Article information pushing method combining article attributes and user preferences

Abstract

The invention provides an article information pushing method combining article attributes and user preferences, and the method comprises the steps: obtaining digestive system disease types, digestive tract absorption capability, digestive juice secretion conditions and physiological feature data related to digestive functions of a plurality of users, grouping the users through a clustering algorithm, user groups with different digestion function characteristics are obtained; the method comprises the following steps: acquiring dosage form, particle size and surface area attribute information of the health-care product, and performing dimensionality reduction on the health-care product by adopting a principal component analysis method according to disintegration time, dissolution rate and osmotic pressure characteristics of the dosage form to obtain a key attribute set of the health-care products of different dosage forms; and acquiring health care product information searched by the user and digestion condition data of the user, and judging whether the user belongs to a group with relatively weak digestive tract absorption capacity or a group with normal digestion function according to a set digestion threshold value.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for pushing item information combining item attributes and user preferences. Background Art

[0002] There are challenges in the personalized matching of health product dosage forms with user digestive functions. Different users have different types of digestive system diseases, digestive tract absorption capacity and digestive juice secretion, and health product dosage forms, particle sizes and surface area properties are also different. It is an important task to accurately evaluate the user's digestive function characteristics and accurately match them with the characteristics of health product dosage forms. Especially for groups with weak digestive tract absorption capacity, it is particularly important to choose a more easily absorbed dosage form. At the same time, the user's digestive function status may change over time, and the actual absorption effect of health products is difficult to accurately evaluate. This makes it difficult to establish a dynamic and personalized recommendation mechanism. Whether based on static evaluation or dynamic tracking, current methods cannot fully take into account the individual differences of users, the characteristics of health products and the dynamic changes of digestive function. These factors together constitute a complex technical problem, that is, how to achieve the optimal matching of health product dosage forms with user digestive functions while considering multiple factors such as individual differences of users, characteristics of health products and dynamic changes, so as to provide a truly personalized and precise health care plan. This means that the convenience and accuracy of collaborative matching in the recommendation of personalized health product dosage forms are currently low. Summary of the invention

[0003] The present invention provides a method for pushing item information combining item attributes and user preferences, which mainly includes:

[0004] Obtain digestive system disease types, digestive tract absorption capacity, digestive fluid secretion and physiological characteristics data related to digestive function of multiple users, group the users by clustering algorithm, and obtain user groups with different digestive function characteristics;

[0005] Obtain the dosage form, particle size, and surface area attribute information of health products. According to the disintegration time, solubility, and osmotic pressure characteristics of the dosage form, use the principal component analysis method to reduce the dimension of health products and obtain the key attribute set of health products with different dosage forms.

[0006] Analyze the characteristics of digestive tract motility frequency, mucosal barrier function, and flora balance state for different digestive function groups. Combined with the adhesion, release control mechanism, and targeted delivery properties of health care product dosage forms, build a correlation matrix between user digestive characteristics and health care product dosage form adaptability.

[0007] According to the association matrix, the key attribute sets of health products with different dosage forms are obtained, and the collaborative filtering algorithm is used to calculate the acceptance and absorption efficiency scores of different user groups for various dosage forms of health products, and the user and dosage form score matrix is ​​obtained. The digestion threshold is set to judge the group with weak digestive tract absorption capacity and the group with normal digestive function;

[0008] Obtain the health product information that the user searches for, as well as the user's digestion data, and determine whether the user belongs to the group with weak digestive tract absorption capacity or the group with normal digestive function based on the set digestion threshold;

[0009] If the user has a weak digestive absorption capacity, then based on the health product information searched by the user and the user and dosage form rating matrix, liquid and soft capsule health products with high solubility, small particle size, and large surface area are recommended first. If the user has normal digestive function, then based on the tolerance of the digestive tract to specific ingredients, various dosage forms of health products suitable for the user's preferences are recommended to obtain the final recommendation strategy.

[0010] Continuously track user feedback on recommended health products, obtain user absorption efficiency evaluation and satisfaction scores, dynamically optimize recommendation strategies, improve the matching degree between health product dosage forms and user digestive characteristics, and provide users with personalized and precise health care plans.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The present invention discloses a method for pushing item information that combines item attributes and user preferences. The method obtains user digestive system data and health product dosage form attributes, and classifies users and health products using clustering and principal component analysis. Then, a correlation matrix between user digestive characteristics and health product dosage form adaptability is constructed, and a collaborative filtering algorithm is used to calculate a scoring matrix and set a digestion threshold. According to the health product information and digestion conditions searched by the user, the user group type is judged and corresponding recommendations are given. For groups with weak digestion and absorption abilities, liquid and soft capsule health products with high solubility and small particle size are preferentially recommended; for groups with normal digestive function, suitable dosage forms are recommended based on their tolerance. The present invention also continuously optimizes the recommendation strategy through a reinforcement learning algorithm, improves the matching degree, and provides users with personalized and precise health care plans, effectively solving the problem of mismatch between health product dosage forms and user digestion characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flowchart of a method for pushing item information that combines item attributes and user preferences.

[0014] Figure 2 A schematic diagram of a method for pushing item information combining item attributes and user preferences according to the present invention.

[0015] Figure 3 It is another schematic diagram of a method for pushing item information combining item attributes and user preferences according to the present invention. DETAILED DESCRIPTION

[0016] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0017] like Figure 1-3 In this embodiment, a method for pushing item information combining item attributes and user preferences may specifically include:

[0018] Step S101, obtaining digestive system disease types, digestive tract absorption capacity, digestive juice secretion and physiological characteristic data related to digestive function of multiple users, grouping the users through a clustering algorithm to obtain user groups with different digestive function characteristics.

[0019] The user's digestive system disease type, digestive tract absorption capacity, digestive fluid secretion and physiological characteristics data were obtained, and the original data were cleaned and preprocessed to obtain the standardized user feature vector. The clustering algorithm was used to analyze the user feature vector, and the optimal number of clusters was determined according to the silhouette coefficient. If the silhouette coefficient reaches the maximum value, the corresponding number of clusters is determined as the optimal value. According to the preliminary grouping results, the boundary users were divided twice in combination with the hierarchical clustering method. The similarity of user characteristics within the group was calculated by the AGNES algorithm, and the similarity was measured by cosine similarity. If the similarity between users is lower than the set threshold, the user is eliminated. According to the clustering result labels, the distribution of digestive system disease types, digestive tract absorption capacity, digestive fluid secretion and physiological characteristics of each group of users were statistically analyzed. The proportion of different digestive system disease types, the mean and standard deviation of digestive tract absorption capacity, the median and interquartile range of digestive fluid secretion and the distribution of main physiological characteristics in each group were calculated. Based on the statistical results, a set of user digestive function characteristics discrimination rules was constructed. New users were screened step by step through the discrimination rule set to determine the group to which the new users belonged.

