Traditional Chinese medicine health-preserving scheme personalized recommendation method and system based on big data

Abstract

According to the big data-based traditional Chinese medicine health-preserving scheme personalized recommendation method and system provided by the invention, the interest preference and demand of the user in the field of traditional Chinese medicine health-preserving knowledge are accurately captured by constructing the first health-preserving interest map of the target user. The knowledge strengthening coefficient and the interest mapping factor are utilized to perform deep mining and personalized matching on knowledge nodes in the traditional Chinese medicine health preserving knowledge field, and an optimized health preserving interest map is generated, so that the accuracy and the personalized degree of traditional Chinese medicine health preserving scheme recommendation are improved, the recommendation result is continuously iterated and optimized through a loop optimization mechanism, and the recommendation efficiency is improved. The traditional Chinese medicine health maintenance demand of the target user is effectively met, and the practicability and user satisfaction of the health maintenance scheme are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for personalized recommendation of traditional Chinese medicine health regimen based on big data. Background Art

[0002] With the rapid development of big data technology and people's increasing attention to health and wellness, the field of TCM health care has gradually merged with big data technology, aiming to provide individuals with more accurate and personalized health care plans through data analysis. However, the existing TCM health care recommendation system mostly relies on traditional static recommendation methods or simple data analysis, which makes it difficult to fully capture users' dynamic health needs and interest preferences.

[0003] Traditional TCM health care recommendation methods are often based on fixed recommendation rules or expert experience, lacking in-depth exploration and dynamic adjustment of user personalized needs. At the same time, existing data analysis methods often have difficulty in effectively extracting and utilizing the correlation between knowledge nodes when dealing with the complex knowledge system in the field of TCM health care, resulting in inaccurate and incomplete recommendation results. Summary of the invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for personalized recommendation of TCM health regimen based on big data, the method comprising: According to multiple health feedback data of the target user and the tracking parameters corresponding to each of the health feedback data, a first health interest map in the field of TCM health knowledge of the target user is constructed; the first health interest map includes a first correlation between each knowledge node in the field of TCM health knowledge and the TCM health knowledge portrait of the target user; According to the first relevance and the predefined prior attention trend of each knowledge node, respectively determine the knowledge reinforcement coefficient and the interest mapping factor of the corresponding knowledge node; According to the knowledge enhancement coefficient, all knowledge nodes in the field of TCM health knowledge are subjected to knowledge enhancement processing to generate a plurality of extended mining nodes, and according to the interest mapping factor, interest mapping is performed on the plurality of extended mining nodes to generate an interest mapping vector for each of the extended mining nodes; According to the interest mapping vector of the extended mining node, the first health interest map of the target user is cyclically optimized to generate an optimized health interest map of the TCM health knowledge field of the target user; A TCM health regimen is constructed for the target user based on the optimized health regimen interest graph, and a TCM health regimen recommended to the target user is generated.

[0005] In a possible implementation of the first aspect, constructing a first health interest graph in the field of TCM health knowledge of the target user based on multiple health feedback data of the target user and tracking parameters corresponding to each of the health feedback data includes: Extracting health care scope information and health care monitoring parameters in the field of TCM health care knowledge from the tracking parameters; Determine the TCM health knowledge field of the target user according to the health scope information; Determining attribute information of each knowledge node in the TCM health knowledge field according to the health monitoring parameters; Loading the attribute information and the plurality of health feedback data into a TCM health knowledge association model, generating a first degree of association between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user; By utilizing the first correlation between each knowledge node in the TCM health care knowledge field and the TCM health care knowledge portrait of the target user, a first health care interest graph of the TCM health care knowledge field of the target user is constructed.

[0006] In a possible implementation of the first aspect, determining the knowledge enhancement coefficient and the interest mapping factor of the corresponding knowledge node respectively according to the first association degree and the predefined prior attention trend of each knowledge node includes: Determine the knowledge enhancement coefficient of the corresponding knowledge node according to the first association degree of each knowledge node; According to the first relevance of each knowledge node and a predefined priori attention trend, an interest mapping factor of the corresponding knowledge node is determined.

[0007] In a possible implementation manner of the first aspect, determining the knowledge enhancement coefficient of the corresponding knowledge node according to the first association degree of each knowledge node includes: Get predefined knowledge enhancement functions; Determine the knowledge concentration of the knowledge reinforcement processing in the current round of cyclic optimization; the knowledge concentration is a characteristic value determined based on a priori optimized health interest map, and the priori optimized health interest map is an optimized health interest map generated in a previous round of cyclic optimization before the current round of cyclic optimization; the knowledge concentration includes a first knowledge concentration and a second knowledge concentration, and the first knowledge concentration is greater than the second knowledge concentration; when cyclically optimizing the first health interest map, iteratively determine the knowledge reinforcement coefficient based on the first knowledge concentration and the second knowledge concentration; The knowledge aggregation degree and the first association degree of each knowledge node are loaded into the knowledge enhancement function to generate a knowledge enhancement coefficient of the corresponding knowledge node.

[0008] In a possible implementation manner of the first aspect, determining the interest mapping factor of the corresponding knowledge node according to the first association degree of each knowledge node and a predefined prior attention trend includes: Obtaining a health association vector of each of the knowledge nodes mapped to the TCM health knowledge portrait node of the target user, and determining a deviation value of a deviation between the health association vector and the prior attention trend; Obtain a predefined interest mapping function, and determine candidate adjustment parameters of interest mapping in this round of cyclic optimization; the candidate adjustment parameters are characteristic values ​​determined based on a priori optimized health interest map, and the priori optimized health interest map is an optimized health interest map generated in a previous round of cyclic optimization before this round of cyclic optimization; The candidate adjustment parameters and the deviation value are loaded into the interest mapping function to generate an interest mapping factor of a corresponding knowledge node.

[0009] In a possible implementation of the first aspect, the knowledge enhancement processing is performed on all knowledge nodes in the field of TCM health knowledge according to the knowledge enhancement coefficient to generate multiple extended mining nodes, including: According to the knowledge enhancement coefficient, in the prior attention trend corresponding to the tracking parameter, the knowledge nodes in the field of traditional Chinese medicine health knowledge are subjected to knowledge enhancement processing to generate the multiple extended mining nodes.

[0010] In a possible implementation of the first aspect, based on the knowledge enhancement coefficient, in the prior attention trend corresponding to the tracking parameter, knowledge enhancement processing is performed on the knowledge nodes in the field of traditional Chinese medicine health knowledge to generate the multiple extended mining nodes, including: For a plurality of knowledge nodes on each of the plurality of prior attention trends corresponding to the tracking parameter, obtaining a knowledge reinforcement coefficient of each of the plurality of knowledge nodes; The knowledge nodes whose knowledge reinforcement coefficient is greater than the threshold value are output as expansion mining nodes on the prior focus trend.

[0011] In a possible implementation manner of the first aspect, performing interest mapping on the multiple extended mining nodes according to the interest mapping factor to generate an interest mapping vector for each of the extended mining nodes includes: Encoding and representing the plurality of health feedback data to generate a first encoding and representing vector for each of the extended mining nodes; Acquire a first encoding representation vector of a plurality of extended mining nodes on each prior focus trend and an interest mapping factor of each of the extended mining nodes; Determine a fusion calculation result between the interest mapping factor of each of the extended mining nodes and the first encoding representation vector; According to the prior attention trend, by performing interest mapping on the fusion calculation results, the fusion calculation results of each extended mining node are accumulated respectively, and an interest mapping vector mapped to the TCM health knowledge portrait of the target user is generated.

[0012] In a possible implementation of the first aspect, the step of cyclically optimizing the first health interest graph of the target user according to the interest mapping vector of the extended mining node to generate an optimized health interest graph of the target user's TCM health knowledge field includes: Obtaining an interest mapping vector and an actual interest vector of an extended mining node at the TCM health knowledge portrait of the target user; Determining a characteristic distance between the interest mapping vector and the actual interest vector; Determining an error parameter of a TCM health knowledge association model according to the characteristic distance; the TCM health knowledge association model is used to predict each first association degree in the first health interest graph; Optimizing the model parameter information of the TCM health-preserving knowledge association model according to the error parameter to generate an optimized TCM health-preserving knowledge association model; Utilizing the optimized TCM health knowledge association model to cyclically optimize the first health interest graph of the target user, to generate an optimized health interest graph in the TCM health knowledge field of the target user; The step of constructing a TCM health regimen for the target user based on the optimized health regimen interest graph and generating a TCM health regimen recommended to the target user comprises: Analyze the optimized health care interest map, identify various knowledge nodes and the associations between the various knowledge nodes, and generate a basic framework of a health care plan including multiple health care sections; For each health care section, analyze the depth and breadth of the health care information contained in the corresponding knowledge node in the optimized health care interest map, obtain the knowledge node information, and make preliminary plans for each health care section based on the knowledge node information, generate preliminary planning content for each health care section, integrate the preliminary planning contents of each health care section, and form a complete Chinese medicine health care plan.

