Evaluation method and system of sleep quality regulation in TCM psychosomatic medicine

Abstract

A method for evaluating sleep quality in TCM psychosomatic medicine treatment, comprising: establishing a corresponding first directed weighted graph model and a second directed weighted graph model based on all daily physiological data, physical examination index data and sleep monitoring data as sample data sets; training the first directed weighted graph model and the second directed weighted graph model in combination with an attention mechanism to obtain a first graph embedding matrix and a second graph embedding matrix, respectively; predicting the sample data of each individual based on the first graph embedding matrix and the second graph embedding matrix; obtaining all daily physiological data, physical examination index data and sleep monitoring data of the current individual to be treated, inputting them into the first directed weighted graph model and the second directed weighted graph model, and outputting the value of the corresponding loss function, so as to evaluate the quality of TCM treatment.

Description

Technical Field

[0001] The present invention belongs to the technical field of traditional Chinese medicine psychosomatic medicine, and more specifically, relates to an evaluation method and system for regulating sleep quality in traditional Chinese medicine psychosomatic medicine. Background Art

[0002] Sub-health refers to the intermediate state between illness and health, so it is also called "floating state" and "gray state". Among them, insomnia is one of the most prominent manifestations of sub-health. Sub-health belongs to the category of "not being sick" in traditional Chinese medicine. "Not being sick" does not mean being disease-free, but the precursor state and initial stage of the disease. At this time, the disease has not yet manifested. According to the view of traditional Chinese medicine, the body has already appeared in the imbalance of yin and yang, qi and blood, and viscera and defense when it is sub-healthy. "Suwen·Four Qi Regulating Spirit Theory" says: "The sage does not treat the disease that has already occurred, but treats the disease that has not yet occurred. If the disease has already occurred and then the medicine is given, and the disorder has already occurred and then the treatment is given, it is like digging a well when thirsty, and casting an awl when fighting. Isn't it too late?" It can be seen that traditional Chinese medicine has long recognized the importance of "preparing for a rainy day and preventing trouble before it happens" for diseases.

[0003] Sleep is an indispensable basic physiological activity for the human body. Studies have found that high-quality and sufficient sleep can regulate the repair of the immune system and reduce the possibility of illness. Entering the post-epidemic era, the pace of life has suddenly accelerated, and the "involution" of all walks of life has become more serious, causing great troubles to people's lives, work, and study. Stress insomnia is mainly caused by various mental factors, such as anxiety, excessive thinking, and other related manifestations, which in turn lead to sleep disorders. Studies have found that when the stress factors are removed, about 40% of patients still develop chronic insomnia. The chronicity of insomnia can induce anxiety, depression and other negative emotions, and become a high-risk factor for diseases such as diabetes, hypertension, and coronary atherosclerotic heart disease. Healthy people often have abnormalities in some physical examination indicators, such as abnormal liver function, abnormal kidney function, etc., which will in turn become stress factors to aggravate the original sleep disorders. Some people even have serious psychosomatic problems such as depression and anxiety over time.

[0004] With the development and popularization of IoT technology and wearable devices, technological advances have made it possible to collect and analyze individual health data in real time and continuously monitor individual physiological parameters, thus providing strong data support for personalized medical care and health management. Through these advanced devices, multi-dimensional health information including heart rate, blood pressure, sleep patterns, activity levels, etc. can be captured, providing users and medical professionals with a way to gain a deeper understanding of their personal health status.

[0005] However, current big data and artificial intelligence analysis often lacks personalized considerations for individual specific situations. For the specific problem of sleep disorders, especially in the category of "insomnia" in traditional Chinese medicine, it is necessary to combine modern technology with traditional medicine to develop more refined and personalized solutions. Summary of the invention

[0006] In order to solve the deficiencies in the prior art, the purpose of the present invention is to solve the above-mentioned defects and further propose an evaluation method and system for regulating sleep quality with traditional Chinese medicine psychosomatic medicine.

[0007] The present invention adopts the following technical solution.