[0020] Specifically, the user's digestive system disease type, digestive tract absorption capacity, digestive fluid secretion, and physiological characteristics data are obtained, and the original data is cleaned and preprocessed. The missing value ratio threshold is set to 20%, and samples exceeding the threshold are eliminated. The 3-times standard deviation method is used to detect outliers for continuous data, and the frequency analysis method is used to detect outliers for discrete data. The minimum and maximum normalization processing is performed on continuous data, and the one-hot encoding conversion is performed on discrete data to generate standardized user feature vectors. K-means clustering is used to analyze user feature vectors, and the silhouette coefficient is used to determine the optimal number of clusters. By calculating the silhouette coefficient under different numbers of clusters, the number of clusters with the largest silhouette coefficient is selected as the optimal value. The K-means++ method is used to select the initial center point, and each cluster center is iteratively calculated until the change in the center point position is less than the preset threshold or the maximum number of iterations is reached. Users are divided into the nearest cluster according to the Euclidean distance to obtain preliminary user grouping results. Based on the preliminary grouping results, the boundary users are divided twice in combination with hierarchical clustering. The AGNES (Agglomerative Nesting) algorithm was used to gradually merge the most similar clusters starting from a single sample. The similarity of user characteristics within the group was calculated, and the cosine similarity was used to measure the similarity between users. The similarity threshold was set to the 75% quantile of the overall similarity distribution, and similarity filtering was performed to remove users below the threshold. The principal component analysis was performed on the user characteristics of each group to reduce the dimension, and the first two principal components were selected for visualization to obtain the optimized user groups with different digestive function characteristics. According to the clustering result labels, the distribution of digestive system disease types, digestive tract absorption capacity, digestive fluid secretion, and physiological characteristics data of each group of users were statistically analyzed. The proportion of different digestive system disease types in each group, the mean and standard deviation of digestive tract absorption capacity, the median and interquartile range of digestive fluid secretion, and the distribution of major physiological characteristics were calculated. Based on these statistical results, a user digestive function characteristic discrimination rule set was constructed to achieve rapid grouping and feature identification of new users. The discrimination rule set includes four parts: disease type judgment, absorption capacity assessment, secretion classification, and physiological characteristic matching. The group to which the new user belongs is determined by step-by-step screening. When obtaining user digestive system data, we collected disease types such as gastric ulcer and colitis, digestive tract absorption capacity indicators such as glucose absorption rate, digestive juice secretion such as gastric acid secretion, and physiological characteristics such as body mass index. After cleaning the original data, it was found that there were 25% missing values ​​in gastric acid secretion, which exceeded the 20% threshold and these samples were removed. The 3-times standard deviation method was used to detect outliers. For example, an outlier with a body mass index of 45 was removed. Continuous data such as glucose absorption rate were normalized to the minimum and maximum values ​​and mapped to the 0-1 interval. Discrete data such as disease types were encoded with one-hot encoding, such as encoding gastric ulcer as [1,0,0] and colitis as [0,1,0].K-means cluster analysis was performed on the processed user feature vectors, and the silhouette coefficient was used to determine the optimal number of clusters. The silhouette coefficients were calculated when the number of clusters was 2 to 10, and it was found that the maximum silhouette coefficient was 0.68 when the number of clusters was 5, so 5 was selected as the optimal number of clusters. The K-means++ method was used to select the initial center point, and each cluster center was iteratively calculated. The maximum number of iterations was set to 100, and the threshold for the change in the center point position was 0.001. After 87 iterations, the change in the center point position was less than the threshold, and the iteration was stopped. Users were divided into the nearest cluster according to the Euclidean distance to obtain the preliminary grouping results. Based on the preliminary grouping results, the AGNES algorithm was used to perform secondary division on the boundary users. Starting from 5000 single samples, the most similar clusters were gradually merged. The cosine similarity between users was calculated, and the similarity threshold was set to 0.85 (the 75% quantile of the overall similarity distribution). For user pairs with a similarity lower than 0.85, they were divided into different clusters. Principal component analysis was performed on the user features of each group, and the first two principal components were selected for visualization. The first principal component explains 65% of the variance, which is mainly related to the absorption capacity of the digestive tract; the second principal component explains 20% of the variance, which is mainly related to the secretion of digestive juice. According to the clustering results, statistical analysis is performed on each group of users. For example, in the first group, 40% of the patients with gastric ulcers and 35% of the patients with colitis; the average glucose absorption rate is 0.72±0.15; the median gastric acid secretion is 2.5mEq / h, and the interquartile range is 1.2mEq / h; the average body mass index is 23.5±3.2. Based on these statistical results, a set of rules for distinguishing the characteristics of user digestive function is constructed. For example, if the glucose absorption rate of a new user is lower than 0.6 and the gastric acid secretion is less than 1.5mEq / h, it will be classified as a group with weak digestive function.

[0021] Step S102, obtaining the dosage form, particle size, and surface area attribute information of the health care product, and using the principal component analysis method to reduce the dimension of the health care product according to the disintegration time, solubility, and osmotic pressure characteristics of the dosage form to obtain a set of key attributes of health care products with different dosage forms.

[0022] The dosage form information and physicochemical property data of health products are obtained. The physicochemical property data include particle size data, surface area properties, disintegration time determination results, dissolution curve data, and osmotic pressure measurement values. The Z-score method is used to detect outliers in the physicochemical property data. If the absolute value of the Z-score of a data point is greater than the preset threshold, the data point is marked as an outlier. For missing values ​​in the physicochemical property data, the mean of the corresponding attribute is used to fill. The continuous data in the physicochemical property data is standardized to obtain standardized values. The dosage form information is converted into a unique hot encoding to obtain the encoded dosage form data. Based on the standardized values ​​and the encoded dosage form data, a standardized health product feature matrix is ​​generated. The Pearson correlation coefficient between the attributes in the standardized health product feature matrix is ​​calculated to obtain a correlation coefficient matrix. According to the correlation coefficient matrix, the strength of the correlation between the attributes is determined. If the absolute value of the correlation coefficient is greater than the preset threshold, it is determined to be significantly correlated. Select one from the significantly correlated attribute pairs as the input variable of the principal component analysis to obtain the principal component analysis input variable set. The principal component analysis method is used to reduce the dimension of the principal component analysis input variable set. Calculate the covariance matrix of the input variable set of principal component analysis, and solve the eigenvalues ​​and eigenvectors of the covariance matrix. Arrange the eigenvalues ​​in descending order, and calculate the contribution rate and cumulative contribution rate of each eigenvalue. Select the eigenvectors whose cumulative contribution rate reaches the preset threshold as the principal components, and obtain the characteristic representation of health products after dimensionality reduction. Explain the physical meaning of each principal component in the characteristic representation of health products after dimensionality reduction, and determine the key attributes corresponding to each principal component. Analyze the distribution of different dosage forms on the principal components, and calculate the mean and variance of each dosage form on the main principal components. According to the mean and variance, extract the attributes with significant differences between groups, and obtain the key attribute set of health products with different dosage forms.