[0013] On the other hand, an embodiment of the present invention also provides a personalized recommendation system for TCM health care plans based on big data, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0014] Based on the above aspects, the embodiment of the present application accurately captures the user's interest preferences and needs in the field of TCM health knowledge by constructing the first health interest map of the target user. By using the knowledge reinforcement coefficient and interest mapping factor, the knowledge nodes in the field of TCM health knowledge are deeply mined and personalized matching is performed to generate an optimized health interest map, which not only improves the accuracy and personalization of TCM health plan recommendations, but also continuously iterates and optimizes the recommendation results through a cyclic optimization mechanism, effectively meeting the TCM health needs of the target user and improving the practicality and user satisfaction of the health plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the execution flow of the method for personalized recommendation of TCM health regimen based on big data provided by an embodiment of the present invention.

[0016] Figure 2 It is a schematic diagram of the hardware architecture of a personalized recommendation system for TCM health care solutions based on big data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 The figure is a flow chart of a personalized recommendation method for a TCM health regimen based on big data provided by an embodiment of the present invention. The personalized recommendation method for a TCM health regimen based on big data is introduced in detail below.

[0018] Step S110, constructing a first health interest graph in the TCM health knowledge field of the target user based on multiple health feedback data of the target user and the tracking parameters corresponding to each of the health feedback data. The first health interest graph includes a first correlation between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user.

[0019] In this embodiment, it is assumed that the server receives health feedback data of the target user, which comes from the user's daily health monitoring equipment, such as sleep data and exercise data recorded by a smart bracelet, as well as diet records and physical discomfort symptoms filled out by the user on a regular basis on the health platform. Each health feedback data has corresponding tracking parameters, such as the tracking parameters of sleep data may include sleep time, sleep quality score, deep sleep duration, etc.; the tracking parameters of exercise data may include exercise type, exercise duration, exercise intensity, etc.; the tracking parameters of diet records include food types, intake, etc.

[0020] The server first extracts the health range information and health monitoring parameters in the field of TCM health knowledge from these tracking parameters. For example, by analyzing the user's diet records, if it is found that the user consumes a lot of spicy food and is prone to getting angry, the health range information may involve the concept of "getting angry" in TCM, which belongs to the category of dietary health in the field of TCM health knowledge. The health monitoring parameters may be the specific intake and frequency of spicy food.

[0021] Determine the TCM health knowledge field of the target user based on the health scope information. For example, the diet mentioned above is related to "getting angry", so it is determined to be in the field of TCM diet health. Then, based on the health monitoring parameters, determine the attribute information of each knowledge node in the TCM health knowledge field. In the field of TCM diet health, knowledge nodes may include the nature and taste of food (such as cold, hot, etc.), the efficacy of food (such as clearing heat and reducing fire, nourishing yin and moisturizing dryness, etc.), and the attribute information is related to the user's specific diet data. For example, the attribute of the peppers that users often eat is hot and may cause getting angry.

[0022] The server loads these attribute information and multiple health feedback data into the TCM health knowledge association model. This model is pre-built and contains a large amount of TCM health knowledge and the associations between them. For example, the model stores the association between cold foods and clearing heat and reducing fire, as well as the adaptability relationship with people of different constitutions. By loading the user's pepper intake (attribute information) and other health feedback data (such as symptoms of getting angry) into the model, the first correlation between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user can be generated. For example, the correlation between the knowledge node of pepper and the user's TCM health knowledge portrait may be high because it is closely related to the user's symptoms of getting angry; while the correlation between the knowledge node of cold foods such as mung beans and the user's portrait may be low because the user has no relevant needs. Finally, these correlations are used to construct the first health interest map in the TCM health knowledge field of the target user. This map is like a map, showing the degree of correlation between each knowledge node (such as different foods, health methods, etc.) and the user's TCM health knowledge portrait.

[0023] Step S120, determining the knowledge enhancement coefficient and interest mapping factor of the corresponding knowledge node respectively according to the first relevance and the predefined prior attention trend of each knowledge node.

[0024] In this embodiment, after the server has constructed the first health interest map, this step is started. Let's first look at the part of determining the knowledge reinforcement coefficient. Assume that the predefined knowledge reinforcement function is (f(x, y)), where (x) represents the knowledge aggregation and (y) represents the first correlation. In this round of cyclic optimization stage, the server must determine the knowledge aggregation of the knowledge reinforcement processing. This knowledge aggregation is based on the characteristic value determined by the prior optimized health interest map. The prior optimized health interest map is an optimized health interest map generated in the previous round of cyclic optimization stage before this round of cyclic optimization. For example, suppose that the prior optimized health interest map shows that in the field of diet and health, the aggregation of food nutrition knowledge nodes was higher, while the aggregation of food cooking method knowledge nodes was lower. The knowledge aggregation here includes a first knowledge aggregation and a second knowledge aggregation. The first knowledge aggregation is greater than the second knowledge aggregation. For example, the first knowledge aggregation may be the aggregation for the core health knowledge node, and the second knowledge aggregation is the aggregation for the auxiliary health knowledge node. When the first health interest graph is optimized in a loop, the server iteratively determines the knowledge reinforcement coefficient based on the first knowledge aggregation and the second knowledge aggregation. For example, for a knowledge node of chili peppers that has a high correlation with the user's internal heat symptoms (the first correlation), if it is in an area where the previous knowledge aggregation was high (the area corresponding to the first knowledge aggregation), then loading the knowledge aggregation and the first correlation of this knowledge node into the knowledge reinforcement function (f(x, y)) may generate a higher knowledge reinforcement coefficient; while for a knowledge node of mung bean that has a low correlation and is in an area with a low knowledge aggregation, the generated knowledge reinforcement coefficient is lower.

[0025] Let's look at the part of determining the interest mapping factor. The server obtains the health association vector of each knowledge node mapped to the TCM health knowledge portrait node of the target user. For example, for the knowledge node of pepper, its health association vector may include the association between the stimulation to the body, the impact on the digestive system, and the user's health knowledge portrait node (such as internal heat constitution, gastrointestinal function, etc.). Determine the deviation value of the deviation between this health association vector and the prior attention trend. Assuming that the prior attention trend is to tend to reduce the intake of irritating foods to improve internal heat symptoms, and the health association vector of pepper indicates that it is an irritating food, then the deviation value will be relatively large. At the same time, the server obtains the predefined interest mapping function (g(x, y)) and determines the candidate adjustment parameters of the interest mapping in this round of cyclic optimization. This candidate adjustment parameter is also a feature value determined based on the prior optimization health interest map. For example, the prior optimization health interest map shows that the sensitivity to food adjustment in improving internal heat symptoms is high, and this sensitivity is part of the candidate adjustment parameter. Then, the candidate adjustment parameters and the deviation value are loaded into the interest mapping function (g(x, y)) to generate the interest mapping factor of the corresponding knowledge node. For the chili knowledge node, due to the large deviation value and the influence of the candidate adjustment parameters, a large interest mapping factor may be generated, indicating that it will receive more attention and adjustment in subsequent processing.

[0026] Step S130, based on the knowledge enhancement coefficient, all knowledge nodes in the field of traditional Chinese medicine health knowledge are subjected to knowledge enhancement processing to generate multiple extended mining nodes, and based on the interest mapping factor, interest mapping is performed on the multiple extended mining nodes to generate an interest mapping vector for each of the extended mining nodes.

[0027] In this embodiment, the server performs knowledge reinforcement processing on the knowledge nodes in the field of traditional Chinese medicine health knowledge in the prior attention trend corresponding to the tracking parameters based on the knowledge reinforcement coefficient, and generates multiple extended mining nodes. For example, for the multiple knowledge nodes on each prior attention trend in the multiple prior attention trends corresponding to the tracking parameters, first look at the knowledge nodes related to the prior attention trend of diet and health. The server obtains the knowledge reinforcement coefficient of each of these knowledge nodes. For the previously mentioned pepper knowledge node, its knowledge reinforcement coefficient is relatively high. Assuming that the threshold value is set to 0.5, the knowledge reinforcement coefficient of the pepper knowledge node is 0.8, which is greater than the threshold value, then the server outputs it as an extended mining node on the prior attention trend (diet and health). The knowledge reinforcement coefficient of the mung bean knowledge node is 0.3, which is less than the threshold value, and it is not output as an extended mining node.

