Evaluation method of stratified randomized controlled traditional Chinese medicine compound intervention for prediabetes

Abstract

This invention relates to the field of clinical evaluation technology, specifically a stratified randomized controlled trial method for evaluating the intervention of traditional Chinese medicine compound formulas in prediabetes. The method includes the following steps: constructing stratified codes for subjects based on fasting blood glucose differences; constructing a blood glucose change difference sequence by combining the differences before and after intervention; using a regression function to obtain characteristic inflection points and establish a dose interval mapping relationship; obtaining the peak blood glucose time based on glucose tolerance tests and segmenting the time axis; calculating the slope characteristic parameters of each segment; and classifying and judging the comprehensive parameters and mapping relationship to construct a efficacy grading evaluation index. In this invention, by comparing multiple fasting blood glucose differences with stable parameters to extract fluctuation characteristics, a stratified coding mechanism is constructed to avoid the risk of stratification mismatch in a single test. The intervention difference is extracted and used to introduce piecewise linear regression to calculate the slope and capture the response characteristic inflection points. Combined with standard comparison, the degree of recovery of the body is determined, and the curve shape is transformed into a multi-dimensional efficacy grading index, effectively improving clinical precision.

Description

Technical Field

[0001] This invention relates to the field of clinical evaluation technology, and in particular to a stratified randomized controlled method for evaluating the intervention of traditional Chinese medicine compound prescriptions for prediabetes. Background Technology

[0002] The field of clinical evaluation technology primarily involves a research system that assesses the safety and effectiveness of medical interventions through systematic research methods. This field typically encompasses core aspects such as clinical trial design, participant selection, randomization methods, controlled trials, construction of efficacy evaluation indicator systems, follow-up observation, and statistical data analysis. Its research subjects include drug therapies, medical devices, interventions, and comprehensive treatment plans. Standardized research processes are used to objectively evaluate treatment plans. In this field, researchers typically determine research subjects based on disease diagnostic criteria, screen the population by establishing inclusion and exclusion criteria, and record data on blood glucose levels, physiological indicators, symptom presentation, and quality of life under uniform intervention conditions. Simultaneously, statistical methods are used to compare and analyze differences between different interventions, thus forming a systematic clinical evaluation research system.

[0003] The traditional stratified randomized controlled trial of traditional Chinese medicine (TCM) compound intervention for prediabetes refers to a method that uses TCM compound intervention in the prediabetic population and evaluates the intervention through a clinical research design. This method typically begins by identifying the prediabetic subject population based on fasting blood glucose levels, glucose tolerance test results, and TCM syndrome differentiation. Then, the subjects are stratified according to age range, gender ratio, body mass index, and TCM syndrome differentiation. Within each stratum, subjects are randomly assigned to either a TCM compound intervention group or a control group using a pre-defined randomization sequence. The intervention group receives a compound preparation containing multiple TCM ingredients according to a predetermined prescription, taken at a fixed dosage and for a fixed period. The control group receives basic intervention or other control measures according to a predetermined study protocol. During the study, fasting blood glucose levels, postprandial blood glucose levels, glucose tolerance test results, and related metabolic indicators are collected regularly. Questionnaires are used to record information such as quality of life scores and psychological status scores. Follow-up records are kept on whether subjects develop diabetes or return to normal blood glucose levels during the study period. Finally, the collected clinical data are statistically analyzed and compared to complete a systematic evaluation of the TCM compound intervention for prediabetes.