[0023] Specifically, the dosage form information, particle size data, surface area attributes, disintegration time determination results, dissolution curve data, and osmotic pressure measurement values ​​of health products are obtained. The Z-score method is used to detect outliers, and the threshold is set to ±3. Data points outside this range are marked as abnormal. The attribute mean is used to fill missing values. Min-Max standardization is performed on continuous data such as particle size, surface area, disintegration time, dissolution, and osmotic pressure, and the values ​​are mapped to the [0,1] interval. The discrete data such as dosage form is converted into a unique encoding to generate a standardized health product feature matrix. According to the standardized feature matrix, the Pearson correlation coefficient matrix between each attribute is calculated to determine the strength of the correlation between the attributes. The absolute value of the correlation coefficient is set to be greater than 0.7 as a significant correlation. For significantly correlated attribute pairs, such as particle size and surface area, disintegration time and dissolution, one of them is selected as the input variable of the principal component analysis to reduce redundant information. The input variable set of the principal component analysis is determined by the correlation analysis results, including key characteristics such as dosage form, particle size, disintegration time, and osmotic pressure. The principal component analysis method is used to reduce the dimension of the input variable set. The covariance matrix is ​​calculated, and the eigenvalues ​​and eigenvectors are solved. The eigenvalues ​​are arranged in descending order, and the contribution rate and cumulative contribution rate of each eigenvalue are calculated. The cumulative contribution rate threshold is set to 85%, and the eigenvectors corresponding to the first few eigenvalues ​​whose cumulative contribution rate reaches this threshold are selected as principal components. The original data are projected onto these principal components to obtain the characteristic representation of health products after dimensionality reduction. For the characteristic representation after dimensionality reduction, the physical meaning of each principal component is explained. The key attributes are determined according to the composition of the principal components. For example, the first principal component may mainly reflect the physical properties of the health product, and the second principal component may mainly reflect its chemical properties. Combined with the original dosage form information, the distribution of different dosage forms on each principal component is analyzed. The mean and variance of each dosage form on the main principal components are calculated, and the attributes with significant differences between groups are extracted. By determining the key attribute set of health products with different dosage forms, the corresponding relationship between the dosage form of health products and the key attributes is obtained. For example, tablets may pay more attention to disintegration time, while capsules may pay more attention to dissolution. When acquiring health supplement data, information on 100 different health supplements was collected, including tablets, capsules, and granules. The particle size range was recorded as 50-500 μm, and the surface area distribution was 0.5-5 m 2 / g, disintegration time is 5-30 minutes, dissolution curve data points are 10 time points within 0-120 minutes, and osmotic pressure measurements are in the range of 200-800mOsm / kg. The Z-score method is used to detect outliers, and the threshold is set to ±3. Five abnormal data points are found, such as an outlier with a particle size of 800μm. The attribute mean is used to fill in 3% of missing values. Min-Max standardization is performed on continuous data, such as mapping the particle size of 50-500μm to the interval of 0-1. The dosage form is encoded with one-hot encoding, such as encoding tablets as [1,0,0] and capsules as [0,1,0], generating a 100×8 standardized feature matrix. The Pearson correlation coefficient matrix between the attributes is calculated, and it is found that the correlation coefficient between particle size and surface area is -0.85, and the correlation coefficient between disintegration time and dissolution is -0.78, both exceeding the significant correlation threshold of 0.7. Particle size and disintegration time were selected as the input variables of principal component analysis, and surface area and dissolution were eliminated. Finally, dosage form, particle size, disintegration time, and osmotic pressure were determined as the input variable set of principal component analysis. The input variable set was subjected to principal component analysis, and the four eigenvalues ​​were calculated to be 2.5, 0.8, 0.5, and 0.2, respectively. The cumulative contribution rate was calculated, and it was found that the cumulative contribution rate of the first two eigenvalues ​​was 82.5%, and the third eigenvalue reached 95%, exceeding the threshold of 85%. The eigenvectors corresponding to the first three eigenvalues ​​were selected as principal components. The original data were projected onto these three principal components to obtain a 100×3 feature representation after dimensionality reduction. The physical meaning of the principal components was analyzed, and it was found that the first principal component was mainly composed of particle size and disintegration time, reflecting the physical properties of health products; the second principal component was mainly composed of osmotic pressure, reflecting chemical properties; and the third principal component was mainly composed of dosage form information. Calculate the distribution of different dosage forms on each principal component. For example, the mean of tablets on the first principal component is 0.7 and the variance is 0.1, while the mean of capsules is 0.4 and the variance is 0.08. Extract the attributes with significant differences, determine that the key attributes of tablets are disintegration time, the key attributes of capsules are dissolution, and the key attributes of granules are particle size and osmotic pressure, and form a corresponding relationship between the dosage form of health products and the key attributes.

[0024] Step S103, for different digestive function groups, analyze the characteristics of digestive tract peristalsis frequency, mucosal barrier function, and flora balance, and combine the adhesion, release control mechanism, and targeted delivery ability properties of health care product dosage forms to construct a correlation matrix between user digestive characteristics and health care product dosage form adaptability.

[0025] The digestive tract peristalsis frequency data, mucosal barrier function indicators, and microflora balance parameters of different digestive function groups were obtained. The digestive tract peristalsis frequency data were processed using the Z-score standardization method, the mucosal barrier function indicators were normalized, and the microflora balance parameters were logarithmically transformed and standardized. According to the standardized data, the preliminary mapping relationship between user digestion characteristics and health product dosage form attributes was constructed, and the association rules between digestive tract peristalsis frequency and dosage form release control mechanism, the association rules between mucosal barrier function and dosage form adhesion, and the association rules between microflora balance and targeted delivery ability were set. The principal component analysis was used to reduce the dimension of the adhesion experimental data, release control mechanism test results, and targeted delivery ability evaluation indicators of health product dosage forms, and the principal component with a cumulative contribution rate reaching the preset threshold was selected as the feature vector. The feature vectors were clustered using the K-means clustering algorithm to obtain the clustering results of dosage form attributes. According to the clustering results of dosage form attributes, the Apriori algorithm is used to mine association rules, with user digestion characteristics as the antecedent and dosage form attribute clustering results as the consequent. The support threshold, confidence threshold and minimum lift are set to mine the association rule set that meets the threshold conditions and generate a preliminary association matrix. For the preliminary association matrix, combined with the preset digestion characteristics and dosage form adaptability rules in the knowledge base, the fuzzy comprehensive evaluation method is used to correct the association strength, and adaptability, safety and effectiveness are selected as evaluation indicators. The hierarchical analysis method is used to set weights, calculate the corrected association degree, and construct the final user digestion characteristics and health product dosage form adaptability association matrix.