[0028] Next, the server performs interest mapping on multiple extended mining nodes according to the interest mapping factor, and generates an interest mapping vector for each extended mining node. The server encodes and represents multiple health feedback data, and generates a first encoding representation vector for each extended mining node. For example, for the extended mining node of pepper, the user's health feedback data on the symptoms of getting angry (such as the degree of oral inflammation, constipation, etc.) is encoded to form a first encoding representation vector. At the same time, the first encoding representation vectors of multiple extended mining nodes on each prior attention trend and the interest mapping factor of each extended mining node are obtained. For example, under the prior attention trend of diet and health, in addition to pepper, there are other extended mining nodes such as ginger (assuming that ginger also meets the conditions to become an extended mining node), and their respective first encoding representation vectors and interest mapping factors are obtained. Then determine the fusion calculation result between the interest mapping factor and the first encoding representation vector of each extended mining node. For the pepper extended mining node, its interest mapping factor may be specially fused with the encoding part related to the symptoms of getting angry in the first encoding representation vector, which may be weighted summation or other specific calculation methods. Finally, according to the prior attention trend, the fusion calculation results are mapped to the interest, and the fusion calculation results of each extended mining node are accumulated to generate an interest mapping vector mapped to the TCM health knowledge portrait of the target user. For example, under the prior attention trend of dietary health, the fusion calculation results of extended mining nodes such as peppers and ginger are accumulated according to certain rules, and finally the interest mapping vector mapped to the user's TCM health knowledge portrait is obtained. This interest mapping vector can reflect the comprehensive impact of these extended mining nodes on the user's health.

[0029] Step S140, based on the interest mapping vector of the extended mining node, the first health interest map of the target user is cyclically optimized to generate an optimized health interest map of the TCM health knowledge field of the target user.

[0030] In this embodiment, the server first obtains the interest mapping vector and actual interest vector of the extended mining node at the TCM health knowledge portrait of the target user. For example, for the previous pepper extended mining node, the interest mapping vector may reflect its theoretical impact on the user's health, while the actual interest vector may be obtained by analyzing the user's actual dietary choice behavior (although pepper may have an impact on getting angry, the user still eats it frequently). Then determine the feature distance between the interest mapping vector and the actual interest vector. If the interest mapping vector indicates that peppers should be eaten as little as possible to improve the symptoms of getting angry, but the actual interest vector shows that the user eats it frequently, then this feature distance is relatively large.

[0031] The error parameter of the TCM health knowledge association model is determined based on this feature distance. The TCM health knowledge association model is used to predict each first degree of association in the first health interest map. For example, if the feature distance is large, it means that there is a large deviation between the degree of association predicted by the model and the actual situation, then the error parameter will be large. Then, the model parameter information of the TCM health knowledge association model is optimized based on the error parameter to generate an optimized TCM health knowledge association model. For example, the weight parameters of the relationship between food and constitution in the model may be adjusted.

[0032] Finally, the optimized TCM health knowledge association model is used to cyclically optimize the first health interest map of the target user to generate the optimized health interest map of the TCM health knowledge field of the target user. In this process, the optimized health interest map is parsed, various knowledge nodes and the associations between various knowledge nodes are identified, and a basic framework of a health plan containing multiple health sections is generated. For example, the dietary health section, exercise health section, emotional health section, etc. are identified, and the relationship between knowledge nodes in each section (such as food, exercise methods, emotional regulation methods, etc.) is clarified. For each health section, the depth and breadth of the health information contained in the corresponding knowledge node in the optimized health interest map are analyzed to obtain the knowledge node information. For example, in the dietary health section, the depth and breadth of the information such as the nutritional content, efficacy, and applicable population of different food knowledge nodes are analyzed. According to the knowledge node information, each health section is preliminarily planned to generate the preliminary planning content of each health section. For example, in the dietary health section, preliminary plans are made to increase the intake of cold foods and reduce the intake of hot foods, and the preliminary planning contents of each health section are integrated to form a complete Chinese medicine health plan.

[0033] Step S150, constructing a TCM health regimen for the target user based on the optimized health regimen interest graph, and generating a TCM health regimen recommended to the target user.

[0034] In this embodiment, the server constructs a TCM health regimen according to the previously generated optimized health interest map. On the basis of the basic framework of the health regimen, the content of each health regimen section is further refined. Taking the diet health regimen section as an example, specific food recommendations are determined based on the correlation between the food knowledge nodes in the optimized health interest map and the user's health knowledge portrait and other relevant information. If the user is prone to getting angry and has a hot constitution, the correlation between cold foods and user portraits in the optimized health interest map is high, then the diet health regimen section in the health regimen will recommend in detail the intake and frequency of cold foods such as bitter melon and winter melon.

[0035] For the sports and health section, based on the information of the sports knowledge nodes in the health interest map, such as the correlation between the effects of different sports methods (such as Tai Chi, jogging, etc.) on the circulation of Qi and blood and the regulation of physical fitness and the user's health needs, determine the appropriate sports method, exercise time and exercise intensity for the user. If the map shows that Tai Chi has a high correlation with the user's regulation of Qi and blood and soothing emotions, the health plan will recommend that the user practice Tai Chi a certain number of times and for a certain length of time each week.

[0036] In the emotional health section, based on the knowledge nodes of the relationship between emotions and body organs in the health interest map, such as the relationship between long-term anxiety and liver function, users are provided with suggestions for emotional regulation. If the user is under great work pressure and prone to anxiety, the health plan may suggest that the user regulate emotions through meditation, listening to soothing music, etc., to protect liver health.

[0037] Through such detailed planning of each health care section, the server finally generates a complete TCM health care plan for the target users. This TCM health care plan comprehensively considers the user's health feedback data, various knowledge nodes in the field of TCM health care knowledge and the relationship between them, and can provide users with personalized TCM health care guidance.

[0038] Based on the above steps, the embodiment of the present application accurately captures the user's interest preferences and needs in the field of TCM health knowledge by constructing the first health interest map of the target user. By using the knowledge reinforcement coefficient and interest mapping factor, the knowledge nodes in the field of TCM health knowledge are deeply mined and personalized matching is performed to generate an optimized health interest map, which not only improves the accuracy and personalization of TCM health plan recommendations, but also continuously iterates and optimizes the recommendation results through a cyclic optimization mechanism, effectively meeting the TCM health needs of the target user and improving the practicality and user satisfaction of the health plan.

[0039] In a possible implementation, step S110 includes: Step S111, determining the TCM health knowledge field of the target user according to the health scope information.

[0040] Step S112, determining the attribute information of each knowledge node in the TCM health knowledge field according to the health monitoring parameters.

[0041] Step S113, loading the attribute information and the multiple health feedback data into the TCM health knowledge association model, generating a first degree of association between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user.

[0042] Step S114, using the first correlation between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user, construct a first health interest graph in the TCM health knowledge field of the target user.

[0043] In this embodiment, first, the health range information and health monitoring parameters in the field of TCM health knowledge are extracted from the tracking parameters. For example, the target user provides a wealth of health feedback data through smart devices and health record platforms. These data include information such as daily steps, heart rate changes, food types and intake, sleep duration and quality, and each data has a corresponding tracking parameter. From the tracking parameters of the number of steps, the server can extract health range information such as exercise intensity. For example, if the number of steps is continuous and the speed is fast, it may involve the moderate labor category in the exercise health in TCM health; and the health monitoring parameters corresponding to the number of steps may be specific step values, step distribution in different time periods, etc. From the perspective of diet-related tracking parameters, if the user records the types and amounts of vegetables, meat, grains, etc. consumed every day, the health range information may involve the nutritional balance in TCM diet health, and the health monitoring parameters are the actual intake of various foods, intake frequency, etc.

[0044] Next, determine the TCM health knowledge field of the target user based on the health scope information. As mentioned earlier, if the health scope information contains a lot of content such as exercise intensity and number of steps and points to the impact of moderate exercise on the body, then it may be determined that the user is in the TCM sports health knowledge field; if the diet-related health scope information is dominant, such as the type of food and nutritional combination, then it is determined to be in the TCM diet health knowledge field. Assuming that the user's diet record shows that he consumes a lot of greasy food and little dietary fiber, and the health scope information revolves around an unreasonable diet structure, it is clear that the user is in the TCM diet health knowledge field.