[0004] Traditional evaluation methods rely on a few fasting blood glucose values ​​to stratify subjects. The conventional operating mode is easily affected by the random fluctuations of a single test, resulting in subject mismatch and disrupting the balance of the trial stratification structure. In the intervention of traditional Chinese medicine compound prescriptions, fixed doses are often used, and efficacy is judged only by the absolute change before and after the intervention. It is difficult to establish a quantitative mapping relationship between dose change and blood glucose regulation intensity. In addition, in the glucose tolerance assessment, only the endpoint blood glucose level is considered, ignoring the peak time axis shift and curve morphology changes, which severely limits the refinement of clinical evaluation. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a stratified randomized controlled method for evaluating the intervention of traditional Chinese medicine compound prescriptions in prediabetes, comprising the following steps: To achieve the above objectives, the present invention employs the following technical solution: a stratified randomized controlled method for evaluating the intervention of traditional Chinese medicine compound prescriptions in prediabetes, comprising the following steps: S1: Collect fasting blood glucose measurements from prediabetic subjects, obtain the maximum fasting blood glucose difference, compare the maximum fasting blood glucose difference with the blood glucose stability interval parameter, obtain the average fasting blood glucose value, compare the average fasting blood glucose value with the preset threshold boundary parameter, and construct subject stratified coding. S2: Based on the subject stratification coding, the dosage gradient of ginseng and astragalus compound was extracted for the traditional Chinese medicine compound intervention for prediabetic subjects. The fasting blood glucose value after the intervention cycle and the baseline blood glucose value before the intervention cycle were collected. The difference was calculated and combined with the dosage gradient of ginseng and astragalus compound to construct a blood glucose change difference sequence. S3: Input the dose gradient of the ginseng and astragalus compound and the difference sequence of blood glucose changes into a piecewise linear regression function to obtain the inflection point of the dose response feature and merge them to construct a dose interval mapping relationship; S4: Collect glucose tolerance test values ​​of prediabetic subjects to obtain peak blood glucose time parameters. Based on the peak blood glucose time parameters, obtain the rising segment time data sequence and the falling segment time data sequence, and calculate the slope to obtain the rising segment slope characteristic parameters and the falling segment slope characteristic parameters. S5: Compare the blood glucose peak time parameter with the standard peak time parameter, the descending slope characteristic parameter with the preset recovery threshold parameter, and based on the comparison results, the ascending slope characteristic parameter and the dose interval mapping relationship, classify the prediabetic subjects and construct efficacy grading evaluation indicators.

[0006] As a further aspect of the present invention, the subject stratified coding includes a steady-state grading label, a blood glucose fluctuation penalty term, and an intervention matching index; the blood glucose change difference sequence includes single-dose attenuation magnitude, cumulative net response value, and baseline drift constant; the dose interval mapping relationship includes an effective onset threshold, an effect saturation boundary, and a preferred treatment window; the rising slope characteristic parameters include glucose absorption burst rate, early impedance calculus, and pre-peak secretory kinetic energy; the falling slope characteristic parameters include late fallback lag rate, islet compensatory half-life, and glucose clearance constant; and the efficacy grading evaluation indicators include physiological reversal grading, relapse risk weighting, and long-term steady-state score.

[0007] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Collect fasting blood glucose measurement values ​​of prediabetic subjects, sort the fasting blood glucose measurement values ​​by time series, establish a blood glucose time detection sequence, perform difference calculation based on adjacent nodes of the blood glucose time detection sequence, extract the single detection difference sequence, call the single detection difference sequence to perform extreme value screening and comparison, select the upper bound span extreme value parameter, and obtain the maximum fasting blood glucose difference. S102: Collect the blood glucose stability interval parameters from the standard configuration library, compare the difference limit between the maximum fasting blood glucose difference and the blood glucose stability interval parameters, determine the blood glucose fluctuation trend, and if the blood glucose fluctuation trend meets the baseline stability condition, call the fasting blood glucose measurement value, perform span value accumulation on the fasting blood glucose measurement value, and perform equal division conversion in combination with the total number of recorded nodes to obtain the average fasting blood glucose value. S103: Obtain preset threshold boundary parameters, perform segmented interval value mapping comparison based on the average fasting blood glucose value and the preset threshold boundary parameters, determine the risk tier, associate the corresponding preset stratification code with the risk tier and extract the basic identity identifier of the subject, perform segment character sequence concatenation of the stratification code and identity identifier, combine the risk feature file, and establish subject stratification coding.