[0008] The first aspect of the present invention discloses a method for evaluating sleep quality in TCM psychosomatic medicine, the method comprising steps 1 to 4;

[0009] Step 1: for each individual receiving TCM treatment, based on all their daily physiological data, physical examination index data and sleep monitoring data as sample data sets, respectively establish a corresponding first directed weighted graph model and a second directed weighted graph model;

[0010] Step 2: Combined with the attention mechanism, the first directed weighted graph model and the second directed weighted graph model are trained to obtain the first graph embedding matrix and the second graph embedding matrix respectively;

[0011] Step 3: predict the sample data of each individual based on the first image embedding matrix and the second image embedding matrix;

[0012] Step 4, obtain all daily physiological data, physical examination index data and sleep monitoring data of the individual to be treated, input them into the first directed weighted graph model and the second directed weighted graph model, and output the value of the corresponding loss function to evaluate the quality of TCM treatment.

[0013] Specifically, the sample data used to construct the first directed weighted graph model is the sample data of the individual before receiving TCM treatment; correspondingly, the sample data used to construct the second directed weighted graph model is the sample data of the individual after receiving TCM treatment.

[0014] Specifically, the sample data set can be expressed as Indicates that at time t i The kth type of sample data collected under the above conditions may be heart rate data, sleep time data, liver function index data, etc., where i=1,2,…,N and k=1,2,…,M.

[0015] Specifically, obtaining the first image embedding matrix specifically includes steps 2.1 to 2.6;

[0016] Step 2.1, extract the feature vector of each vertex after mapping in the directed weighted graph according to the data type of the vertex;

[0017] Step 2.2, calculate the neighborhood feature vector of any vertex with respect to any type;

[0018] Step 2.3, calculate the attention score of any vertex with respect to any type according to the neighborhood feature vector of any vertex with respect to any type;

[0019] Step 2.4, based on the attention score of any vertex with respect to any type, calculate the correlation strength between any two vertices in the previous and next time periods;

[0020] Step 2.5, according to the correlation strength between two vertices in the same type in the previous and next time, calculate the weighted feature vector of each type;

[0021] Step 2.6, aggregate all weighted feature vectors to obtain the final first-graph embedding matrix.

[0022] Specifically, each vertex v j The mapped feature vector As shown below:

[0023]

[0024] Among them, U k is the preset weight learning matrix, is the parameter to be learned, and its initial value is randomly set;

[0025] Any vertex v j About the neighborhood eigenvectors of any type k As shown below:

[0026]

[0027] in, is a vertex v in a directed weighted graph j The set of vertices of type k in the neighborhood of j For vertex v j In a directed weighted graph, the adjacency matrix is Represents vertex v u The feature representation after mapping, where u = 1, 2, ..., N × M; A j Represents vertex v j The adjacency matrix of , I is the identity matrix;

[0028] Any vertex v j Attention score for any type k As shown below:

[0029]

[0030] Among them, σ(·) represents the activation function, is the preset attention vector of the kth type, which is the parameter to be learned, and its initial value is randomly set. T represents the transpose of the vector;

[0031] The two vertices in the previous and next time are v i ,v j , the strength of the correlation between them As shown below:

[0032]

[0033] Among them, t ij Represents two vertices v i ,v j The time difference between two vertices v i ,v j The corresponding ones are and Then t ij =t v -t u , l=1,2,…,M;

[0034] The weighted eigenvector Z of the kth type k As shown below:

[0035]

[0036] Among them, z i For vertex v j The weighted eigenvalue of ;

[0037] The first graph embedding matrix G can be expressed as follows:

[0038] Q k =W Q Z k

[0039] K k =W V Z k

[0040] V k =W V Z k

[0041]

[0042] Where d is the dimension of the key vector, W Q ,W K ,W V is the weight matrix in the attention mechanism model, which is the parameter to be learned and its initial value is randomized.

[0043] Specific, predicted results As shown below:

[0044]

[0045] Among them, e and b are the parameters to be learned, where e is a C×M matrix and b is a C×1 vector. C represents the number of types of TCM treatment;

[0046] The loss function Loss is shown as follows:

[0047]

[0048] Among them, Y c Indicates the type of treatment actually selected by the individual. If the treatment type is selected, its value is 1, otherwise it is 0. Representation vector The cth value of .

[0049] Specifically, if the value Loss of the loss function corresponding to the output is less than the preset threshold, it means that the effect of the adjustment has reached the expected result, otherwise it is considered that the effect has not reached the expected result.