[0045] Then, based on the health monitoring parameters, the attribute information of each knowledge node in the field of TCM health knowledge is determined. In the field of TCM diet health, if the health monitoring parameters show that the user consumes a lot of greasy food, the attribute information of the knowledge node of greasy food may be rich in fat, high in calories, and difficult to digest. These attribute information are based on TCM's understanding of food characteristics. For example, greasy food may increase the burden on the spleen and stomach, and there is a connection with the knowledge related to the spleen and stomach function in TCM theory. For another example, if the user's sleep duration is short and the quality is poor, in the field of TCM sleep health knowledge, the attribute information of the knowledge node of sleep is insufficient sleep time, which may affect the conditioning of qi and blood, because in TCM, sleep is closely related to qi and blood, and insufficient sleep may affect the filling and operation of qi and blood.

[0046] After that, the attribute information and multiple health feedback data are loaded into the TCM health knowledge association model to generate the first correlation between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user. This TCM health knowledge association model is a knowledge base that contains a large amount of TCM health knowledge and their interrelationships. For example, for the greasy food knowledge node, its attribute information (high calorie, difficult to digest, etc.) and the user's health feedback data (such as overweight, occasional indigestion symptoms, etc.) are loaded into the model. The model stores information such as the correlation between greasy food and weight, spleen and stomach function in TCM knowledge. By analyzing this information, the correlation between the greasy food knowledge node and the TCM health knowledge portrait of the target user can be obtained. If the user has indigestion symptoms and is overweight, and greasy food is prone to cause these problems, then the correlation between them will be relatively high. For healthy food knowledge nodes such as vegetables, due to the low intake of users and health problems, their correlation with the user's TCM health knowledge portrait may be relatively low, because the user's current health status is related to the lack of vegetable intake, but it is not taken seriously in actual behavior.

[0047] Finally, the first correlation between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user is used to construct the first health interest map in the TCM health knowledge field of the target user. This map is like a map showing the relationship between the user and TCM health knowledge. Taking the TCM diet health field as an example, the greasy food knowledge node in the map may be more prominent in the map due to its high correlation with the user's TCM health knowledge portrait, and is surrounded by related health problems (such as overweight, indigestion) and related concepts in TCM theory (such as spleen and stomach transportation); while the vegetable knowledge node is relatively marginal in the map due to its low correlation. This first health interest map provides a basis for subsequent TCM health analysis and program formulation for users. Through this map, we can intuitively see the degree of correlation between each knowledge node and the user's health status, thereby providing a basis for further adjustment of health strategies.

[0048] In a possible implementation, step S120 includes: Step S121, determining the knowledge enhancement coefficient of the corresponding knowledge node according to the first association degree of each knowledge node.

[0049] Step S122: determining the interest mapping factor of the corresponding knowledge node according to the first relevance of each knowledge node and the predefined prior attention trend.

[0050] In a possible implementation, step S121 includes: Step S1211, obtaining a predefined knowledge enhancement function.

[0051] Step S1212, determine the knowledge concentration of the knowledge enhancement processing in this round of cyclic optimization. The knowledge concentration is a characteristic value determined based on the prior optimized health interest map, and the prior optimized health interest map is an optimized health interest map generated in the previous round of cyclic optimization before this round of cyclic optimization. The knowledge concentration includes a first knowledge concentration and a second knowledge concentration, and the first knowledge concentration is greater than the second knowledge concentration. When the first health interest map is cyclically optimized, the knowledge enhancement coefficient is iteratively determined based on the first knowledge concentration and the second knowledge concentration.

[0052] Step S1213: Load the knowledge aggregation degree and the first association degree of each knowledge node into the knowledge enhancement function to generate a knowledge enhancement coefficient of the corresponding knowledge node.

[0053] In this embodiment, first, the server determines the knowledge enhancement coefficient of the corresponding knowledge node based on the first correlation degree of each knowledge node. In this process, the server must first obtain a predefined knowledge enhancement function. This knowledge enhancement function is pre-set. It is a mathematical expression constructed based on the TCM health knowledge system and the analysis results of a large amount of past user data. It aims to calculate a suitable knowledge enhancement coefficient based on different input parameters. For example, this function may be a complex multivariate function that comprehensively considers multiple factors that affect knowledge enhancement.

[0054] Next, the server determines the degree of knowledge concentration of knowledge reinforcement processing in this round of cyclic optimization. This knowledge concentration is based on the characteristic value determined by the prior optimization health interest map, which is the optimized health interest map generated in the previous round of cyclic optimization before this round of cyclic optimization. Taking the aforementioned field of Chinese medicine diet health as an example, it is assumed that the prior optimization health interest map shows that the degree of concentration of knowledge nodes for the nutritional components of food is high, while the degree of concentration of knowledge nodes for the cooking methods of food is low. The knowledge concentration here includes the first knowledge concentration and the second knowledge concentration, and the first knowledge concentration is greater than the second knowledge concentration. The first knowledge concentration may represent the degree of concentration of more core and key knowledge nodes in the health system. For example, in Chinese medicine diet health, the degree of concentration of knowledge nodes related to basic nutritional characteristics such as food nature, taste, meridians, etc. may be classified as the first knowledge concentration, because these characteristics directly affect the effect of food on human health; while the second knowledge concentration may be the degree of concentration of relatively minor but influential knowledge nodes, such as the degree of concentration of knowledge nodes related to information such as food color and origin. When the first health interest graph is cyclically optimized, the server iteratively determines the knowledge reinforcement coefficient based on the first knowledge concentration and the second knowledge concentration.

[0055] Taking a specific knowledge node as an example, suppose there is a knowledge node of greasy food in the field of TCM diet and health. If the knowledge node of greasy food has a high first correlation with the TCM health knowledge portrait of the target user in the previous first health interest graph, and it is in the area with high first knowledge concentration in the prior optimization health interest graph, for example, it is closely related to the food nutrition knowledge node (because the high fat and other nutrients of greasy food are its important characteristics). The server loads the knowledge concentration (here is the high first knowledge concentration) and the first correlation of the greasy food knowledge node into the knowledge reinforcement function. Assuming that the knowledge reinforcement function is (f(x, y)), where (x) represents the knowledge concentration and (y) represents the first correlation, the corresponding knowledge reinforcement coefficient will be generated by function calculation (f(high knowledge concentration, high first correlation)). This coefficient will reflect the degree to which the greasy food knowledge node should be strengthened in the subsequent health knowledge optimization process. Because it has a high correlation and is in a high knowledge concentration area, the generated knowledge reinforcement coefficient may be relatively high.

[0056] In a possible implementation, step S122 includes: Step S1221, obtaining the health association vector of each of the knowledge nodes mapped to the TCM health knowledge portrait node of the target user, and determining the deviation value of the deviation between the health association vector and the prior attention trend.

[0057] Step S1222, obtain a predefined interest mapping function, and determine candidate adjustment parameters of interest mapping in this round of cyclic optimization. The candidate adjustment parameters are feature values ​​determined based on a priori optimized health interest map, which is an optimized health interest map generated in a previous round of cyclic optimization before this round of cyclic optimization.

[0058] Step S1223: Load the candidate adjustment parameter and the deviation value into the interest mapping function to generate an interest mapping factor of the corresponding knowledge node.

[0059] On the other hand, the server determines the interest mapping factor of the corresponding knowledge node based on the first relevance of each knowledge node and the predefined prior attention trend. The server first obtains the health association vector of each knowledge node mapped to the TCM health knowledge portrait node of the target user. Taking the greasy food knowledge node in the field of TCM diet health as an example, its health association vector may include its impact on human blood lipid levels, the burden on the spleen and stomach digestive function, and the association with obesity. These aspects are associated with the TCM health knowledge portrait node of the target user (such as the user's blood lipid status, spleen and stomach function, weight status, etc.). Then determine the deviation value of the deviation between this health association vector and the prior attention trend. Assuming that the prior attention trend is developing in the direction of improving the user's blood lipid level, reducing the burden on the spleen and stomach, and controlling weight, and the health association vector of greasy food indicates that it has an effect on increasing blood lipids, aggravating the burden on the spleen and stomach, and may cause weight gain, then this deviation value will be relatively large.