[0008] As a further aspect of the present invention, the process of setting the threshold boundary parameter specifically involves: collecting fasting blood glucose baseline recording sequences from individuals diagnosed with prediabetes in a clinical historical database; performing kernel density estimation and cumulative probability distribution calculation on the fasting blood glucose baseline recording sequences; extracting morphological inflection points and discrete feature nodes corresponding to preset percentiles in the probability density distribution curve; extracting the theoretical boundary interval between impaired glucose tolerance and impaired fasting blood glucose in standard clinical disease diagnosis guidelines; performing an intersection mapping comparison between the extracted discrete feature nodes and the theoretical boundary interval; and extracting node data that are within the theoretical boundary interval and close to the trough of the probability density distribution curve as the critical judgment criterion for risk stratification to obtain the threshold boundary parameter.

[0009] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Perform clinical record retrieval for the subject stratified coding, extract the list of traditional Chinese medicine compound preparations, divide the list of traditional Chinese medicine compound preparations into concentration increments, determine the dosage parameters, perform sequence combination of the dosage parameters according to the time sequence arrangement rules, perform vectorized coding in combination with the interval parameter, and establish the dosage gradient of ginseng and astragalus compound preparations. S202: Based on the dosage gradient of the ginseng and astragalus compound, collect the pre-intervention period physiological records of prediabetic subjects in the pre-set sample data, extract the baseline blood glucose value before intervention, obtain the fasting blood glucose value after intervention from the physiological records of prediabetic subjects after the intervention period of the traditional Chinese medicine compound, perform cleaning on the fasting blood glucose value after intervention and the baseline blood glucose value before intervention, remove abnormal isolated points, perform pair mapping on the remaining values, and obtain the blood glucose monitoring pairs of the subjects. S203: Call the subject's blood glucose monitoring pair, perform parsing and extract periodic measurement parameters, perform node subtraction operation on the periodic measurement parameters, calculate the node concentration fluctuation deviation, perform timestamp alignment and merging based on the node concentration fluctuation deviation and the dose gradient of the ginseng and astragalus compound, establish dose-effect key-value pairs, perform ascending sorting and reconstruction of the dose-effect key-value pairs according to the dose increment direction, and generate a blood glucose change difference sequence.

[0010] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Call the dose gradient of the ginseng and astragalus compound and the difference sequence of blood glucose changes, set the dose gradient of the ginseng and astragalus compound as the independent variable benchmark, set the difference sequence of blood glucose changes as the dependent variable target combination to construct a two-dimensional coordinate discrete point matrix, perform piecewise linear regression operation on the two-dimensional coordinate discrete point matrix, calculate the slope value corresponding to the line connecting the discrete nodes, extract the intercept parameter and regression coefficient corresponding to the slope change stage, and obtain the regression slope feature parameter set; S302: Based on the regression slope feature parameter set, perform critical state detection, extract the absolute value of the difference between adjacent regression coefficients in the regression slope feature parameter set, perform global extreme value retrieval on the absolute value difference sequence, obtain the node index of the sequence where the maximum difference is located, call the node index to perform coordinate mapping addressing to the dosage gradient of the ginseng and astragalus compound, extract the scale value at the corresponding position, and obtain the inflection point of the dose response feature. S303: Based on the inflection point of the dose response feature, perform threshold segmentation on the difference sequence between the dose gradient of the ginseng and astragalus compound and the change in blood glucose, extract the data subset of independent intervals on both sides of the inflection point, perform position recombination on the associated parameters within the data subset according to the ascending order rule, perform fusion splicing on adjacent and overlapping intervals after recombination, construct the associated key-value pairs between the spliced ​​dose values ​​and the magnitude of blood glucose change, and establish a dose interval mapping relationship.

[0011] As a further aspect of the present invention, the process of extracting the intercept parameter and regression coefficient corresponding to the slope variation stage specifically involves: extracting the ratio of the difference between the ordinate and abscissa of each adjacent node in the two-dimensional coordinate discrete point matrix to calculate the local slope sequence; performing a segmentation operation on the local slope sequence according to the preset slope tolerance parameter to obtain monotonically changing sub-intervals; calling the coordinate data of the boundary nodes of the monotonically changing sub-intervals to calculate the intercept parameter and the regression coefficient corresponding to the monotonically changing sub-intervals respectively; arranging the intercept parameter and the regression coefficient corresponding to the monotonically changing sub-intervals in ascending order of value to obtain the regression slope feature parameter set; the specific setting method of the preset slope tolerance parameter is to extract the maximum variation range value of the historical test local slope sequence, and set the product of the maximum variation range value and the inherent error weight as the preset slope tolerance parameter; the specific setting method of the inherent error weight is to call the allowable error limit percentage value in the factory parameters of the test equipment and set it as the inherent error weight.