[0050] The second aspect of the present invention discloses an evaluation system for regulating sleep quality using traditional Chinese medicine psychosomatic medicine, which is used to execute the method described in the first aspect, and the system comprises: a data acquisition module, a graph algorithm module and a logic calculation module;

[0051] For each individual receiving TCM treatment, the data acquisition module is used to obtain all of their daily physiological data, physical examination index data and sleep monitoring data as a sample data set; and to obtain all of the daily physiological data, physical examination index data and sleep monitoring data of the individual currently to be treated;

[0052] The graph algorithm module is used to establish a corresponding first directed weighted graph model and a second directed weighted graph model; and in combination with the attention mechanism, train the first directed weighted graph model and the second directed weighted graph model to obtain a first graph embedding matrix and a second graph embedding matrix respectively; and predict the sample data of each individual based on the first graph embedding matrix and the second graph embedding matrix;

[0053] The logic calculation module is used to obtain all daily physiological data, physical examination index data and sleep monitoring data of the individual to be treated, input them into the first directed weighted graph model and the second directed weighted graph model, and output the value of the corresponding loss function to evaluate the quality of TCM treatment.

[0054] The third aspect of the present invention discloses a terminal, including a processor and a storage medium; the characteristics are:

[0055] The storage medium is used to store instructions;

[0056] The processor is used to operate according to the instructions to execute the steps of the method described in the first aspect.

[0057] The fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0058] The beneficial effects of the present invention are that, compared with the prior art, the present invention has the following advantages:

[0059] (1) The present invention eliminates the problem of inequality between daily physiological data and physical examination index data by establishing a directed weighted graph network model, fully utilizes daily physiological data, and makes the learning and training of the model more sufficient. At the same time, the present invention evaluates the results of TCM treatment through the value of the loss function, eliminating the problem that the output set of sample data is a matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of a method for evaluating sleep quality in TCM psychosomatic medicine. DETAILED DESCRIPTION

[0061] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

[0062] In the process of integrating IoT technology with TCM psychosomatic medicine to evaluate sleep quality, the conventional TCM psychosomatic medicine sleep quality evaluation method can be: first, the individual's daily physiological data (such as heart rate, blood pressure, blood oxygen saturation, etc.) and sleep monitoring data (such as total sleep time, sleep latency, habitual sleep efficiency, etc.) should be collected. These data can be obtained in real time through smart wearable devices. Secondly, the daily physiological data and sleep monitoring data are processed and combined with physical examination index data (such as liver function index, kidney function index, thyroid function index, etc.), and incorporated into the TCM syndrome differentiation results to form a set of comprehensive sample data sets. On this basis, through deep learning of a large number of sample data sets of different individuals, a prediction model can be established to evaluate sleep quality and identify influencing factors. Ultimately, the evaluation results obtained by the deep learning model should not only reflect the individual's sleep quality status, but also better serve sleep health management.

[0063] Since the amount of daily physiological data and sleep monitoring data is much greater than the physical examination index data, the existing prediction model cannot effectively utilize daily physiological data and sleep monitoring data. In addition, conventional algorithms cannot meet the needs of further adding the impact of TCM treatment, that is, they do not pay attention to the introduction of TCM treatment data (for example: drug treatment, acupuncture treatment, massage treatment, etc.), so the algorithm needs to be improved.

[0064] In addition, the sample data (s1, s2, ... | y1, y2, ...) often contains specific results (that is, the output data is either a numerical value or a vector). Among them, (s1, s2, ...) represents the input data, and (y1, y2, ...) represents the output data. However, in an embodiment of the present invention, the structure of the sample data is not the same. Daily physiological data and physical examination index data can be regarded as input data, and sleep monitoring data can be regarded as output data. However, sleep monitoring data is essentially a very large matrix, in which the length and width of the matrix are the number of types of sleep monitoring data and the number of measurements, respectively. At this time, the use of traditional deep learning algorithms will not be applicable to the usage scenarios of the present invention.

[0065] Based on this, the present invention discloses a method for evaluating sleep quality in TCM psychosomatic medicine. The method trains daily physiological data, sleep monitoring data, TCM treatment data, and physical examination index data through a graph neural network model, thereby reasonably evaluating sleep quality. Figure 1 As shown, it includes steps 1 to 4.

[0066] Step 1: For each individual receiving TCM treatment, based on all their daily physiological data, physical examination index data and sleep monitoring data as sample data sets, a corresponding first directed weighted graph model and a second directed weighted graph model are established respectively.