[0026] Specifically, the digestive tract peristalsis frequency data, mucosal barrier function indicators, and flora balance parameters of different digestive function groups were obtained. The digestive tract peristalsis frequency was processed using the Z-score standardization method, the mucosal barrier function indicators were normalized using the Min-Max standardization, and the flora balance parameters were logarithmically transformed and standardized. Discrete data such as digestive function group types were converted into one-hot encoding to generate a standardized user digestion feature matrix. A preliminary mapping relationship between user digestion characteristics and health product dosage form attributes was constructed based on historical data. The association rules between digestive tract peristalsis frequency and dosage form release control mechanism, the association rules between mucosal barrier function and dosage form adhesion, and the association rules between flora balance and targeted delivery ability were set. These preliminary rules were stored in the knowledge base for subsequent construction and optimization of the association matrix. According to the adhesion experimental data, release control mechanism test results, and targeted delivery ability evaluation indicators of health product dosage forms, principal component analysis was used to reduce the dimension of dosage form attributes. The first few principal components with a cumulative contribution rate of 85% were selected as new feature vectors. K-means clustering was used to cluster the dosage form attributes after dimensionality reduction. The number of clusters was set to 5, and the iterative calculation was performed until the cluster center was stable or the maximum number of iterations reached 100. The clustering results of dosage form attributes were obtained, and the Euclidean distance between different dosage forms was calculated as the similarity index. The Apriori algorithm was used to mine association rules, with user digestion characteristics as the antecedent and the dosage form attribute clustering results as the consequent. The support threshold was set to 0.1, the confidence threshold was set to 0.7, and the minimum lift was set to 1.1. The association rule set that met the threshold conditions was mined to generate a preliminary association matrix. For the preliminary association matrix, the fuzzy comprehensive evaluation method was used to correct the association strength in combination with the preset digestion characteristics and dosage form adaptability rules in the knowledge base. Adaptability, safety, and effectiveness were selected as evaluation indicators, and the weights were set using the hierarchical analysis method. The corrected association degree was calculated to construct the final user digestion characteristics and health product dosage form adaptability association matrix. Data of 1,000 users with different digestive function groups were obtained, including gastrointestinal peristalsis frequency, which ranged from 3 to 12 times / minute, mucosal barrier function index, which ranged from 0 to 100 points, and microbial balance parameters, whose Shannon diversity index was 1.5 to 4.5. Z-score standardization was applied to peristalsis frequency to convert the original value into a standard normal distribution; Min-Max standardization was used for mucosal barrier function index to map the value to the interval of 0 to 1; and microbial balance parameters were logarithmically transformed and standardized. The types of digestive function groups, such as normal, weak, and strong, were one-hot encoded to generate a 1,000×6 standardized user digestive feature matrix. Preliminary mapping relationships were constructed based on historical data, such as setting peristalsis frequency of 5 to 8 times / minute to be related to sustained-release dosage forms, mucosal barrier function index greater than 80 points to be related to mucosal adhesion dosage forms, and microbial diversity index greater than 3.5 to be related to enteric dosage forms.The attribute data of 50 health product dosage forms were collected, including adhesion 0-100%, release control time 0-24 hours, and targeted delivery ability 0-10 points. The principal component analysis was used for dimensionality reduction, and the first two principal components with a cumulative contribution rate of 87% were selected as new features. K-means clustering was applied, and the number of clusters was set to 5. After 76 iterations, the cluster center was stable, and the clustering results of dosage form attributes were obtained. The Apriori algorithm was used for association rule mining, and the support was set to 0.1, the confidence was set to 0.7, and the minimum lift was set to 1.1. 50 association rules that met the conditions were mined. For example, "IF peristalsis frequency = low AND mucosal function = medium THEN dosage form = sustained release", the support was 0.15, and the confidence was 0.82. Combined with the preset rules, the fuzzy comprehensive evaluation method was used to correct the association strength. The weights of fitness, safety, and effectiveness were set to 0.5, 0.3, and 0.2, respectively. The corrected association degree was calculated, and the final association degree of the above rule was 0.78. A 1000 × 50 correlation matrix between user digestive characteristics and health product dosage form compatibility was constructed, with the matrix element value ranging from 0 to 1, indicating the degree of adaptation.

[0027] Step S104, based on the association matrix, a set of key attributes of health products in different dosage forms are obtained, and a collaborative filtering algorithm is used to calculate the acceptance and absorption efficiency scores of different user groups for various dosage forms of health products, to obtain a user and dosage form score matrix, and a digestion threshold is set to judge the group with weak digestive tract absorption capacity and the group with normal digestive function.

[0028] Obtain the association matrix between user digestion characteristics and health product dosage form adaptability, and construct a user and dosage form characteristic matrix by combining the key attribute set of health products with different dosage forms. Use the key attributes of the dosage form to optimize the collaborative filtering algorithm, and construct a dosage form attribute matrix based on the release control mechanism, adhesion, and targeted delivery capability of the dosage form. Combine the user dosage form acceptance score and absorption efficiency score to obtain the user and dosage form score matrix. Set the digestion threshold and determine the initial threshold from the score distribution of historical data. Use five-fold cross validation to try different thresholds on the training set, evaluate the classification effect on the validation set, and select the threshold with the highest F1 score as the final digestion threshold. If the user's comprehensive score for the dosage form is lower than the digestion threshold, it is determined that the user belongs to the group with weak digestive tract absorption capacity. Classify all users according to the digestion threshold to obtain the user digestion capacity classification results.

[0029] Specifically, the association matrix between user digestive characteristics and health product dosage form adaptability is obtained, and the user-dosage form feature matrix is ​​constructed by combining the key attribute sets of health products with different dosage forms. One-Hot encoding is used to process discrete features such as user digestive function type and dosage form category, and Min-Max normalization is used to process continuous features such as gastrointestinal motility frequency and mucosal barrier function indicators, and user digestive characteristics and dosage form characteristics are converted into numerical feature vectors. For the key attributes of the dosage form, such as release control mechanism, adhesion, and targeted delivery ability, the expert scoring method is used to quantify them into values ​​of 0-10, which are integrated into the feature vector to obtain the input data set for subsequent calculations. According to the user and dosage form feature matrix, the collaborative filtering algorithm of matrix decomposition is used, and the user's acceptance and absorption efficiency of the dosage form are considered at the same time. Initialize the user latent vector U and the dosage form latent vector V, and the latent vector dimension is set to 20. The stochastic gradient descent method is used to optimize the objective function, which contains the prediction error term, the regularization term, and the dosage form attribute similarity term. In each iteration, the predicted score is calculated based on the current U and V, and the error is obtained by comparing it with the actual score, and then U and V are updated. The number of iterations is set to 100 or the convergence condition is reached. The comprehensive score of the user for the untried dosage form is predicted by the product of U and V. The collaborative filtering algorithm is optimized using the key attributes of the dosage form. The key attributes of the dosage form, such as release control mechanism, adhesion, and targeted delivery ability, are used as additional features to construct the dosage form attribute matrix F. The attribute term is introduced into the matrix decomposition model, that is, R≈U·V^T+U·F, where R is the user and dosage form rating matrix, and V^T is the transpose of the dosage form latent vector V. By minimizing the new objective function, the user latent vector U, the dosage form latent vector V, and the attribute weight matrix are learned simultaneously. This method combines collaborative filtering and content-based recommendation to improve the prediction accuracy of new dosage forms. The user dosage form acceptance score and the absorption efficiency score are merged to generate the user and dosage form rating matrix. Set the digestion threshold and use the score distribution of historical data to determine the initial threshold. Five-fold cross validation was used to try different thresholds on the training set, and the classification effect was evaluated on the validation set. The threshold with the highest F1 score was selected as the final digestion threshold. The F1 score is the average of precision and recall. Precision refers to the proportion of samples predicted to be positive that are actually positive, and recall refers to the proportion of samples actually positive that are correctly predicted to be positive. If the user's comprehensive score for the dosage form is lower than the threshold, the user is judged to belong to the group with weak digestive tract absorption capacity, otherwise it is judged to be a group with normal digestive function. All users are classified according to the threshold to obtain the user digestion ability classification results. The digestion characteristics of 1,000 users and the adaptability association matrix of 50 health product dosage forms were obtained, and a 1,000×70 user-dosage form feature matrix was constructed in combination with a set of 20 dosage form key attributes.One-hot encoding was performed on the user's digestive function types of normal, weak, and strong, and the three types were converted into the form of [1,0,0], [0,1,0], and [0,0,1]. The peristaltic frequency of the digestive tract, ranging from 3-12 times / minute, was normalized by Min-Max and mapped to the interval of 0-1. For the key attributes of the dosage form, such as release control mechanism 0-10 points, adhesion 0-100%, and targeted delivery ability 0-10 points, the expert scoring method was used to quantify and integrate them into the feature vector. The collaborative filtering algorithm of matrix decomposition was used to initialize the 1000×20 user latent vector U and the 50×20 dosage form latent vector V, and the latent vector dimension was set to 20. The stochastic gradient descent method was used to optimize the objective function, with the learning rate set to 0.01 and the regularization parameter set to 0.1. Iteration was performed 100 times, and the mean square error between the predicted score and the actual score was calculated in each iteration. When the error change was less than 0.0001, it was stopped early. By multiplying U and V, the user's comprehensive score for the untried dosage form is predicted to obtain a 1000×50 prediction score matrix. The 20 key attributes of the dosage form are constructed into a 50×20 dosage form attribute matrix F, and the attribute terms are introduced in the matrix decomposition model, that is, R≈U·V^T+U·F, where R is a 1000×50 user-dosage form rating matrix. Minimize the new objective function and learn the user latent vector U, the dosage form latent vector V and the 20×20 attribute weight matrix at the same time. Based on the score distribution of historical data, the initial digestion threshold is set to 6.5 points, with a full score of 10 points. Using five-fold cross validation, different thresholds are tried in the range of 5.5-7.5 points with a step size of 0.1, and the highest F1 score of 6.8 points is selected as the final digestion threshold. 1,000 users were classified, and those with a comprehensive score below 6.8 were identified as a group with weak digestive tract absorption capacity, and those with a comprehensive score above 6.8 were identified as a group with normal digestive function. Finally, the classification results included 620 normal groups and 380 groups with weak absorption capacity.