[0060] At the same time, the server needs to obtain a predefined interest mapping function and determine the candidate adjustment parameters of the interest mapping in this round of cyclic optimization. This interest mapping function is also pre-built and is used to calculate the interest mapping factor based on the input parameters. The candidate adjustment parameter is a feature value determined based on the prior optimization health interest map, which is an optimized health interest map generated in the previous round of cyclic optimization before this round of cyclic optimization. For example, the prior optimization health interest map may show a high sensitivity to food adjustment in improving blood lipids, and this sensitivity may become part of the candidate adjustment parameter. Assume that the interest mapping function is (g(x, y)), where (x) represents the candidate adjustment parameter and (y) represents the deviation value. The server loads the candidate adjustment parameter (such as information such as food adjustment sensitivity) and the deviation value (the deviation of the greasy food health association vector from the prior attention trend) into the interest mapping function (g(x, y)), and then generates the interest mapping factor of the corresponding knowledge node (such as the greasy food knowledge node). Due to the large deviation of greasy food from the prior attention trend and the influence of candidate adjustment parameters (e.g., high sensitivity of food adjustment), the generated interest mapping factor may be relatively large, which indicates that in the subsequent health knowledge processing, the knowledge node of greasy food needs to be adjusted more according to the interest mapping factor to better meet the health goals.

[0061] Taking the exercise intensity knowledge node in the field of TCM sports health as an example, if the exercise intensity knowledge node has a low first correlation with the target user's TCM health knowledge profile in the first health interest map, and it is in an area with a low second knowledge concentration in the prior optimization health interest map, for example, it is close to the relatively minor environmental factor knowledge node in sports health. Loading this low knowledge concentration and low first correlation into the knowledge reinforcement function (f(x, y)), the generated knowledge reinforcement coefficient may be relatively low.

[0062] For the determination of the interest mapping factor of the exercise intensity knowledge node, its health association vector may include the impact on physical fatigue recovery, the pressure on the cardiovascular system and other aspects of the association with the target user's TCM health knowledge portrait node (such as the user's fatigue recovery ability, cardiovascular health status, etc.). Assuming that the prior attention trend is to develop in the direction of improving fatigue recovery ability and reducing cardiovascular system pressure, if the exercise intensity is too high, its health association vector will deviate from the prior attention trend. The server obtains the interest mapping function (g(x, y)), and determines the candidate adjustment parameters of the interest mapping (such as the sensitivity to exercise intensity adjustment in improving cardiovascular health, etc.) based on the prior optimization of the health interest map, loads the candidate adjustment parameters and the deviation value into the interest mapping function, and generates the interest mapping factor of the exercise intensity knowledge node. If the deviation value is small and the sensitivity in the candidate adjustment parameter is low, then the generated interest mapping factor may be relatively small, which means that in the subsequent health knowledge processing, the adjustment range of this knowledge node is relatively small.

[0063] In the entire TCM health knowledge system, the determination process of the knowledge reinforcement coefficient and interest mapping factor of different knowledge nodes is similar. In this way, the server can accurately determine the knowledge reinforcement coefficient and interest mapping factor of each knowledge node based on the correlation between the knowledge node and the user's health profile and the prior attention trend, thereby providing an important basis for the subsequent optimization of health knowledge and the formulation of personalized health plans. For example, in the field of TCM emotional health, the knowledge node of the emotional regulation method (such as regulating anxiety through meditation) determines its knowledge reinforcement coefficient and interest mapping factor based on its first correlation with the user's health profile (such as the severity of the user's anxiety), the knowledge aggregation degree in the prior optimization health interest map (such as the importance of emotional regulation in the entire health system), and the prior attention trend (such as the direction of alleviating anxiety and improving mental health). In this way, the server can perform targeted processing on the knowledge nodes in different health knowledge fields according to these knowledge reinforcement coefficients and interest mapping factors in the subsequent processing to better meet the health needs of the target users.

[0064] In a possible implementation, step S130 includes: According to the knowledge enhancement coefficient, in the prior attention trend corresponding to the tracking parameter, the knowledge nodes in the field of traditional Chinese medicine health knowledge are subjected to knowledge enhancement processing to generate the multiple extended mining nodes.

[0065] In a possible implementation, based on the knowledge enhancement coefficient, in the prior attention trend corresponding to the tracking parameter, knowledge enhancement processing is performed on the knowledge nodes in the field of TCM health knowledge to generate the multiple extended mining nodes, including: Step S131, for a plurality of knowledge nodes on each of a plurality of a priori focus trends corresponding to the tracking parameter, obtain a knowledge reinforcement coefficient of each of the plurality of knowledge nodes.

[0066] Step S132, outputting the knowledge nodes whose knowledge enhancement coefficients are greater than the threshold value as expansion mining nodes on the prior focus trend.

[0067] In this embodiment, first, the server shall, based on the knowledge enhancement coefficient, perform knowledge enhancement processing on the knowledge nodes in the field of TCM health knowledge in the prior attention trend corresponding to the tracking parameter to generate multiple extended mining nodes. The prior attention trend corresponding to the tracking parameter here is a direction determined based on the previous analysis of the user's health data and health goals. Taking the field of TCM diet and health as an example, the tracking parameters may include the type of food, intake, nutritional components, etc. The prior attention trend may be developed in the direction of improving the user's blood lipid status, regulating spleen and stomach function, or controlling weight.

[0068] Then, for the multiple knowledge nodes on each prior attention trend in the multiple prior attention trends corresponding to the tracking parameters, the server obtains the knowledge reinforcement coefficient of each of these knowledge nodes. For example, in the field of TCM diet and health preservation, under the prior attention trend towards improving blood lipid status, the knowledge nodes may include various foods such as greasy foods (such as fried foods), high-fiber foods (such as oats), foods rich in unsaturated fatty acids (such as fish oil), etc., and may also include knowledge nodes such as dietary regularity (such as eating small meals and frequent meals). For each such knowledge node, the server has determined its knowledge reinforcement coefficient based on previous calculations. For the greasy food knowledge node, since it may have adverse effects on blood lipids and is closely related to the user's health in previous analysis (such as the user's high blood lipids and frequent consumption of greasy food), its knowledge reinforcement coefficient may be relatively high; and for the high-fiber food knowledge node, since it helps to improve blood lipids and may be lacking in the user's diet (assuming that the user's dietary fiber intake is insufficient), its knowledge reinforcement coefficient may also be high; for the dietary regularity knowledge node, if the user's diet is irregular and this has a certain impact on blood lipids, its knowledge reinforcement coefficient may also be at a high level.

[0069] Finally, the server outputs the knowledge nodes whose knowledge reinforcement coefficient is greater than the threshold value as the extended mining nodes on the prior attention trend. Assuming that the threshold value is set to 0.6, for the knowledge node of greasy food, if its knowledge reinforcement coefficient is 0.8, which is greater than the threshold value, then the server will output the knowledge node of greasy food as the extended mining node on the prior attention trend of improving blood lipid status. This means that in the subsequent process of optimizing health knowledge and formulating health plans, the knowledge node of greasy food will be focused on and further mined for its related knowledge, such as further analyzing the specific mechanism of the effect of greasy food on blood lipids, how to reduce the intake of greasy food, or how to choose relatively healthy substitutes for greasy food. For the knowledge node of high-fiber food, if its knowledge reinforcement coefficient is 0.7, which is also greater than the threshold value, it will also be output as an extended mining node. At this time, for high-fiber food, it may be further mined for the differences in the improvement of blood lipids of different types (such as oats, whole wheat, etc.), the optimal intake, and how to match other foods to better play the effect of improving blood lipids. For the dietary rule knowledge node, if its knowledge reinforcement coefficient is 0.65, which is greater than the threshold value, it will also become an expanded mining node, and then conduct in-depth research on the specific pattern of eating small meals frequently that is suitable for users, such as the interval time between each meal, the approximate food intake ratio of each meal, etc.