[0067] It is understandable that the sample dataset can be represented as Indicates that at time t i The kth type of sample data collected under the above conditions may be heart rate data, sleep time data, liver function index data, etc., where i=1,2,…,N and k=1,2,…,M.

[0068] Among them, the sample data used to construct the first directed weighted graph model is the sample data of the individual before receiving TCM treatment; correspondingly, the sample data used to construct the second directed weighted graph model is the sample data of the individual after receiving TCM treatment.

[0069] It is understandable that no matter what type of sample data is, it should at least include the value and the corresponding time t i .

[0070] For a directed weighted graph G = (V, E), no matter it is the first directed weighted graph or the second directed weighted graph, it at least includes a vertex set V and an edge set E and a corresponding adjacency matrix A. The number of vertices in the vertex set is |V| = N × M. Considering the problem of form matching and for a unified description, each In the following text, it represents a vertex, that is, The value of v j , where j = 1, 2,…, N×M.

[0071] Step 2: Combined with the attention mechanism, the first directed weighted graph model and the second directed weighted graph model are trained to obtain the first graph embedding matrix G and the second graph embedding matrix G′ respectively.

[0072] Taking the first graph embedding matrix G as an example, step 2 may specifically include steps 2.1 to 2.6.

[0073] Step 2.1, according to the data type of the vertex, extract the feature vector of each vertex after mapping in the directed weighted graph.

[0074] More specifically, each vertex v j The mapped feature vector As shown below:

[0075]

[0076] Among them, U k is the preset weight learning matrix, is the parameter to be learned, and its initial value is randomly set.

[0077] Step 2.2, calculate the neighborhood eigenvector of any vertex with respect to any type.

[0078] Specifically, any vertex v j About the neighborhood eigenvectors of any type k As shown below:

[0079]

[0080] in, is a vertex v in a directed weighted graph j The set of vertices of type k in the neighborhood of j For vertex v j In a directed weighted graph, the adjacency matrix is Represents vertex v u The feature representation after mapping, where u = 1, 2, ..., N × M; A j Represents vertex v jThe adjacency matrix of , I is the identity matrix.

[0081] It can be understood that in the adjacency matrix, the edge formed by vertices of the same type is the difference between the values ​​of the two vertices, and the edge formed by vertices of different types is 0, wherein the direction of the edge is from the vertex with an earlier time to the vertex with a later time.

[0082] Step 2.3, according to any vertex v j With respect to the neighborhood feature vector of any type k, the attention score of any vertex with respect to any type is calculated.

[0083] Specifically, any vertex v j Attention score for any type k As shown below:

[0084]

[0085] Among them, σ(·) represents the activation function, || is the concatenation operation, is the preset attention vector of the kth type, is the parameter to be learned, its initial value is randomly set, and T represents the transpose of the vector.

[0086] Step 2.4, according to any vertex v j For any type k of attention score, the strength of the association between any two vertices in the previous and next time is calculated.

[0087] More specifically, the two vertices in the previous and next time are v i ,v j , the strength of the correlation between them As shown below:

[0088]

[0089] Among them, t ij Represents two vertices v i ,v j The time difference between two vertices v i ,v j The corresponding ones are and Then t ij =t v -t u , l=1,2,…,M.

[0090] Step 2.5, calculate the weighted feature vector of each type according to the strength of the correlation between two vertices in the previous and next time under the same type.

[0091] More specifically, the weighted feature vector Z of the kth type k As shown below:

[0092]

[0093] Among them, z i For vertex v j The weighted eigenvalues ​​of .

[0094] Step 2.6, aggregate all weighted feature vectors to obtain the final first graph embedding matrix G.

[0095] In some embodiments, all weighted feature vectors may be aggregated based on an attention mechanism, and the first graph embedding matrix G may be as follows:

[0096] Q k =W Q Z k

[0097] K k =W K Z k

[0098] V k =W V Z k

[0099]

[0100] Where d is the dimension of the key vector, W is a scaling factor used to control the size of the dot product to avoid excessive dot products causing the softmax function to saturate in areas where its gradient is small. Q ,W K ,W V is the weight matrix in the attention mechanism model, is the parameter to be learned, and its initial value is randomized. And the query vector Q is calculated separately k , key vector K k Sum value vector V k .

