. It represents the frequency of occurrence of the i-th symptom in the sample data corresponding to the f-th syndrome except the target syndrome in the r-th target sample data.

[0058] in, The larger the value of , the more other syndromes that occur concurrently with the target syndrome in the rth target sample data. The more likely the symptoms exhibited by the rth target sample data are symptoms related to other syndromes, the smaller the calculated reliability of the correlation between the sample and the target syndrome. Similarly, The larger the value, the greater the frequency of the i-th symptom in other syndromes, the more likely the i-th symptom is to be related to other syndromes, and the smaller the calculation reliability of the correlation with the target syndrome, the smaller the frequency adjustment parameter The smaller.

[0059] It can be understood that the frequency adjustment parameter represents the credibility of the symptom to be analyzed in the target sample data and the target syndrome. The greater the credibility, the larger the corresponding frequency adjustment parameter, and the smaller the credibility, the smaller the corresponding frequency adjustment parameter. The credibility specifically indicates the strong correlation between the symptom to be analyzed in the target sample data and the target syndrome, that is, the smaller the frequency adjustment parameter, the more likely the symptom to be analyzed is caused by other syndromes, and the larger the frequency adjustment parameter, the more likely the symptom to be analyzed is caused by the target syndrome.

[0060] Furthermore, in some embodiments of the present invention, the target correlation coefficient of the symptom to be analyzed is determined based on the frequency of occurrence of the symptom to be analyzed in the target sample data and the frequency adjustment parameters of the symptom to be analyzed in all target sample data, including: taking the mean of the frequency adjustment parameters of the symptom to be analyzed in all target sample data as the overall adjustment parameter of the symptom to be analyzed; normalizing the product of the frequency of occurrence of the symptom to be analyzed in the target sample data and the overall adjustment parameter to obtain the target correlation coefficient of the symptom to be analyzed.

[0061] Among them, the higher the frequency of the symptom to be analyzed in all target sample data, the stronger the correlation between the symptom to be analyzed and the target syndrome. The smaller the frequency adjustment parameter is, the more it indicates that the symptom to be analyzed is caused by other syndromes, and the larger the frequency adjustment parameter is, the more it indicates that the symptom to be analyzed is caused by the target syndrome. Therefore, by directly calculating the product of the frequency of the symptom to be analyzed in the target sample data and the overall adjustment parameter, and normalizing the product value, the target correlation coefficient of the symptom to be analyzed is obtained. The target correlation coefficient is the correlation coefficient between the symptom to be analyzed and the target syndrome. The larger its value is, the stronger the correlation between the symptom to be analyzed and the target syndrome, that is, the appearance of the symptom to be analyzed indicates that the patient is affected by the target syndrome.

[0062] Furthermore, in some embodiments of the present invention, the information gain index of the symptom to be analyzed in constructing a decision tree is determined, including: based on the ID3 algorithm, taking the difference between the entropy of the training set of the decision tree and the entropy of the training set of the decision tree under the given symptom to be analyzed as the information gain index.

[0063] Among them, the ID3 algorithm is a partitioning algorithm for constructing a decision tree. The information gain is obtained by considering the change in information entropy under given preset conditions. In the embodiment of the present invention, the preset conditions are the symptoms to be analyzed, and the information gain is the information gain index. Since the ID3 algorithm is a commonly used algorithm for constructing a decision tree, the embodiment of the present invention will not be further described.

[0064] It should be noted that a certain number of symptoms can be combined to construct a decision tree, that is, each decision tree has a corresponding number of symptoms. In the embodiment of the present invention, for example, there are a total of 10 symptom types, and 5 of them can be selected at random to construct a decision tree. Thus, a total of All symptoms in each decision tree were subsequently analyzed.

[0065] Furthermore, in some embodiments of the present invention, a decision tree is constructed for the symptoms to be analyzed based on the target correlation coefficient and the information gain index, including: calculating the product of the target correlation coefficient and the information gain index, and normalizing it as the division weight of the symptoms to be analyzed for training in the decision tree; obtaining the division weights of all symptoms in the decision tree to be trained, and constructing the decision tree with the largest division weight among all symptoms as the first layer, and then calculating the division weights of all remaining symptoms except those that have already participated in the construction of the decision tree, and taking the largest one as the second layer, and so on to complete the construction of the decision tree.