[0030] Step S105, obtaining the health care product information searched by the user and the user's digestion data, and judging whether the user belongs to the group with weak digestive tract absorption ability or the group with normal digestive function according to the set digestion threshold.

[0031] The health care product information searched by the user is obtained, and the feature words are extracted from the health care product information using the TF-IDF algorithm, the TF-IDF value of the feature words is calculated, and the first preset feature words with the highest TF-IDF value are selected as key features; based on the key features, combined with the user's historical search behavior, the frequency of occurrence of the user's search words in the past is counted, and the first preset words with the highest frequency of occurrence are selected to form the user's health care product interest vector; the user's digestion status data is obtained, and the digestive enzyme activity, intestinal permeability, and gastrointestinal motility indicators in the digestion status data are normalized using Z-score to generate the user's digestion feature vector; based on the user's digestion feature vector, a decision tree algorithm is used to construct a user digestion ability classification model, and the decision tree algorithm uses the information gain ratio as the feature selection and splitting criterion; if the user's digestion feature vector meets the condition of weak digestive tract absorption capacity, the corresponding content is selected from the preset suggestion library to generate health suggestions, and if the user's digestion feature vector does not meet the condition of weak digestive tract absorption capacity, suggestions for maintaining intestinal health are provided.

[0032] Specifically, the health care product information and digestion data of users are obtained, and the TF-IDF algorithm is used to extract health care product related feature words from the search information, and the TF-IDF value of each word is calculated, and the top 10 words with the highest weight are selected as key features. Combined with the user's historical search behavior, the word frequency statistics method is used to construct the user's health care product interest vector, and the frequency of user search words in the past 30 days is counted, and the top 20 words with the highest frequency are selected to form the interest vector. At the same time, the user's digestion data is standardized, and the digestive enzyme activity, intestinal permeability, gastrointestinal motility and other indicators are standardized by Z-score to generate the user's digestion feature vector. According to the pre-set digestion threshold, the C4.5 decision tree algorithm is used to construct the user's digestion ability classification model. The user's digestion feature vector is used as input, and the information gain ratio is used as the feature selection and splitting criterion. Starting from the root node, the information gain ratio of each feature is calculated, and the feature with the highest gain ratio is selected as the split node. The minimum number of samples is set to 50, the maximum tree depth is set to 5, and the decision tree is constructed recursively. Pruning technology is used to avoid overfitting. The pessimistic pruning method is adopted to evaluate the pruning effect of each non-leaf node from bottom to top. Pruning operation is performed when the error rate after pruning does not exceed the error rate of unpruned nodes plus one standard deviation. Finally, a decision rule set for judging the user's digestive ability is obtained. The constructed decision rule set is used to classify and judge the user's digestive feature vector. Thresholds for multiple digestive indicators are set, including digestive enzyme activity less than 50U / L, intestinal permeability index greater than 0.05, and gastrointestinal motility index less than 70 points. Starting from the root node of the decision tree, the user's feature value is judged step by step until the leaf node is reached. If the condition of weak digestive tract absorption capacity is met, that is, at least two indicators exceed the threshold, the user is marked as a weak absorption group, otherwise it is marked as a normal digestive function group, and the user's digestive ability classification result is obtained. Based on the user's digestive ability classification result, corresponding health advice and dietary guidance are generated. For the group with weak digestive tract absorption capacity, appropriate content is selected from the preset suggestion library, including suggestions such as increasing the intake of easily digestible food, adjusting the dietary structure, and controlling the eating speed. For people with normal digestive function, general suggestions for maintaining intestinal health are provided. At the same time, based on the user's interest vector for health products, health product usage suggestions related to the user's interests are screened out from the suggestion library, such as choosing a suitable dosage form, arranging the time of taking the product reasonably, etc., to form a personalized health guidance plan. For a user's search for "probiotic capsules", keywords are extracted through the TF-IDF algorithm. For example, the TF-IDF values ​​of "probiotics", "intestines", and "digestion" are 0.8, 0.6, and 0.5 respectively, and these are selected as feature words. Analyzing the user's search history over the past 30 days, it is found that "probiotics" appears 50 times, "vitamins" 30 times, and "protein powder" 20 times, and an interest vector is constructed.At the same time, the digestion data of users were collected, including digestive enzyme activity 45U / L, intestinal permeability index 0.06, and gastrointestinal motility index 65 points. The values ​​after Z-score standardization were -1.2, 1.5, and -0.8, respectively. The C4.5 decision tree algorithm was used to construct a classification model, with the standardized digestion index as input, and the information gain ratio threshold was set to 0.1, the minimum number of samples was 50, and the maximum tree depth was 5. After training, the decision rule was obtained that if the digestive enzyme activity <-1.0 and the intestinal permeability >1.0, it was determined to be a weak absorption group. The rules were applied to the user, and it was found that the conditions were met and marked as a weak absorption group. Based on the classification results, relevant content was selected from the suggestion library, such as "select probiotic powder that is easy to digest and absorb, and take it 2-3 times a day", "increase the intake of foods rich in dietary fiber to improve the intestinal environment", etc. Combined with the user's interest in probiotics, the additional suggestion "select a composite probiotic preparation containing lactic acid bacteria and bifidobacteria, which helps to improve the balance of intestinal flora."

[0033] Step S106, if the user belongs to the group with weak digestive tract absorption capacity, then based on the health care product information searched by the user and the user and dosage form rating matrix, liquid and soft capsule health care products with high solubility, small particle size and large surface area are recommended preferentially; if the user belongs to the group with normal digestive function, then based on the tolerance of the digestive tract to specific ingredients, various dosage forms of health care products suitable for the user's preferences are recommended to obtain the final recommendation strategy.

[0034] Obtain the classification results of the user's digestive tract absorption capacity, health product information, and user and dosage form rating matrix, where the health product information is obtained by user search; according to the health product information, use the TF-IDF algorithm to extract health product keywords and attributes, and construct a health product feature vector containing dissolution, particle size, and surface area attribute values; judge the classification results of the user's digestive tract absorption capacity. If the user belongs to a group with weak digestive tract absorption capacity, set the dissolution, particle size, and surface area thresholds to screen out qualified liquid and soft capsule health products; calculate the similarity between users based on the user and dosage form rating matrix, select the user with the highest similarity to construct a neighbor set, combine the rating data of neighbor users and the historical rating of the current user, and use the weighted average method to predict the user's preference score for each dosage form; obtain the user's digestive tract tolerance data for specific ingredients, and use the weighted scoring method to rank the candidate health products, where the weighted scoring method includes product attribute matching, user preference score, and tolerance factor, calculate the comprehensive score of each candidate health product, and select the health product with the highest comprehensive score as the final recommendation list.