[0101] Step 3: Based on the first graph embedding matrix G and the second graph embedding matrix G ′ , and make predictions for each individual’s sample data.

[0102] In some embodiments, a softmax function can be used for prediction, and the predicted result is It can be shown as follows:

[0103]

[0104] Among them, e and b are parameters to be learned, where e is a C×M matrix and b is a C×1 vector. C represents the number of types of TCM treatment. For example, TCM treatment includes three major categories: drug treatment, acupuncture treatment, and massage treatment, then C=3. Correspondingly, if the drug treatment category can also include heart-nourishing and tranquilizing drug treatment, heat-clearing and tranquilizing drug treatment, and qi-regulating and tranquilizing drug treatment, then the value of C can be expanded to a larger value (for example, C=5), but the classification granularity of C should depend on the number of sample data.

[0105] In addition, a loss function needs to be constructed to train the parameters to be learned of the directed weighted graph model. The loss function Loss can be shown as follows:

[0106]

[0107] Among them, Y c Indicates the type of treatment actually selected by the individual. If the treatment type is selected, its value is 1, otherwise it is 0. Representation vector The cth value of .

[0108] Step 4, obtain all daily physiological data, physical examination index data and sleep monitoring data of the individual to be treated, input them into the first directed weighted graph model and the second directed weighted graph model, and output the value of the corresponding loss function to evaluate the quality of TCM treatment.

[0109] In step 4, a specific threshold may be pre-set. If the value Loss of the loss function corresponding to the output is less than the threshold, it means that the effect of the adjustment has reached the expected result. Otherwise, it is considered that the effect has not reached the expected result.

[0110] Correspondingly, the present invention also discloses an evaluation system for regulating sleep quality using TCM psychosomatic medicine, comprising: a data acquisition module, a graph algorithm module and a logic calculation module;

[0111] For each individual receiving TCM treatment, the data acquisition module is used to obtain all of their daily physiological data, physical examination index data and sleep monitoring data as a sample data set; and to obtain all of the daily physiological data, physical examination index data and sleep monitoring data of the individual currently to be treated;

[0112] The graph algorithm module is used to establish a corresponding first directed weighted graph model and a second directed weighted graph model; and in combination with the attention mechanism, train the first directed weighted graph model and the second directed weighted graph model to obtain a first graph embedding matrix and a second graph embedding matrix respectively; and predict the sample data of each individual based on the first graph embedding matrix and the second graph embedding matrix;

[0113] The logic calculation module is used to obtain all daily physiological data, physical examination index data and sleep monitoring data of the individual to be treated, input them into the first directed weighted graph model and the second directed weighted graph model, and output the value of the corresponding loss function to evaluate the quality of TCM treatment.