[0066] Among them, the larger the value of the target correlation coefficient is, the stronger the correlation between the symptoms to be analyzed and the target syndrome is. The target correlation coefficient is weighted by the information gain index to obtain the partition weight, and the decision tree is constructed based on the partition weight.

[0067] S103: Determine the combined correlation coefficient of the two symptom manifestation categories based on the frequency of any two symptom manifestation categories appearing in the training sample data and the target sample data, and the target correlation coefficients corresponding to the two symptom manifestation categories; determine the decision weight of the decision tree based on the combined correlation coefficient of any two symptom manifestation categories in any decision tree, and the target correlation coefficients of all symptoms for the target syndrome.

[0068] It should be noted that the decision weights of each decision tree in the relevant technology are the same. However, in actual practice, for some symptoms, their correlation with a certain syndrome alone may not be strong, but when some symptoms occur at the same time as other symptoms, the correlation between the two symptoms and the corresponding syndrome is very strong. In this case, the impact of the combined correlation between symptoms on the correlation should be considered. For multiple features whose correlation increases after association, when they appear in the same decision tree at the same time, the decision tree in which they are located will have a higher weight on the result.

[0069] Therefore, in this scheme, the combined correlation coefficient is calculated, and based on the combined correlation coefficient and the target correlation coefficient of all symptoms for the target syndrome, the decision weight is determined to allocate the target syndrome weight to each decision tree.

[0070] Furthermore, in some embodiments of the present invention, the combined correlation coefficient of two symptom manifestation categories is determined, including: taking the sum of the target correlation coefficients of the two symptom manifestation categories as the first relevant influence indicator; determining the second relevant influence indicator of the combination of the two symptoms based on the frequency ratio of the two symptoms occurring together and the frequency ratio of the two symptoms occurring separately; calculating the product of the first relevant influence indicator and the second relevant influence indicator, and performing maximum and minimum value normalization processing to obtain the combined correlation coefficient.

[0071] For example, taking sallow face and constipation as a combination, for wind-heat syndrome, the sum of the target correlation coefficients corresponding to sallow face and constipation respectively is used as the first relevant influence indicator. The larger the value, the greater the influence of the combination of sallow face and constipation on wind-heat, that is, the symptoms of wind-heat are likely to include sallow face and constipation at the same time.

[0072] Among them, it is also possible that yellow face or constipation alone may affect the wind-heat syndrome. Therefore, it is necessary to eliminate the impact of the frequency of the two symptoms occurring together and the frequency of their occurring alone based on the difference between the two symptoms occurring together and the frequency of their occurring alone.

[0073] Furthermore, in some embodiments of the present invention, the calculation formula corresponding to the second relevant impact indicator is:

[0074]

[0075] In the formula, It represents the second related impact index between the symptom of category i and the symptom of category v, It indicates the frequency of the i-th symptom and the v-th symptom appearing in all target sample data at the same time. represents the frequency of the i-th symptom and the v-th symptom appearing in all training sample data at the same time, Indicates the frequency of occurrence of the ith symptom in all target sample data, Indicates the frequency of occurrence of the vth symptom in all target sample data, represents the frequency of occurrence of the i-th symptom in all training sample data, Represents the frequency of occurrence of the vth type of symptoms in all training sample data.

[0076] in, represents the frequency ratio of the two symptoms of category i and category v occurring together, and It represents the frequency ratio of the two symptoms of category i and category v occurring separately. The second relevant influence index is obtained by difference. The larger the value of the second relevant influence index is, the more the correlation between the two symptoms of category i and category v is improved after combination, that is, the correlation between the symptoms and the target syndrome is improved more after combination.

[0077] In summary, the product of the first relevant influence index and the second relevant influence index is calculated, and the maximum and minimum values ​​are normalized to obtain the combined correlation coefficient, which represents the influence of the two symptoms on the target syndrome when they occur together, so as to facilitate the calculation of decision weights based on the combined correlation coefficient in the future.

[0078] Furthermore, in some embodiments of the present invention, the decision weight of the decision tree is determined based on the combined correlation coefficient of any two symptom manifestation categories in any decision tree and the target correlation coefficients of all symptoms, including: taking the mean of the target correlation coefficients of all symptoms in the same decision tree as the first weight indicator; taking the mean of the combined correlation coefficients corresponding to all combinations of any two symptoms in the same decision tree as the second weight indicator; calculating the product of the first weight indicator and the second weight indicator, and mapping it to between [-1,1] based on the Premnmx function to obtain the decision weight.