[0070] Taking the field of TCM sports health preservation as an example, the tracking parameters may include exercise type, exercise intensity, exercise time, etc. The a priori focus trend assumption is to improve cardiopulmonary function and enhance body flexibility. Knowledge nodes may include aerobic exercise (such as jogging, swimming), strength training (such as weightlifting, push-ups), flexibility training (such as yoga, Pilates), etc. For aerobic exercise knowledge nodes, if the user's cardiopulmonary function is weak and previous analysis shows that aerobic exercise has great potential to improve its cardiopulmonary function, its knowledge reinforcement coefficient may be high; for strength training knowledge nodes, if it has a relatively small effect on improving cardiopulmonary function and has a low correlation with the user's current needs, its knowledge reinforcement coefficient may be low; for flexibility training knowledge nodes, if it directly helps improve body flexibility and meets the user's health needs (such as the user's body flexibility is poor and hopes to improve), its knowledge reinforcement coefficient may be high. Assuming the threshold value is 0.5, and the knowledge reinforcement coefficient of the aerobic exercise knowledge node is 0.7, which is greater than the threshold value, it will be output as an extended mining node on the prior trend of improving cardiopulmonary function, and may subsequently further explore knowledge such as aerobic exercise intensity, time, and frequency that are suitable for improving the user's cardiopulmonary function. If the knowledge reinforcement coefficient of the flexibility training knowledge node is 0.6, which is greater than the threshold value, it will also become an extended mining node, and further explore knowledge such as the targetedness of different yoga poses for improving body flexibility and the best practice time. If the knowledge reinforcement coefficient of the strength training knowledge node is 0.4, which is less than the threshold value, it will not be output as an extended mining node.

[0071] In the field of TCM emotional health, tracking parameters may include emotion types (such as the frequency and degree of anxiety, depression, etc.), coping methods (such as talking, self-regulation, etc.), and the prior attention trend is assumed to reduce anxiety and improve psychological stress resistance. Knowledge nodes may include meditation, deep breathing, social activities, etc. For meditation knowledge nodes, if the user's anxiety is serious and meditation is considered to be an effective way to relieve anxiety, its knowledge reinforcement coefficient may be high; for deep breathing knowledge nodes, if it is helpful to relieve anxiety but the effect is relatively weak, its knowledge reinforcement coefficient may be at a medium level; for social activity knowledge nodes, if it has positive significance in reducing anxiety and improving psychological stress resistance and is in line with the user's situation (such as the user has few social activities and is anxious), its knowledge reinforcement coefficient may be high. Set the threshold value to 0.55, and the knowledge reinforcement coefficient of the meditation knowledge node to 0.7. If it is greater than the threshold value, it will be output as an extended mining node on the prior attention trend of reducing anxiety. After that, it may be further mined, such as the difference in the effect of different meditation methods (such as mindfulness meditation, transcendental meditation) on user anxiety relief, the optimal meditation duration, and other knowledge. If the knowledge enhancement coefficient of a social activity knowledge node is 0.6, which is greater than the threshold value, it will also become an extended mining node to further mine knowledge such as what kind of social activities (such as participating in interest groups and community activities) are most effective in reducing anxiety and improving psychological stress resistance. However, if the knowledge enhancement coefficient of a deep breathing knowledge node is 0.5, which is less than the threshold value, it will not be output as an extended mining node. In this way, the server can filter out knowledge nodes that are worthy of focus and in-depth mining under various prior attention trends from a large number of knowledge nodes, providing a key foundation for subsequent health knowledge optimization and health plan construction.

[0072] In a possible implementation, step S130 further includes: Step S133: Encode and represent the multiple health feedback data to generate a first encoding representation vector for each of the extended mining nodes.

[0073] Step S134, obtaining first encoding representation vectors of multiple extended mining nodes on each prior focus trend and an interest mapping factor of each of the extended mining nodes.

[0074] Step S135, determining a fusion calculation result between the interest mapping factor of each of the extended mining nodes and the first encoding representation vector.

[0075] Step S136, based on the prior attention trend, by performing interest mapping on the fusion calculation results, accumulating the fusion calculation results of each extended mining node respectively, and generating an interest mapping vector mapped to the TCM health knowledge portrait of the target user.

[0076] In this embodiment, first, the server encodes and represents multiple health feedback data to generate the first encoding representation vector of each extended mining node. Taking the field of traditional Chinese medicine diet and health preservation as an example, it is assumed that the previously determined extended mining nodes are greasy food and high-fiber food. The health feedback data includes the user's blood lipid level, weight change, eating habits (such as the type and amount of food for each meal), and some related symptoms of the body (such as indigestion, etc.). For the extended mining node of greasy food, the server will encode these health feedback data. For example, the blood lipid level is encoded according to different intervals, such as the normal range is 0, the mild high is 1, and the moderate high is 2; if the weight change is increased, it is encoded as 1, reduced as -1, and unchanged as 0; the frequency of greasy food intake in eating habits is high and encoded as 1, and low as 0; the presence of indigestion symptoms is encoded as 1, and the absence is encoded as 0. Combining these codes forms the first encoding representation vector of the extended mining node of greasy food, which may be [1, 1, 1, 1] (this is just a simple example, the actual encoding will be more complicated). For the high-fiber food expansion mining node, the health feedback data is encoded according to similar rules. It is assumed that the first encoding representation vector is [-1, 0, 0, 0], because high-fiber food helps lower blood lipids (if blood lipids are high, it is encoded as -1 to indicate an improvement), has little to do with weight gain (encoded as 0), users may consume less (encoded as 0), and has little to do with indigestion (encoded as 0).

[0077] Next, the server obtains the first encoding representation vectors of multiple extended mining nodes on each prior attention trend and the interest mapping factor of each extended mining node. Taking the improvement of blood lipid status as a prior attention trend in the field of traditional Chinese medicine diet and health preservation as an example, for the greasy food extended mining node, assuming that its interest mapping factor is 0.8 (this value is previously determined based on factors such as the deviation between the health preservation association vector and the prior attention trend), its first encoding representation vector is [1, 1, 1, 1]; for the high-fiber food extended mining node, its interest mapping factor is 0.6, and the first encoding representation vector is [-1, 0, 0, 0].

[0078] Then, the server determines the fusion calculation result between the interest mapping factor of each extended mining node and the first coding representation vector. For the greasy food extended mining node, assuming that the weighted multiplication fusion calculation method is adopted (it may actually be a more complex calculation method), the interest mapping factor 0.8 is multiplied by each element in the first coding representation vector [1, 1, 1, 1], and the fusion calculation result is [0.8, 0.8, 0.8, 0.8]. For the high-fiber food extended mining node, the interest mapping factor 0.6 is multiplied by the first coding representation vector [-1, 0, 0, 0], and the fusion calculation result is [-0.6, 0, 0, 0].

[0079] Finally, the server performs interest mapping on the fusion calculation results based on the prior attention trend, accumulates the fusion calculation results of each extended mining node, and generates an interest mapping vector mapped to the TCM health knowledge portrait of the target user. Under the prior attention trend of improving blood lipid status, the server will accumulate the fusion calculation results of the greasy food and high-fiber food extended mining nodes. The accumulation method here may be to add the corresponding elements, add [0.8, 0.8, 0.8, 0.8] of greasy food and [-0.6, 0, 0, 0] of high-fiber food, and get [0.2, 0.8, 0.8, 0.8]. This result is the interest mapping vector mapped to the TCM health knowledge portrait of the target user. This vector reflects the degree of influence of these extended mining nodes on the user's health status under the prior attention trend of improving blood lipid status.

[0080] Looking at the field of TCM sports health preservation, assuming that the prior trend of concern is to improve cardiopulmonary function, the extended mining nodes include aerobic exercise and flexibility training. Health feedback data includes the user's cardiopulmonary function indicators (such as vital capacity, heart rate recovery time, etc.), exercise habits (such as the number of exercises per week, the duration of each exercise, etc.), and physical fatigue. For the aerobic exercise extended mining node, the health feedback data is encoded to obtain the first encoding representation vector, such as [1, 1, 1] (assuming that low vital capacity is encoded as 1 to indicate the need for improvement, the number of exercises is encoded as 1, and the strong fatigue after exercise is encoded as 1), and its interest mapping factor is 0.7; for the flexibility training extended mining node, the first encoding representation vector after encoding is [0, 0, 0] (because flexibility training has little effect on cardiopulmonary function indicators, exercise habits and fatigue are not closely related to it), and the interest mapping factor is 0.3. After fusion calculation, aerobic exercise obtains [0.7, 0.7, 0.7], and flexibility training obtains [0, 0, 0]. After accumulation, [0.7, 0.7, 0.7] is obtained as the interest mapping vector mapped to the TCM health knowledge portrait of the target user.