[0035] Specifically, the classification results of the user's digestive tract absorption capacity, the searched health care product information and the user and dosage form rating matrix are obtained. The TF-IDF algorithm is used to extract health care product keywords and attributes from the search information, and the health care product feature vector is constructed, including attribute values ​​such as dissolution, particle size, and surface area. At the same time, the user's rating data for different dosage forms are extracted from the user and dosage form rating matrix. A user and ingredient tolerance matrix is ​​established, and the tolerance data is quantified into a score of 0-1 for the subsequent recommendation process. According to the classification results of the user's digestive tract absorption capacity, conditional judgment is performed. If the user belongs to the group with weak digestive tract absorption capacity, specific thresholds are set, including dissolution greater than 80%, particle size less than 100 microns, and surface area greater than 1 square meter / gram. Liquid and soft capsule health care products that meet these conditions are screened out to form a preliminary recommendation list. For such users, the physical and chemical properties of the product are given priority to ensure better absorption effect. If the user belongs to the group with normal digestive function, a user-based collaborative filtering algorithm is used. The similarity between users is calculated based on the user and dosage form rating matrix, and the cosine similarity formula is used for calculation. The 10 users with the highest similarity are selected to construct a neighbor set. Combined with the rating data of neighbor users and the historical ratings of the current user, the weighted average method is used to predict the user's preference score for various dosage forms. Combined with the user's digestive tract tolerance data for specific ingredients, the weighted scoring method is used to rank the candidate health products. The weight distribution is set as 0.4 for product attribute matching, 0.4 for user preference score, and 0.2 for tolerance factor. The comprehensive score of each candidate health product is calculated, taking into account product characteristics, user preferences, and tolerance. The top 5 health products with the highest comprehensive scores are selected as the final recommendation list, ensuring that the recommendation results are in line with the user's digestive ability and preference, and take into account the tolerance to specific ingredients, and obtain personalized recommendation strategies for users with different digestive abilities. For a user's search for "probiotic soft capsules", the TF-IDF algorithm is used to extract keywords, such as "probiotics", "soft capsules", and "intestinal" with TF-IDF values ​​of 0.8, 0.7, and 0.5, respectively. Construct a health product feature vector, including 85% dissolution, 80 microns particle size, and 1.2 square meters / gram surface area. The user's score for soft capsules is extracted from the user and dosage form rating matrix, which is 4.5 points out of a full 5 points. A user and ingredient tolerance matrix is ​​established, and the user's tolerance score for probiotics is 0.9. It is determined that the user belongs to a group with weak digestive tract absorption capacity, so the physical and chemical properties are given priority. Soft capsule health products that meet the conditions, with solubility>80%, particle size<100 microns, and surface area>1 square meter / gram, are screened out to form a preliminary recommendation list, including 5 products. The weighted scoring method is used for the final sorting, and the weights are set as product attribute matching 0.4, user preference score 0.4, and tolerance factor 0.2.Calculate the comprehensive score. For example, the score of a probiotic soft capsule is 0.4×0.95 (attribute matching) + 0.4×0.9 (user preference) + 0.2×0.9 (tolerance) = 0.92. Finally, select the three products with the highest comprehensive scores as the recommendation results to form a personalized recommendation strategy for users with weak digestive ability.

[0036] Step S107, continuously tracking user feedback on recommended health products, obtaining user absorption efficiency evaluation and satisfaction scores, dynamically optimizing recommendation strategies, improving the matching degree between health product dosage forms and user digestive characteristics, and providing users with personalized and accurate health care plans.

[0037] Receive feedback data carrying a user identification number, generate a user feedback feature vector based on the feedback data, the user feedback feature vector includes objective physiological indicators and subjective evaluation data; use the Q-learning reinforcement learning algorithm to build a dynamic optimization model, the dynamic optimization model is trained according to the user feedback feature vector; update the Q value table, the Q value table is obtained according to the training results of the dynamic optimization model; each update uses the formula Q(s, a) = Q(s, a) + α[r + γ*max(Q(s', a'))-Q(s, a)], where Q(s, a) is the Q value of taking action a in state s, a is the action taken in the current state, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ is the new state reached by the agent after executing action a, and a′ is the next state. The actions that can be taken in state s′, max(Q(s', a')) is the Q value of the action a′ with the highest Q value in the next state s′; judging whether the Q value table meets the adjustment conditions, including: querying the recommendation strategy database; wherein the recommendation strategy database includes strategy data with a recommendation strategy identification number; judging whether there is a recommendation strategy identification number matching the Q value table in the recommendation strategy database; if there is a recommendation strategy identification number matching the Q value table in the recommendation strategy database, adjusting the existing recommendation strategy; generating a personalized precision health care plan according to the adjusted recommendation strategy, the personalized precision health care plan including a recommended combination of health products, dosage and time schedule; sending a push message carrying the personalized precision health care plan, the push message is sent by the system to the user terminal through a mobile application.

[0038] Specifically, obtain user feedback data on recommended health products, including absorption efficiency evaluation and satisfaction rating. Use smart bracelets to monitor user heart rate, blood pressure, sleep quality and other obtainable physiological indicators in real time, combined with subjective ratings filled out daily by users through mobile applications, such as digestion status, energy level, etc. Integrate these data into a user feedback feature vector, which includes two parts: objective physiological indicators and subjective evaluation. Set weekly feedback reminders to ensure continuous acquisition of user data. Based on the user feedback feature vector, use the Q-learning reinforcement learning algorithm to build a dynamic optimization model. Define the state space as user digestion characteristics, such as digestibility index, intestinal health score and currently recommended health product combination. The action space includes adjusting the dosage form, adjusting the dosage, changing the type of health products, etc. The reward function is set to the weighted sum of the absorption efficiency score, weight 0.6 and the satisfaction score, weight 0.4. Initialize the Q value table and assign a random value to each state-action pair. Update the Q value through iteration, and use the formula for each update

[0039] Q(s,a)=Q(s,a)+α[r+γ*max(Q(s',a'))-Q(s,a)], where Q(s,a) is the Q value of taking action a in state s, that is, the expected value of the long-term reward that can be obtained by taking a certain action in the current state. s is the current state, which refers to the position or condition of the agent in the environment. a is the action taken in the current state. α is the learning rate, which determines the extent to which new information affects the old Q value. The value range of α is usually (0,1]. If α=1, then the new information will completely replace the old information; if α is close to 0, the impact of the new information is small. r is the immediate reward, which is the reward signal received by the agent from the environment after performing action a. This reward can be positive or negative. It can be negative, depending on the quality of the action. γ is a discount factor used to measure the importance of future rewards. The value range of γ is [0,1]. If γ approaches 0, the agent will only consider immediate rewards; if γ is close to 1, the agent will consider future rewards more. s′ is the next state, that is, the new state reached by the agent after executing action a. a′ is the action that can be taken in the next state s′. max(Q(s',a')) is the Q value of action a′ with the highest Q value among all possible actions in the next state s′. Use the updated Q value table to adjust the existing recommendation strategy. Calculate the matching score for different combinations of user digestive characteristics and health product dosage forms. Use the cosine similarity meter Calculate the similarity between the user feature vector and the health product feature vector to obtain the matching score. Combine the Q value and matching score, use the weighted summation method to comprehensively consider the absorption efficiency, satisfaction and dosage form characteristics, and generate an optimized recommendation list. The matching score weight is set to 0.3, and the Q value weight is set to 0.7. Based on the optimized recommendation list, combined with the user's historical data and current digestive status, generate a personalized and precise health care plan. Select the action with the highest Q value in the current state from the Q value table to determine the recommended health product combination, dosage and schedule. Set the feedback monitoring cycle to one week, and update the user status and Q value table every week. Push personalized plans to users through mobile applications, and remind users to provide timely feedback. Continuously collect user data and review it regularly, such as monthly. A new Q-learning model is trained to achieve dynamic optimization of the recommendation strategy. For a user taking probiotic soft capsules, the average heart rate of 70 beats / min, blood pressure of 120 / 80 mmHg and sleep quality of 7 hours / night are collected through a smart bracelet. The user fills in the digestion status, 8 points / 10 points and energy level, 7 points / 10 points every day through the mobile application. These data are integrated into the user feedback feature vector [70,120,80,7,8,7]. The state space is defined as the user's digestion ability index 0.8 and the currently recommended probiotic soft capsule, 1 billion live bacteria / capsule. The action space includes adjusting the dosage by ±20% and changing the dosage form, tablets / powder. The reward function is set to 0.6×8+0.4×7=7.6.The initial Q value table is randomly assigned between -1 and 1. After 100 iterations, the learning rate α = 0.1, the discount factor γ = 0.9, and the Q value table converges. The user feature vector [0.8, 10] and the health product feature vector are calculated.