[0079] Among them, the Premnmx function is a normalization function in MATLAB, which is used to normalize the data to a value range of [-1,1]. It is a technology well known to those skilled in the art and will not be described in detail.

[0080] Among them, since both the target correlation coefficient and the combined correlation coefficient represent the correlation between symptoms and syndromes in the decision tree, the first weight index is calculated to characterize the influence of a single symptom, and the second weight index is calculated to characterize the influence of combined symptoms, thereby integrating the single symptom situation and the symptom combination situation to obtain a more accurate, objective and reliable decision weight, which is convenient for achieving a more effective decision tree decision effect in subsequent steps.

[0081] S104: Decision weighting is performed on each decision tree according to the decision weight, a syndrome recognition model is constructed, and intelligent syndrome recognition of the four-diagnosis medical records is performed according to the syndrome recognition model.

[0082] In the embodiments of the invention, see Figure 2 In the random forest model in the related art, when all decision trees are combined into a random forest model, the default weight of each decision tree is 1. When the random forest finally outputs the result, the average value of all decision tree results is taken. Taking wind-heat syndrome as an example, in K decision trees, the label value of "is wind-heat syndrome" is regarded as 1, and the label value of "is not wind-heat syndrome" is regarded as 0. If the result obtained by 0.25K decision trees is "is wind-heat syndrome", and the result obtained by 0.75K decision trees is "is not wind-heat syndrome", then the probability of wind-heat syndrome output by the random forest is 0.25.

[0083] This solution obtains the decision weight of each decision tree according to the above steps and uses it as the weight of each decision tree when adding up. The decision tree with more relevant features has a greater weight. The decision weight is introduced into the random forest model to obtain a syndrome recognition model.

[0084] Input the medical records of the four diagnoses to be identified, and after being processed by the syndrome identification model, the final output probability of wind-heat syndrome is the weighted average of the label values ​​output by all decision trees, thereby obtaining the corresponding probability of wind-heat syndrome.

[0085] It should be noted that the probability of wind-heat syndrome is only used to assist in syndrome identification and plays an auxiliary role. It cannot be directly used as a basis for judgment. It still needs to be judged in combination with actual symptoms.

[0086] In an embodiment of the present invention, in order to address the problem that the influence weight of each decision tree in the random forest constructed in the related technology on the output result is the same, which leads to low accuracy and reliability of syndrome identification, this solution determines the target correlation coefficient of each symptom to each syndrome through frequency analysis, and then realizes the construction of the decision tree; then, through the combination of two symptoms and the target correlation coefficient of all symptoms in the decision tree to the syndrome, the decision weight of the decision tree is determined, and the decision weight is the weight information of the corresponding decision tree itself when performing decision analysis. The decision weight is used to weight the decision of each decision tree, and a syndrome identification model is constructed, so that different decision trees can be given different weights, and more accurate random forest output results can be obtained, so as to improve the accuracy and reliability of intelligent syndrome identification.

[0087] The present invention also provides a TCM syndrome intelligent identification system, see Figure 3 , Figure 3 This is a structural diagram of a system for intelligent identification of TCM syndromes provided in one embodiment of the present invention. The system 600 includes a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and executable on the processor 602. When the processor 602 executes the computer program 603, the steps of the aforementioned method for intelligent identification of TCM syndromes are implemented.

[0088] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the steps of the aforementioned method for intelligent identification of TCM syndromes are implemented.