[0081] In the field of TCM emotional health care, the prior focus trend is to reduce anxiety, and the extended mining nodes are meditation and social activities. Health feedback data includes anxiety level score, social frequency, emotional stability, etc. For the meditation extended mining node, the first encoded representation vector after encoding is [1, 0, 1] (assuming that high anxiety level is encoded as 1, social frequency is not very related to meditation is encoded as 0, and emotional instability is encoded as 1), and the interest mapping factor is 0.8; for the social activity extended mining node, the first encoded representation vector after encoding is [-1, 1, 0] (social activities help to reduce anxiety and are encoded as -1, high social frequency is encoded as 1, and not very related to emotional stability is encoded as 0), and the interest mapping factor is 0.6. After fusion calculation, meditation obtains [0.8, 0, 0.8], and social activities obtain [-0.6, 0.6, 0]. After accumulation, [0.2, 0.6, 0.8] is obtained as the interest mapping vector mapped to the TCM health care knowledge portrait of the target user. Through such an interest mapping process, the server can more accurately grasp the relationship between the extended mining nodes and the health needs of the target users, providing an important basis for the subsequent optimization of the health interest graph and the construction of health plans.

[0082] In a possible implementation, step S140 includes: Step S141, obtaining the interest mapping vector and the actual interest vector of the extended mining node at the TCM health knowledge portrait of the target user.

[0083] Step S142, determining a characteristic distance between the interest mapping vector and the actual interest vector.

[0084] Step S143, determining an error parameter of a TCM health knowledge association model according to the characteristic distance. The TCM health knowledge association model is used to predict each first association degree in the first health interest graph.

[0085] Step S144, optimizing the model parameter information of the TCM health-preserving knowledge association model according to the error parameter to generate an optimized TCM health-preserving knowledge association model.

[0086] Step S145, using the optimized TCM health care knowledge association model to cyclically optimize the first health care interest map of the target user, to generate an optimized health care interest map of the TCM health care knowledge field of the target user.

[0087] Wherein, step S150 includes: Step S151, analyzing the optimized health care interest graph, identifying various knowledge nodes and the associations between various knowledge nodes, and generating a basic framework of a health care plan including multiple health care sections.

[0088] Step S152, for each health care section, analyze the depth and breadth of the health care information contained in the corresponding knowledge node in the optimized health care interest map, obtain the knowledge node information, and make preliminary plans for each health care section based on the knowledge node information, generate preliminary planning content for each health care section, integrate the preliminary planning content of each health care section, and form a complete Chinese medicine health care plan.

[0089] In this embodiment, first, the server obtains the interest mapping vector and actual interest vector of the extended mining node at the TCM health knowledge portrait of the target user. Taking the field of TCM diet and health as an example, the previously determined extended mining nodes such as greasy food and high-fiber food have corresponding interest mapping vectors. Assume that the interest mapping vector for the greasy food extended mining node is [0.2, 0.8, 0.8, 0.8] (this vector is obtained after comprehensive consideration of various factors, indicating the mapping of greasy food in different health aspects). The actual interest vector is obtained by analyzing the user's actual behavior and preferences. For example, by analyzing the user's diet record, it is found that the user actually often eats greasy food, although greasy food may have many disadvantages from a health perspective, then the actual interest vector may be [1, 1, 1, 1] (here it indicates the actual high tendency to greasy food in terms of blood lipids, weight, eating habits, digestion, etc.). For the high-fiber food expansion mining node, the interest mapping vector is assumed to be [-0.6, 0, 0, 0], while the actual interest vector may be [0, 0, 0, 0] because the user may rarely eat high-fiber food and has low actual interest in it in various health-related aspects.

[0090] Next, the server determines the characteristic distance between the interest mapping vector and the actual interest vector. The calculation of this characteristic distance may involve multiple methods, such as the Euclidean distance calculation method. For the greasy food extension mining node, the interest mapping vector [0.2, 0.8, 0.8, 0.8] and the actual interest vector [1, 1, 1, 1] are substituted into the Euclidean distance formula to calculate the distance between the two. This distance value reflects the degree of difference between the theoretical health advice (interest mapping vector) and the actual user behavior (actual interest vector). For the high-fiber food extension mining node, the characteristic distance between its interest mapping vector [-0.6, 0, 0, 0] and the actual interest vector [0, 0, 0, 0] is also calculated.

[0091] Then, the error parameter of the TCM health knowledge association model is determined based on the feature distance. The TCM health knowledge association model is an important tool for predicting each first degree of association in the first health interest map. Since the feature distance reflects the difference between the interest mapping vector and the actual interest vector, the greater the difference, the greater the error that the model may have when predicting the degree of association. For example, if the feature distance of the greasy food extension mining node is large, it means that the TCM health knowledge association model may have a large deviation when predicting the degree of association between greasy food and the TCM health knowledge portrait of the target user, thereby determining a correspondingly larger error parameter based on this feature distance. For the high-fiber food extension mining node, if the feature distance is small, the corresponding error parameter will also be small.

[0092] Afterwards, the model parameter information of the TCM health knowledge association model is optimized based on the error parameter to generate an optimized TCM health knowledge association model. It is assumed that the TCM health knowledge association model contains parameters such as the nutritional components of various foods, their effects on the body, their relationships with different physiques, and model parameter information such as the weight relationship between these parameters. If the error parameter corresponding to greasy food is large, the server may adjust the parameters related to greasy food, such as adjusting the weight parameter of the effect of the high fat content of greasy food on blood lipids, or adjusting the parameters of the relationship between greasy food and spleen and stomach function. For high-fiber foods, if the error parameter is small, only minor adjustments may be made or the relevant parameters may not be adjusted. Through such adjustments, an optimized TCM health knowledge association model is generated.

[0093] Finally, the optimized TCM health knowledge association model is used to cyclically optimize the first health interest map of the target user to generate the optimized health interest map of the TCM health knowledge field where the target user is located. In this process, the optimized health interest map is parsed, various knowledge nodes and the associations between various knowledge nodes are identified, and a basic framework of a health plan containing multiple health sections is generated. For example, in the field of TCM health knowledge, health sections such as diet health, exercise health, and emotional health are identified. In the diet health section, knowledge nodes may include various types of food (such as greasy food and high-fiber food mentioned earlier), dietary rules, etc.; knowledge nodes in the exercise health section may include various exercise methods (such as running, yoga), exercise intensity, etc.; knowledge nodes in the emotional health section include emotional regulation methods (such as meditation, social activities), etc., and the associations between these knowledge nodes are clarified, such as the mutual influence between diet and emotion (bad diet may affect emotions, and emotions may also affect dietary choices), the relationship between exercise and diet in energy metabolism, etc., so as to construct a basic framework of the health plan.

[0094] For each health care section, analyze and optimize the depth and breadth of the health care information contained in the corresponding knowledge node in the health care interest map, obtain the knowledge node information, and make preliminary plans for each health care section based on the knowledge node information, generate the preliminary planning content of each health care section, integrate the preliminary planning content of each health care section, and form a complete TCM health care plan. In the diet health care section, for the greasy food knowledge node, analyze its health care information depth in the optimized health care interest map, such as in-depth understanding of the different types of greasy food (the difference between animal oil greasy food and vegetable oil greasy food), the difference in greasy degree and health impact under different cooking methods, etc.; in terms of breadth, consider the relationship between greasy food and other health care factors, such as the relationship with season, region, and personal physique. Based on this knowledge node information, preliminarily plan the content of the diet health care section, such as recommending to reduce the intake of greasy food, especially animal oil greasy food, which should be strictly controlled in summer and when the physique is hot and humid. For the sports health section, the knowledge nodes of sports methods are analyzed. For example, the depth of health information for running includes the effects of different running speeds and distances on cardiopulmonary function and muscle strength, and the breadth involves the synergy between running and other health methods (such as diet and emotions). Based on this information, the content of the sports health section is preliminarily planned, such as suggestions to choose the appropriate running intensity and distance according to personal physique and health goals, and pay attention to dietary adjustments before and after exercise. For the emotional health section, the knowledge nodes of emotion regulation methods are analyzed. For example, the depth of meditation information includes the differences in the effects of different meditation schools and techniques on emotion regulation, and the breadth involves the mutual influence of meditation and lifestyle (such as diet and exercise). Based on these preliminary plans, the content of each health section is finally integrated to form a complete TCM health plan, such as comprehensive diet, exercise, and emotional advice, to provide users with comprehensive TCM health guidance.

[0095] Figure 2 The hardware structure of the personalized recommendation system 100 for TCM health regimen based on big data provided by the embodiment of the present invention for implementing the personalized recommendation method for TCM health regimen based on big data is shown as follows: Figure 2 As shown, the personalized recommendation system 100 for TCM health regimen based on big data may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0096] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the big data-based TCM health plan personalized recommendation system 100 uses to execute or use to complete the exemplary method described in the present invention.