[0012] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Obtain the glucose tolerance test values ​​corresponding to multiple measurement nodes of prediabetic subjects in the authorized public historical test dataset, perform time node concentration correlation comparison and value traversal for the glucose tolerance test values, extract the global maximum concentration extreme value node in the concentration sequence, call the global maximum concentration extreme value node mapping timestamp identifier, and obtain the blood glucose peak time parameter. S402: Based on the blood glucose peak time parameter, the glucose tolerance test values ​​are divided into time axes, and the concentration values ​​corresponding to the preceding interval of the blood glucose peak time parameter are extracted and combined to establish a pre-sequence. The concentration values ​​corresponding to the following interval of the blood glucose peak time parameter are extracted and combined to establish a post-sequence, thus obtaining the rising segment time data sequence and the falling segment time data sequence. S403: Perform a quotient operation by dividing the first and last concentration difference in the rising segment time data sequence by the time span to obtain the positive climbing rate parameter; perform a quotient operation by dividing the first and last concentration deviation in the falling segment time data sequence by the time span to obtain the negative falling rate parameter; establish a feature pair between the positive climbing rate parameter and the negative falling rate parameter to generate the rising segment slope feature parameter and the falling segment slope feature parameter.

[0013] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Obtain the standard peak time parameter within the authorized clinical medical knowledge graph node, call the blood glucose peak time parameter and the standard peak time parameter to perform time axis time domain difference calculation, calibrate the time delay span index, perform linear normalization mapping between the time delay span index and the preset physiological extreme value, and generate the peak time deviation vector. S502: Call the preset recovery threshold parameter, perform bidirectional numerical boundary comparison and detection between the descending slope feature parameter and the preset recovery threshold parameter, identify the attenuation fallback compliance status, perform discrete category variable encoding and assignment based on the attenuation fallback compliance status, extract the scalar value of the blockage risk, and establish the glucose clearance blockage coefficient. S503: Perform multi-dimensional feature splicing and combination based on the peak time deviation vector and glucose clearance blockade coefficient to reconstruct the individual metabolic baseline tensor. Call the rising slope feature parameter and the individual metabolic baseline tensor to perform matrix-dot product operation to calibrate the basal absorption and metabolism score parameter. Perform addressing mapping on the basal absorption and metabolism score parameter within the dose interval mapping relationship to match the drug reversal estimate. Call the drug reversal estimate to perform unsupervised spatial clustering to define the intervention response echelon of the subjects and obtain efficacy grading evaluation indicators.