[0040] The cosine similarity of [0.9,12] is 0.98. Combining the Q value, with a weight of 0.7 and the matching score, with a weight of 0.3, the optimal recommendation is to increase the dose by 20%. The generated personalized solution is probiotic soft capsules, 1.2 billion live bacteria / capsule, twice a day. Set weekly feedback reminders, continuously collect data and update the Q-learning model monthly to achieve dynamic optimization.

[0041] The above embodiment is only one of the preferred implementation modes of the present invention and should not be used to limit the protection scope of the present invention. Any changes or modifications that are made to the main design concept and spirit of the present invention and have no substantive significance, and the technical problems they solve are still consistent with the present invention, should be included in the protection scope of the present invention.

Claims

1. A method for pushing item information combining item attributes and user preferences, characterized in that: The method comprises: Obtain digestive system disease types, digestive tract absorption capacity, digestive fluid secretion and physiological characteristics data related to digestive function of multiple users, group the users by clustering algorithm, and obtain user groups with different digestive function characteristics; Obtain the dosage form, particle size, and surface area attribute information of health products. According to the disintegration time, solubility, and osmotic pressure characteristics of the dosage form, use the principal component analysis method to reduce the dimension of health products and obtain the key attribute set of health products with different dosage forms. Analyze the characteristics of digestive tract motility frequency, mucosal barrier function, and flora balance state for different digestive function groups. Combined with the adhesion, release control mechanism, and targeted delivery properties of health care product dosage forms, build a correlation matrix between user digestive characteristics and health care product dosage form adaptability. According to the association matrix, the key attribute sets of health products with different dosage forms are obtained, and the collaborative filtering algorithm is used to calculate the acceptance and absorption efficiency scores of different user groups for various dosage forms of health products, and the user and dosage form score matrix is ​​obtained. The digestion threshold is set to judge the group with weak digestive tract absorption capacity and the group with normal digestive function; Obtain the health product information that the user searches for, as well as the user's digestion data, and determine whether the user belongs to the group with weak digestive tract absorption capacity or the group with normal digestive function based on the set digestion threshold; If the user has a weak digestive absorption capacity, then based on the health product information searched by the user and the user and dosage form rating matrix, liquid and soft capsule health products with high solubility, small particle size, and large surface area are recommended first. If the user has normal digestive function, then based on the tolerance of the digestive tract to specific ingredients, various dosage forms of health products suitable for the user's preferences are recommended to obtain the final recommendation strategy. Continuously track user feedback on recommended health products, obtain user absorption efficiency evaluation and satisfaction scores, dynamically optimize recommendation strategies, improve the matching degree between health product dosage forms and user digestive characteristics, and provide users with personalized and precise health care plans.

2. The method according to claim 1, characterized in that The method of obtaining digestive system disease types, digestive tract absorption capacity, digestive fluid secretion and physiological characteristic data related to digestive function of multiple users, and grouping the users by clustering algorithm to obtain user groups with different digestive function characteristics includes: Obtain the user's digestive system disease type, digestive tract absorption capacity, digestive fluid secretion and physiological characteristics data, clean and pre-process the original data, and obtain a standardized user feature vector; A clustering algorithm is used to analyze the user feature vector and the optimal number of clusters is determined based on the silhouette coefficient; If the silhouette coefficient reaches the maximum value, the corresponding number of clusters is determined as the optimal value; According to the preliminary grouping results, the hierarchical clustering method is used to perform secondary division of border users; The similarity of user features within a group is calculated using the AGNES algorithm, and the similarity is measured using cosine similarity; If the similarity between users is lower than the set threshold, the user will be removed; According to the clustering result labels, statistical analysis was performed on the distribution of digestive system disease types, digestive tract absorption capacity, digestive fluid secretion, and physiological characteristics of each group of users; The proportion of different digestive system disease types, the mean and standard deviation of digestive tract absorption capacity, the median and interquartile range of digestive fluid secretion, and the distribution of main physiological characteristics in each group were calculated; Based on the statistical results, a set of rules for distinguishing the characteristics of user digestive functions is constructed; New users are screened step by step through a set of discrimination rules to determine the group to which they belong.

3. The method according to claim 1, characterized in that The dosage form, particle size, and surface area attribute information of the health care product are obtained, and the principal component analysis method is used to reduce the dimension of the health care product according to the disintegration time, solubility, and osmotic pressure characteristics of the dosage form to obtain a set of key attributes of health care products of different dosage forms, including: Obtaining dosage form information and physicochemical property data of health products, including particle size data, surface area properties, disintegration time determination results, dissolution curve data and osmotic pressure measurement values; The Z-score method is used to detect outliers in the physicochemical property data. If the absolute value of the Z-score of a data point is greater than the preset threshold, the data point is marked as an outlier; For missing values ​​in the physical and chemical attribute data, the mean of the corresponding attribute was used to fill them; Standardize the continuous data in the physical and chemical property data to obtain standardized values; Perform one-hot encoding conversion on the dosage form information to obtain the encoded dosage form data; Generate a standardized health product feature matrix based on standardized values ​​and coded dosage form data; Calculate the Pearson correlation coefficients between the attributes in the standardized health product feature matrix to obtain a correlation coefficient matrix; According to the correlation coefficient matrix, determine the strength of the correlation between attributes; If the absolute value of the correlation coefficient is greater than the preset threshold, it is judged to be significantly correlated; Select one from the significantly correlated attribute pairs as an input variable for principal component analysis to obtain a principal component analysis input variable set; The principal component analysis method is used to reduce the dimension of the principal component analysis input variable set; Calculate the covariance matrix of the principal component analysis input variable set and solve the eigenvalues ​​and eigenvectors of the covariance matrix; Arrange the eigenvalues ​​in descending order and calculate the contribution rate and cumulative contribution rate of each eigenvalue; Select the feature vectors whose cumulative contribution rate reaches the preset threshold as the principal component to obtain the feature representation of health products after dimensionality reduction; Explain the physical meaning of each principal component in the feature representation of health products after dimensionality reduction, and determine the key attributes corresponding to each principal component; Analyze the distribution of different dosage forms on the principal components and calculate the mean and variance of each dosage form on the main principal components; According to the mean and variance, the attributes with significant differences between groups are extracted to obtain the key attribute set of health products with different dosage forms.