[0114] The applicant of the present invention has made a detailed explanation and description of the implementation examples of the present invention in conjunction with the drawings in the specification. However, those skilled in the art should understand that the above implementation examples are only preferred implementation schemes of the present invention, and the detailed description is only to help readers better understand the spirit of the present invention, but not to limit the scope of protection of the present invention. On the contrary, any improvements or modifications based on the inventive spirit of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for evaluating sleep quality in TCM psychosomatic medicine, characterized in that: The method comprises steps 1 to 4; Step 1: for each individual receiving TCM treatment, based on all their daily physiological data, physical examination index data and sleep monitoring data as sample data sets, establish a corresponding first directed weighted graph model and a second directed weighted graph model respectively; Step 2: Combined with the attention mechanism, the first directed weighted graph model and the second directed weighted graph model are trained to obtain the first graph embedding matrix and the second graph embedding matrix respectively; Step 3: predict the sample data of each individual based on the first image embedding matrix and the second image embedding matrix; Step 4, obtaining all daily physiological data, physical examination index data and sleep monitoring data of the individual to be treated, inputting them into the first directed weighted graph model and the second directed weighted graph model, and outputting the value of the corresponding loss function to evaluate the quality of TCM treatment; The sample dataset is represented as Indicates that at time t i The kth type of sample data collected under the above conditions is the heart rate data, sleep time data, and liver function index data, where i = 1, 2, ..., N, k = 1, 2, ..., M, N and M are the number of different times and different types of data collected, respectively. In the following text, each represents a vertex. The value of v j , where j = 1, 2, ..., N × M; Obtaining the first image embedding matrix specifically includes steps 2.1 to 2.6; Step 2.1, extract the feature vector of each vertex after mapping in the directed weighted graph according to the data type of the vertex; Step 2.2, calculate the neighborhood feature vector of any vertex with respect to any type; Step 2.3, calculate the attention score of any vertex with respect to any type according to the neighborhood feature vector of any vertex with respect to any type; Step 2.4, based on the attention score of any vertex with respect to any type, calculate the correlation strength between any two vertices in the previous and next time periods; Step 2.5, according to the correlation strength between two vertices in the same type in the previous and next time, calculate the weighted feature vector of each type; Step 2.6, aggregate all weighted feature vectors to obtain the final first-graph embedding matrix; Each vertex v j The mapped feature vector As shown below: Among them, U k is the preset weight learning matrix, is the parameter to be learned, and its initial value is randomly set; Any vertex v j About the neighborhood eigenvectors of any type k As shown below: in, is a vertex v in a directed weighted graph j The set of vertices of type k in the neighborhood of j For vertex v j In a directed weighted graph, the adjacency matrix is Represents vertex v u The feature representation after mapping, where u = 1, 2, ..., N × M; A j Represents vertex v j The adjacency matrix of , I is the identity matrix; Any vertex v j Attention score for any type k As shown below: Among them, σ(·) represents the activation function, || is the concatenation operation, is the preset attention vector of the kth type, which is the parameter to be learned, and its initial value is randomly set. T represents the transpose of the vector; The two vertices in the previous and next time are v i ,v j , the strength of the correlation between them As shown below: Among them, t ij Represents two vertices v i ,v j The time difference between two vertices v i ,v j The corresponding and t ij =t v -t u , l=1,2,…,M; The weighted eigenvector Z of the kth type k As shown below: WITH k =[…,of i ,…],i∈k Among them, z i For vertex v j The weighted eigenvalue of , i∈k represents the vertex v i The type is k, Z k represents all z under type k i A collection of; The first image embedding matrix G is as follows: Q k =W Q Z k K k =In K WITH k V k =W v Z k Where d is the dimension of the key vector, W Q ,W K ,W V is the weight matrix in the attention mechanism model, which is the parameter to be learned and its initial value is randomized; Predicted results As shown below: Among them, e and b are the parameters to be learned, G ′ is the embedding matrix of the second graph, where e is a C×M matrix, b is a C×1 vector, and C represents the number of types of TCM treatment, including drug treatment, acupuncture treatment, and massage treatment; The loss function Loss is shown as follows: Among them, Y c Indicates the type of treatment actually selected by the individual. If the treatment type is selected, its value is 1, otherwise it is 0. Representation vector The cth value of ; The sample data used to construct the first directed weighted graph model is the sample data of the individual before receiving TCM treatment; correspondingly, the sample data used to construct the second directed weighted graph model is the sample data of the individual after receiving TCM treatment; If the value Loss of the loss function corresponding to the output is less than the preset threshold, it means that the effect of the adjustment has reached the expected result, otherwise it is considered that the effect has not reached the expected result.

2. A sleep quality assessment system for TCM psychosomatic medicine, characterized in that: The system is used to execute the method according to claim 1, characterized in that the system comprises: a data acquisition module, a graph algorithm module and a logic calculation module; For each individual receiving TCM treatment, the data acquisition module is used to obtain all of their daily physiological data, physical examination index data and sleep monitoring data as a sample data set; and to obtain all of the daily physiological data, physical examination index data and sleep monitoring data of the individual currently to be treated; The graph algorithm module is used to establish a corresponding first directed weighted graph model and a second directed weighted graph model; and in combination with the attention mechanism, train the first directed weighted graph model and the second directed weighted graph model to obtain a first graph embedding matrix and a second graph embedding matrix respectively; and predict the sample data of each individual based on the first graph embedding matrix and the second graph embedding matrix; The logic calculation module is used to obtain all daily physiological data, physical examination index data and sleep monitoring data of the individual to be treated, input them into the first directed weighted graph model and the second directed weighted graph model, and output the value of the corresponding loss function to evaluate the quality of TCM treatment.

3. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to claim 1 .

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to claim 1 are implemented.

Images