[0014] As a further aspect of the present invention, the method for obtaining the standard peak time parameter is as follows: obtaining continuous blood glucose monitoring data of healthy population samples with normal glucose tolerance characteristics from a preset sample dataset; extracting the time span records from the glucose load start node to the blood glucose concentration reaching the global maximum extreme value node from the continuous blood glucose monitoring data; performing central trend statistical calculations on the entire time span record set to obtain the time mean or median parameter; and setting the time node parameter characterizing the normal physiological metabolic absorption rhythm as the standard peak time parameter. The recovery threshold parameter is obtained by collecting blood glucose test records of the normal metabolic control group in the authorized clinical database during the decline phase. A baseline decline slope set is constructed by performing a quotient operation between the concentration decline difference and the corresponding time span for the blood glucose test records during the decline phase. Statistical distribution analysis is performed on the baseline decline slope set to extract the lower limit critical slope value of the preset probability confidence interval. The rate index that characterizes the boundary of the normal body's glucose clearance compensation ability is set as the recovery threshold parameter.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by performing difference calculations on multiple fasting blood glucose measurement values ​​and comparing them with stable parameters to extract fluctuation characteristics, a hierarchical coding mechanism based on the joint judgment of mean and boundary is constructed to avoid the risk of hierarchical mismatch in a single test. The difference before and after intervention in the incremental dosing cycle is extracted to construct a response sequence. A piecewise linear regression function is introduced to calculate the slope to capture the inflection point of the dose response characteristics. A quantitative mapping relationship between the dose interval and the intensity of blood glucose regulation is established. The peak time of the glucose tolerance test sequence is located to divide the time axis. The slope parameters of the rising and falling segments are calculated and compared with the standard peak time to judge the degree of recovery of the body. The evolution of the curve shape is transformed into a multi-dimensional efficacy grading index, which effectively improves the precision of clinical evaluation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0019] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0020] Please see Figure 1 This invention provides a stratified randomized controlled method for evaluating the intervention of traditional Chinese medicine compound prescriptions in prediabetes, including the following steps: S1: Collect fasting blood glucose measurement values ​​of prediabetic subjects, perform difference calculation on the fasting blood glucose measurement values ​​to obtain the maximum fasting blood glucose difference, compare the maximum fasting blood glucose difference with the blood glucose stability interval parameter, calculate the mean of the fasting blood glucose measurement values ​​based on the comparison results to obtain the average fasting blood glucose value, compare the average fasting blood glucose value with the preset threshold boundary parameter, and construct subject stratified coding based on the comparison results of the preset threshold boundary parameter. S2: Based on the subject stratified coding, the dosage gradient of ginseng and astragalus compound was extracted for the traditional Chinese medicine compound intervention of prediabetic subjects. Fasting blood glucose values ​​after the intervention cycle and baseline blood glucose values ​​before the intervention cycle were collected. The difference between the fasting blood glucose value after the intervention and the baseline blood glucose value before the intervention was calculated. Based on the calculated difference and the dosage gradient of ginseng and astragalus compound, a blood glucose change difference sequence was constructed. S3: Input the dose gradient of the ginseng and astragalus compound and the difference sequence of blood glucose changes into a piecewise linear regression function to calculate the slope, obtain the inflection point of the dose response feature, and perform interval sorting and merging calculation on the dose gradient of the ginseng and astragalus compound and the difference sequence of blood glucose changes based on the inflection point of the dose response feature to construct the dose interval mapping relationship. S4: Collect glucose tolerance test values ​​from prediabetic subjects, perform numerical optimization calculations on the glucose tolerance test values ​​to obtain the peak blood glucose time parameter, perform time axis segmentation calculations on the glucose tolerance test values ​​based on the peak blood glucose time parameter, obtain the rising segment time data sequence and the falling segment time data sequence, perform slope calculations on the rising segment time data sequence and the falling segment time data sequence respectively, and obtain the rising segment slope feature parameter and the falling segment slope feature parameter; S5: Compare the peak blood glucose time parameter with the standard peak time parameter, compare the descending slope characteristic parameter with the preset recovery threshold parameter, and classify and judge the prediabetic subjects based on the comparison results of the peak blood glucose time parameter, the comparison results of the descending slope characteristic parameter, the ascending slope characteristic parameter, and the dose interval mapping relationship, and construct efficacy grading evaluation indicators; Subject stratification coding includes steady-state grading labels, blood glucose fluctuation penalty terms, and intervention matching indexes. Blood glucose change difference sequences include single-dose attenuation magnitude, cumulative net response value, and baseline drift constant. Dose interval mapping relationships include effective onset threshold, effect saturation boundary, and optimal treatment window. Rising slope characteristic parameters include glucose uptake burst rate, early impedance calculus, and pre-peak secretory kinetics. Falling slope characteristic parameters include late fallback lag rate, islet compensatory half-life, and glucose clearance constant. Efficacy grading evaluation indicators include physiological reversal grading, relapse risk weights, and long-term steady-state scores.