4. The method according to claim 1, characterized in that: The method analyzes the peristaltic frequency, mucosal barrier function, and flora balance characteristics of the digestive tract for different digestive function groups, and combines the adhesion, release control mechanism, and targeted delivery properties of the health product dosage form to construct a correlation matrix between user digestive characteristics and health product dosage form adaptability, including: The digestive tract motility frequency data, mucosal barrier function indexes and microbiota balance parameters of different digestive function groups were obtained. The digestive tract motility frequency data were processed using the Z-score standardization method, the mucosal barrier function indexes were normalized, and the microbiota balance parameters were logarithmically transformed and then standardized. Based on the standardized data, the preliminary mapping relationship between the user's digestive characteristics and the properties of the health product dosage form is constructed, and the association rules between the gastrointestinal motility frequency and the dosage form release control mechanism, the association rules between the mucosal barrier function and the dosage form adhesion, and the association rules between the microbial balance state and the targeted delivery ability are set; Principal component analysis was used to reduce the dimensionality of the adhesion experimental data, release control mechanism test results and targeted delivery ability evaluation indicators of health care product dosage forms, and the principal components whose cumulative contribution rate reached the preset threshold were selected as feature vectors; The K-means clustering algorithm was used to cluster the feature vectors to obtain the clustering results of dosage form attributes; According to the clustering results of dosage form attributes, the Apriori algorithm is used to mine association rules, with user digestion characteristics as the antecedent and dosage form attribute clustering results as the consequent. The support threshold, confidence threshold and minimum lift are set to mine the association rule set that meets the threshold conditions and generate a preliminary association matrix. For the preliminary association matrix, combined with the preset digestive characteristics and dosage form adaptability rules in the knowledge base, the fuzzy comprehensive evaluation method was used to correct the association strength, and the adaptability, safety and effectiveness were selected as evaluation indicators. The hierarchical analysis method was used to set the weights, and the corrected association degree was calculated to construct the final user digestive characteristics and health product dosage form adaptability association matrix.

5. The method according to claim 1, characterized in that According to the association matrix, the key attribute sets of health products with different dosage forms are obtained, and the collaborative filtering algorithm is used to calculate the acceptance and absorption efficiency scores of different user groups for various dosage forms of health products, to obtain the user and dosage form score matrix, and set the digestion threshold to judge the group with weak digestive tract absorption capacity and the group with normal digestive function, including: Obtain the correlation matrix between user digestive characteristics and health product dosage form adaptability, and construct a user and dosage form feature matrix by combining the key attribute sets of health products with different dosage forms; The collaborative filtering algorithm was optimized using the key attributes of the dosage form to construct a dosage form attribute matrix based on the release control mechanism, adhesion, and targeted delivery capabilities of the dosage form; Combine the user dosage form acceptance score and the absorption efficiency score to obtain the user and dosage form score matrix; Set the digestion threshold and determine the initial threshold from the score distribution of historical data; Five-fold cross validation was used to try different thresholds on the training set, and the classification effect was evaluated on the validation set. The threshold with the highest F1 score was selected as the final digestion threshold. If the user's comprehensive score for the dosage form is lower than the digestion threshold, the user is judged to belong to the group with weak digestive tract absorption capacity; All users are classified according to the digestion threshold to obtain the user digestion ability classification results.

6. The method according to claim 1, characterized in that The obtaining of the health care product information searched by the user and the user's digestion data, and judging whether the user belongs to a group with weak digestive tract absorption capacity or a group with normal digestive function according to a set digestion threshold, includes: Obtain the health care product information searched by the user, use the TF-IDF algorithm to extract feature words from the health care product information, calculate the TF-IDF value of the feature words, and select the first preset feature words with the highest TF-IDF value as key features; Based on the key features and the user's historical search behavior, the frequency of past user search terms is counted, and the pre-set terms with the highest frequency are selected to form the user's health care product interest vector; Obtain user digestion data, and use Z-score standardization to process digestive enzyme activity, intestinal permeability, and gastrointestinal motility indicators in the digestion data to generate a user digestion feature vector; According to the user digestion feature vector, the decision tree algorithm is used to build the user digestion ability classification model. The decision tree algorithm uses the information gain ratio as the feature selection and splitting criterion; If the user's digestive feature vector meets the condition of weak digestive tract absorption capacity, the corresponding content is selected from the preset suggestion library to generate health suggestions. If the user's digestive feature vector does not meet the condition of weak digestive tract absorption capacity, suggestions for maintaining intestinal health are provided.

7. The method according to claim 1, characterized in that If the user belongs to the group with weak digestive tract absorption capacity, then based on the health care product information searched by the user and the user and dosage form rating matrix, liquid and soft capsule health care products with high solubility, small particle size and large surface area are recommended first. If the user belongs to the group with normal digestive function, then based on the tolerance of the digestive tract to specific ingredients, various dosage forms of health care products suitable for the user's preference are recommended to obtain the final recommendation strategy, including: Obtain the classification results of the user's digestive tract absorption capacity, health product information, and user and dosage form rating matrix, where the health product information is obtained by the user's search; According to the health care product information, the TF-IDF algorithm is used to extract health care product keywords and attributes, and a health care product feature vector containing dissolution, particle size and surface area attribute values ​​is constructed; Determine the classification results of the user's digestive tract absorption capacity. If the user belongs to the group with weak digestive tract absorption capacity, set the solubility, particle size and surface area thresholds to screen out liquid and soft capsule health products that meet the conditions; The similarity between users is calculated based on the user and dosage form rating matrix, and the user with the highest similarity is selected to construct a neighbor set. The user's preference score for each dosage form is predicted using a weighted average method based on the neighbor user's rating data and the current user's historical rating. Obtain the user's digestive tract tolerance data to specific ingredients, and use a weighted scoring method to rank the candidate health products, where the weighted scoring method includes product attribute matching, user preference score and tolerance factor. Calculate the comprehensive score of each candidate health product, and select the health products with the highest comprehensive score as the final recommendation list.

8. The method according to claim 1, characterized in that The method of continuously tracking user feedback on recommended health products, obtaining user absorption efficiency evaluation and satisfaction scores, dynamically optimizing recommendation strategies, improving the matching degree between health product dosage forms and user digestive characteristics, and providing users with personalized and accurate health care solutions includes: Receiving feedback data carrying a user identification number, and generating a user feedback feature vector according to the feedback data, wherein the user feedback feature vector includes objective physiological indicators and subjective evaluation data; The Q-learning reinforcement learning algorithm is used to build a dynamic optimization model, which is trained based on the user feedback feature vector; Update the Q value table, which is obtained according to the training results of the dynamic optimization model; Each update uses the formula Q(s,a)=Q(s,a)+α[r+γ*max(Q(s',a'))-Q(s,a)], where Q(s,a) is the Q value of taking action a in state s, a is the action taken in the current state, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ is the new state reached by the agent after taking action a, a′ is the action that can be taken in the next state s′, and max(Q(s',a')) is the Q value of action a′ with the highest Q value in the next state s′; Determining whether the Q value table meets the adjustment condition includes: querying the recommendation strategy database; Wherein, the recommended strategy database includes strategy data with a recommended strategy identification number; Determine whether there is a recommendation strategy identification number matching the Q value table in the recommendation strategy database; If the recommendation strategy database has a recommendation strategy identification number that matches the Q value table, then adjust the existing recommendation strategy; Generate a personalized and precise health care plan based on the adjusted recommendation strategy, which includes a recommended combination of health products, dosage and schedule; Send push information with personalized and precise health care solutions. The push information is sent by the system to the user terminal through the mobile application.

Images