[0089] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0090] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for intelligent identification of TCM syndromes, characterized in that: The method comprises: Obtain training sample data of the four-diagnosis medical records of the diagnosed person, and determine that the training sample data belongs to the target sample data of the target syndrome; obtain the symptom manifestation category in the training sample data, and select one symptom to be analyzed from all the symptom manifestation categories; For any target sample data, according to other syndrome types in the target sample data and the occurrence frequencies of the symptoms to be analyzed belonging to other syndromes, the frequency adjustment parameters of the symptoms to be analyzed in the target sample data are determined; according to the occurrence frequencies of the symptoms to be analyzed in the target sample data and the frequency adjustment parameters of the symptoms to be analyzed in all the target sample data, the target correlation coefficient of the symptoms to be analyzed is determined; the information gain index of the symptoms to be analyzed in constructing a decision tree is determined; according to the target correlation coefficient and the information gain index, a decision tree is constructed for the symptoms to be analyzed; According to the frequency of any two symptom manifestation categories appearing in the training sample data and the target sample data, and the target correlation coefficients corresponding to the two symptom manifestation categories, determine the combined correlation coefficient of the two symptom manifestation categories; based on the combined correlation coefficient of any two symptom manifestation categories in any decision tree, and the target correlation coefficients of all symptoms for the target syndrome, determine the decision weight of the decision tree; Performing decision weighting on each decision tree according to the decision weight, constructing a syndrome recognition model, and performing intelligent syndrome recognition of the four-diagnosis medical records according to the syndrome recognition model; Determining the frequency adjustment parameters of the symptoms to be analyzed in the target sample data includes: Determine the frequency and value of the symptoms to be analyzed in other syndromes; Normalize the inverse of the product of the number of syndrome types and the frequency sum value in the target sample data to use as the frequency adjustment parameter; Determining the target correlation coefficient of the symptom to be analyzed according to the frequency of occurrence of the symptom to be analyzed in the target sample data and the frequency adjustment parameter of the symptom to be analyzed in all the target sample data includes: The mean of the frequency adjustment parameters of the symptoms to be analyzed in all the target sample data is used as the overall adjustment parameter of the symptoms to be analyzed; Normalizing the product of the frequency of occurrence of the symptom to be analyzed in the target sample data and the overall adjustment parameter to obtain the target correlation coefficient of the symptom to be analyzed; The method of determining the combined correlation coefficient of two symptom manifestation categories comprises: The sum of the target correlation coefficients of the two symptom manifestation categories was used as the first related influence indicator; Based on the frequency ratio of the two symptoms occurring together and the frequency ratio of the two symptoms occurring separately, the second relevant impact index of the combination of the two symptoms is determined; Calculating the product of the first relevant impact index and the second relevant impact index, and performing maximum and minimum value normalization processing to obtain a combined correlation coefficient; The calculation formula corresponding to the second relevant impact indicator is: In the formula, It represents the second related impact index between the symptom of category i and the symptom of category v, It indicates the frequency of the i-th symptom and the v-th symptom appearing in all target sample data at the same time. represents the frequency of the i-th symptom and the v-th symptom appearing in all training sample data at the same time, Indicates the frequency of occurrence of the ith symptom in all target sample data, Indicates the frequency of occurrence of the vth symptom in all target sample data, represents the frequency of occurrence of the i-th symptom in all training sample data, Represents the frequency of occurrence of the vth type of symptom in all training sample data; The step of determining the decision weight of the decision tree based on the combined correlation coefficient of any two symptom manifestation categories in any decision tree and the target correlation coefficient of all symptoms includes: The mean of the target correlation coefficients of all symptoms in the same decision tree is used as the first weight indicator; The mean of the combined correlation coefficients corresponding to all combinations of any two symptoms in the same decision tree is used as the second weight indicator; The product of the first weight index and the second weight index is calculated, and mapped to [-1, 1] based on the Premnmx function to obtain the decision weight.

2. A method for intelligent identification of TCM syndromes as claimed in claim 1, characterized in that: Determine the information gain index of the symptoms to be analyzed in building the decision tree, including: Based on the ID3 algorithm, the difference between the entropy of the training set of the decision tree and the entropy of the training set of the decision tree under the given symptom to be analyzed is used as the information gain indicator.

3. The method for intelligent identification of TCM syndromes according to claim 1, characterized in that: According to the target correlation coefficient and information gain index, a decision tree is constructed for the symptoms to be analyzed, including: Calculating the product of the target correlation coefficient and the information gain index, and normalizing the result to be used as the division weight of the symptoms to be analyzed for training in the decision tree; Obtain the division weights of all symptoms in the decision tree to be trained, take the symptom with the largest division weight as the first layer to build the decision tree, and then calculate the division weights of all remaining symptoms except those that have participated in the construction of the decision tree, take the largest one as the second layer, and so on to complete the construction of the decision tree.

4. A system for intelligent identification of TCM syndromes, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the steps of the method for intelligent identification of TCM syndromes described in any one of claims 1 to 3 are implemented.

Images