[0021] Please see Figure 2 The specific steps of S1 are as follows: S101: Collect fasting blood glucose measurement values ​​of prediabetic subjects, sort the fasting blood glucose measurement values ​​by time series, establish a blood glucose time detection sequence, perform difference calculation based on adjacent nodes of the blood glucose time detection sequence, extract the single detection difference sequence, call the single detection difference sequence to perform extreme value screening and comparison, select the upper bound span extreme value parameter, and obtain the maximum fasting blood glucose difference. The system uses a continuous glucose monitoring (CGM) data interface to extract fasting blood glucose measurement values ​​from prediabetic subjects. The collected data is in floating-point format with absolute timestamps. Data cleaning and preprocessing are performed on the extracted fasting blood glucose measurement values. All record nodes are traversed, and records with missing values ​​are identified and removed. A median filtering algorithm with a window size of 5 data points is used to process the remaining data sequence. The measurement values ​​of the current node and the two nodes before and after it are extracted, sorted in ascending order, and the median is used to overwrite the original value of the current node, removing high-frequency spike noise interference. Time series sorting is performed on the denoised and cleaned fasting blood glucose measurement values. The absolute timestamp attribute of each node is read, and a quicksort mechanism is used to reorganize the node memory address mapping positions according to the ascending order of timestamps, establishing a blood glucose time detection sequence. Based on the blood glucose time detection sequence, the measurement data of adjacent nodes are extracted and subtracted. The calculation formula is: Single detection difference = Fasting blood glucose measurement value of the subsequent node - Fasting blood glucose measurement value of the preceding node. All difference results generated by the traversal operation are stored sequentially to establish a single detection difference sequence. The system calls upon the single-test difference sequence to perform extreme value screening and comparison. The initial maximum difference record variable is set to 0. The absolute value of each difference in the sequence is extracted and compared with the current maximum difference record variable. If the extracted absolute value is greater than the current maximum difference record variable, the extracted value is updated in the maximum difference record variable, until the entire sequence is traversed. In actual operation, the fasting blood glucose measurement values ​​obtained at five consecutive time points were 5.2 mmol / L, 5.6 mmol / L, 5.3 mmol / L, 6.1 mmol / L, and 5.8 mmol / L, respectively. Substituting the calculation logic described above into the formula, the first single-detection difference is 5.6 - 5.2 = 0.4 mmol / L, the second single-detection difference is 5.3 - 5.6 = -0.3 mmol / L, the third single-detection difference is 6.1 - 5.3 = 0.8 mmol / L, and the fourth single-detection difference is 5.8 - 6.1 = -0.3 mmol / L. The single-detection difference sequence is extracted as 0.4 mmol / L, -0.3 mmol / L, 0.8 mmol / L, and -0.3 mmol / L. Extreme value screening and comparison are performed, extracting the absolute values ​​0.4, 0.3, 0.8, and 0.3 respectively and comparing them with the initial value 0 multiple times. The upper bound extreme value parameter corresponding to the value 0.8 is selected, obtaining the maximum fasting blood glucose difference as 0.8 mmol / L.