[0097] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the personalized recommendation method of Chinese medicine health care plan based on big data as described in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0098] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned big data-based personalized recommendation system 100 for traditional Chinese medicine health care plans. The implementation principles and technical effects are similar and will not be repeated in this embodiment.

[0099] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the above-mentioned personalized recommendation method for TCM health care plans based on big data is implemented.

[0100] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A personalized recommendation method for TCM health regimen based on big data, characterized in that: The method comprises: According to multiple health feedback data of the target user and the tracking parameters corresponding to each of the health feedback data, a first health interest map in the field of TCM health knowledge of the target user is constructed; the first health interest map includes a first correlation between each knowledge node in the field of TCM health knowledge and the TCM health knowledge portrait of the target user; According to the first relevance and the predefined prior attention trend of each knowledge node, respectively determine the knowledge reinforcement coefficient and the interest mapping factor of the corresponding knowledge node; According to the knowledge enhancement coefficient, all knowledge nodes in the field of TCM health knowledge are subjected to knowledge enhancement processing to generate a plurality of extended mining nodes, and according to the interest mapping factor, interest mapping is performed on the plurality of extended mining nodes to generate an interest mapping vector for each of the extended mining nodes; According to the interest mapping vector of the extended mining node, the first health interest map of the target user is cyclically optimized to generate an optimized health interest map of the TCM health knowledge field of the target user; A TCM health regimen is constructed for the target user based on the optimized health regimen interest graph, and a TCM health regimen recommended to the target user is generated.

2. The method for personalized recommendation of TCM health regimen based on big data according to claim 1, characterized in that: The step of constructing a first health interest graph in the field of TCM health knowledge of the target user based on the multiple health feedback data of the target user and the tracking parameters corresponding to each of the health feedback data includes: Extracting health care scope information and health care monitoring parameters in the field of traditional Chinese medicine health care knowledge from the tracking parameters; Determine the TCM health knowledge field of the target user according to the health scope information; Determining attribute information of each knowledge node in the TCM health knowledge field according to the health monitoring parameters; Loading the attribute information and the plurality of health feedback data into a TCM health knowledge association model, generating a first degree of association between each knowledge node in the TCM health knowledge field and the TCM health knowledge portrait of the target user; By utilizing the first correlation between each knowledge node in the TCM health care knowledge field and the TCM health care knowledge portrait of the target user, a first health care interest graph of the TCM health care knowledge field of the target user is constructed.

3. The method for personalized recommendation of TCM health regimen based on big data according to claim 1, characterized in that: The step of determining the knowledge enhancement coefficient and interest mapping factor of the corresponding knowledge node according to the first relevance of each knowledge node and the predefined prior attention trend comprises: Determine the knowledge enhancement coefficient of the corresponding knowledge node according to the first association degree of each knowledge node; According to the first relevance of each knowledge node and a predefined priori attention trend, an interest mapping factor of the corresponding knowledge node is determined.

4. The method for personalized recommendation of TCM health regimen based on big data according to claim 3, characterized in that: The step of determining the knowledge enhancement coefficient of the corresponding knowledge node according to the first association degree of each knowledge node includes: Get predefined knowledge enhancement functions; Determine the knowledge concentration of the knowledge reinforcement processing in the current round of cyclic optimization; the knowledge concentration is a characteristic value determined based on a priori optimized health interest map, and the priori optimized health interest map is an optimized health interest map generated in a previous round of cyclic optimization before the current round of cyclic optimization; the knowledge concentration includes a first knowledge concentration and a second knowledge concentration, and the first knowledge concentration is greater than the second knowledge concentration; when cyclically optimizing the first health interest map, iteratively determine the knowledge reinforcement coefficient based on the first knowledge concentration and the second knowledge concentration; The knowledge aggregation degree and the first association degree of each knowledge node are loaded into the knowledge enhancement function to generate a knowledge enhancement coefficient of the corresponding knowledge node.

5. The method for personalized recommendation of TCM health regimen based on big data according to claim 3, characterized in that: The step of determining the interest mapping factor of the corresponding knowledge node according to the first relevance of each knowledge node and a predefined priori attention trend includes: Obtaining a health association vector of each of the knowledge nodes mapped to the TCM health knowledge portrait node of the target user, and determining a deviation value of a deviation between the health association vector and the prior attention trend; Obtain a predefined interest mapping function, and determine candidate adjustment parameters of interest mapping in this round of cyclic optimization; the candidate adjustment parameters are characteristic values ​​determined based on a priori optimized health interest map, and the priori optimized health interest map is an optimized health interest map generated in a previous round of cyclic optimization before this round of cyclic optimization; The candidate adjustment parameters and the deviation value are loaded into the interest mapping function to generate an interest mapping factor of a corresponding knowledge node.

6. The method for personalized recommendation of TCM health regimen based on big data according to claim 1, characterized in that: According to the knowledge enhancement coefficient, all knowledge nodes in the field of TCM health knowledge are subjected to knowledge enhancement processing to generate multiple extended mining nodes, including: According to the knowledge enhancement coefficient, in the prior attention trend corresponding to the tracking parameter, the knowledge nodes in the field of traditional Chinese medicine health knowledge are subjected to knowledge enhancement processing to generate the multiple extended mining nodes.

7. The method for personalized recommendation of TCM health regimen based on big data according to claim 6, characterized in that: According to the knowledge enhancement coefficient, in the prior attention trend corresponding to the tracking parameter, the knowledge nodes in the field of TCM health knowledge are subjected to knowledge enhancement processing to generate the plurality of extended mining nodes, including: For a plurality of knowledge nodes on each of the plurality of prior attention trends corresponding to the tracking parameter, obtaining a knowledge reinforcement coefficient of each of the plurality of knowledge nodes; The knowledge nodes whose knowledge reinforcement coefficient is greater than the threshold value are output as expansion mining nodes on the prior focus trend.

8. The method for personalized recommendation of TCM health regimen based on big data according to claim 7, characterized in that: The performing interest mapping on the plurality of extended mining nodes according to the interest mapping factor to generate an interest mapping vector for each of the extended mining nodes includes: Encoding and representing the plurality of health feedback data to generate a first encoding and representing vector for each of the extended mining nodes; Acquire a first encoding representation vector of a plurality of extended mining nodes on each prior focus trend and an interest mapping factor of each of the extended mining nodes; Determine a fusion calculation result between the interest mapping factor of each of the extended mining nodes and the first encoding representation vector; According to the prior attention trend, by performing interest mapping on the fusion calculation results, the fusion calculation results of each extended mining node are accumulated respectively, and an interest mapping vector mapped to the TCM health knowledge portrait of the target user is generated.

9. The method for personalized recommendation of TCM health regimen based on big data according to any one of claims 1 to 8, characterized in that: The step of cyclically optimizing the first health care interest graph of the target user according to the interest mapping vector of the extended mining node to generate an optimized health care interest graph in the field of traditional Chinese medicine health care knowledge of the target user includes: Obtaining an interest mapping vector and an actual interest vector of an extended mining node at the TCM health knowledge portrait of the target user; Determining a characteristic distance between the interest mapping vector and the actual interest vector; Determining an error parameter of a TCM health knowledge association model according to the characteristic distance; the TCM health knowledge association model is used to predict each first association degree in the first health interest graph; Optimizing the model parameter information of the TCM health-preserving knowledge association model according to the error parameter to generate an optimized TCM health-preserving knowledge association model; Utilizing the optimized TCM health knowledge association model to cyclically optimize the first health interest graph of the target user, to generate an optimized health interest graph in the TCM health knowledge field of the target user; The step of constructing a TCM health regimen for the target user based on the optimized health regimen interest graph and generating a TCM health regimen recommended to the target user comprises: Analyze the optimized health care interest map, identify various knowledge nodes and the associations between the various knowledge nodes, and generate a basic framework of a health care plan including multiple health care sections; For each health care section, analyze the depth and breadth of the health care information contained in the corresponding knowledge node in the optimized health care interest map, obtain the knowledge node information, and make preliminary plans for each health care section based on the knowledge node information, generate preliminary planning content for each health care section, integrate the preliminary planning contents of each health care section, and form a complete Chinese medicine health care plan.

10. A personalized recommendation system for TCM health regimen based on big data, characterized in that: The personalized recommendation system for TCM health care plans based on big data includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the personalized recommendation method for TCM health care plans based on big data as described in any one of claims 1 to 9 above.

Images