[0022] S102: Collect the blood glucose stability interval parameters from the standard configuration library, compare the difference limit between the maximum fasting blood glucose difference and the blood glucose stability interval parameters to determine the blood glucose fluctuation trend, and if the blood glucose fluctuation trend meets the baseline stability condition, call the fasting blood glucose measurement value, perform the span value accumulation on the fasting blood glucose measurement value, and perform the average conversion based on the total number of recorded nodes to obtain the average fasting blood glucose value. A query request is sent to the standard configuration database server to collect pre-stored blood glucose stability interval parameters, including the lower and upper limits of the stability interval. Using the maximum fasting blood glucose difference obtained in the previous steps as the input variable, the blood glucose stability interval parameters are called to perform a difference boundary comparison judgment. The calculation formula is: Difference Deviation = Maximum Fasting Blood Glucose Difference - Upper Limit of Stable Interval. If the difference deviation is less than or equal to 0, the blood glucose fluctuation trend is determined to meet the baseline stability condition; if the difference deviation is greater than 0, the blood glucose fluctuation trend is determined to be an abnormal deviation state. For cases where the blood glucose fluctuation trend meets the baseline stability condition, the cleaned fasting blood glucose measurement value set is requested from memory. All fasting blood glucose measurement values ​​in the set are extracted and summed across the range. The initial sum variable is set to 0. Each node value is extracted, added to the sum variable, and the sum variable is overwritten until the entire range is traversed and the total sum is obtained. The total number of recorded nodes is extracted, and the sum of the spans is divided by the total number of recorded nodes. The formula is: Average fasting blood glucose value = Sum of spans / Total number of recorded nodes, thus completing the average conversion. In the actual data simulation, the upper limit of the stable interval was collected from the standard configuration database as 1.0 mmol / L. The calculated maximum difference in fasting blood glucose of 0.8 mmol / L is substituted into the calculation formula, and the difference deviation = 0.8 - 1.0 = -0.2 mmol / L. It is determined that -0.2 mmol / L is less than 0, triggering the subsequent branch path that meets the baseline stability condition. The fasting blood glucose measurement values ​​of the aforementioned 5 nodes are accumulated, and the sum of the spans = 5.2 + 5.6 + 5.3 + 6.1 + 5.8 = 28.0 mmol / L. The total number of recorded nodes is 5. Substituting them into the average conversion formula, the average fasting blood glucose value is 28.0 / 5 = 5.6 mmol / L. Therefore, the average fasting blood glucose value is 5.6 mmol / L.

[0023] S103: Obtain the preset threshold boundary parameters, perform segmented interval numerical mapping comparison based on the average fasting blood glucose value and the preset threshold boundary parameters, determine the risk echelon, associate the corresponding preset stratification code with the risk echelon and extract the basic identity identifier of the subject, perform segment character sequence concatenation of the stratification code and identity identifier, combine the risk feature file, and establish subject stratification coding. The specific process for setting the threshold boundary parameters is as follows: collecting fasting blood glucose baseline recording sequences from individuals diagnosed with prediabetes in the clinical historical database; performing kernel density estimation and cumulative probability distribution calculation on the fasting blood glucose baseline recording sequences; extracting morphological inflection points and discrete feature nodes corresponding to preset percentiles in the probability density distribution curve; extracting the theoretical boundary interval between impaired glucose tolerance and impaired fasting blood glucose in the standard clinical disease diagnosis guidelines; performing intersection mapping comparison between the extracted discrete feature nodes and the theoretical boundary interval; extracting node data that are inside the theoretical boundary interval and close to the trough of the probability density distribution curve as the critical judgment criterion for risk stratification; and obtaining the threshold boundary parameters. A communication connection was established to collect baseline fasting blood glucose records from individuals diagnosed with prediabetes within the past 5 years from a clinical historical database, removing abnormal records containing non-numeric characters. For the baseline fasting blood glucose records, a Gaussian kernel function was used to perform kernel density estimation, with a bandwidth parameter set to 0.1, to construct a probability density distribution curve. The first derivative sequence was calculated by traversing the probability density distribution curve, and the nodes where the first derivative transitioned from negative to positive values ​​were extracted as morphological inflection points. Simultaneously, a cumulative probability distribution calculation was performed on the baseline fasting blood glucose records, and the nodes corresponding to the 85th percentile were extracted as discrete feature nodes corresponding to preset percentiles. The theoretical limit intervals were extracted from the standard clinical disease diagnosis guideline text data, setting the lower limit of impaired glucose tolerance at 7.8 mmol / L and the upper limit of impaired fasting blood glucose at 7.0 mmol / L. The extracted discrete feature nodes are compared with the theoretical limit intervals using an intersection mapping to determine if the discrete feature node values ​​fall within the range of 7.0 to 7.8. Simultaneously, the absolute difference between the discrete feature nodes and the morphological inflection points is calculated. Node data within the theoretical limit intervals with an absolute difference less than 0.2 are selected as the critical criterion for risk stratification, obtaining the threshold boundary parameters. The average fasting blood glucose value obtained in the previous steps is compared with the preset threshold boundary parameters using a segmented interval numerical mapping. If the average fasting blood glucose value is less than the threshold boundary parameters, the node is assigned to the low-risk tier; otherwise, it is assigned to the high-risk tier. The corresponding preset stratification codes are retrieved: 1 for the low-risk tier and 2 for the high-risk tier. The basic identity identifier string of the subject is extracted from the identity registration form. The stratification code and the basic identity identifier of the subject are concatenated using a bit-segment character sequence concatenation operation, calculated as: Subject Stratification Code = Basic Identity Identifier String + Stratification Code. This combination forms a risk feature profile and is stored in the database. In the actual calculation, the morphological inflection point was found to be 6.8 mmol / L, and the discrete feature node was 7.1 mmol / L. 7.1 mmol / L falls within the range of 7.0 to 7.8 mmol / L, and the absolute difference is 7.1 - 6.8 = 0.3. The discrete feature node was readjusted to another trough position of 6.9 mmol / L, with an absolute difference of 6.9 - 6.8 = 0.1, satisfying the condition, and the threshold boundary parameter was obtained as 6.9 mmol / L. The aforementioned average fasting blood glucose value of 5.6 mmol / L was compared with 6.9 mmol / L; 5.6 is less than 6.9, thus determining the risk stratification group as low-risk, and the stratification code 1 was extracted. The basic identity identifier of the subject was obtained as 8859. Substituting into the concatenation formula, the subject stratification code = 8859 + 1 = 88591. The subject stratification code was then established.

[0024] Table 1 Threshold Boundary Parameter Mapping Table Risk Tier Classification Preset layer code Identity Identification Example Threshold boundary parameter lower limit Threshold boundary parameter upper limit Extremely low risk echelon 0 1001 3.9 mmol / L 5.5 mmol / L Low-risk echelon 1 8859 5.6 mmol / L 6.9 mmol / L High-risk echelons 2 3421 7.0 mmol / L 7.8 mmol / L Please see Figure 3 The specific steps of S2 are as follows: S201: Perform clinical record retrieval for subject stratified coding, extract the list of traditional Chinese medicine compound preparations, divide the list of traditional Chinese medicine compound preparations into concentration increments, determine the dosage parameters, perform sequence combination of the dosage parameters according to the time sequence arrangement rules, perform vectorized coding in combination with the interval parameter, and establish the dosage gradient of ginseng and astragalus compound; The system reads the identity identifier segment from the subject's stratified coding and sends a file retrieval request to the medication prescription data table in the clinical medical information system. It then parses the returned Extended Markup Language (EML) format file and extracts the traditional Chinese medicine (TCM) compound preparation list from the prescription detail tags. For the total amount of dry extract powder of the core ingredients Astragalus membranaceus and Salvia miltiorrhiza in the TCM compound preparation list, a concentration-incrementing division operation is performed. The initial baseline starting dose is set to 5.0 grams, the increment step is set to 2.5 grams, and the total number of division levels is set to 4 levels. The medication dosage parameter sequence is determined according to the arithmetic progression rule, and the calculation formula is: Nth level medication dosage parameter = initial baseline starting dose plus the number of preceding levels multiplied by the increment step. The generated medication dosage parameters of the 4 levels are sequentially appended to a one-dimensional array structure according to the chronological order, and sequence combination is performed. The dosing interval parameter is set to 24 hours, and the interval parameter is merged into the one-dimensional array structure as an extended column. Feature expansion is performed on the data of each level, and the one-hot encoding algorithm is used to perform vectorized encoding on the level sequence number to establish the Astragalus membranaceus and Salvia miltiorrhiza compound dosage gradient matrix. In the simulation, the initial dose is 5.0 g, and the increment is 2.5 g. Substituting into the calculation formula, the dosage parameter for level 1 is 5.0 + 0 x 2.5 = 5.0 g, for level 2 it is 5.0 + 1 x 2.5 = 7.5 g, for level 3 it is 5.0 + 2 x 2.5 = 10.0 g, and for level 4 it is 5.0 + 3 x 2.5 = 12.5 g. The sequence combination results are 5.0, 7.5, 10.0, and 12.5. Combining the 24-hour dosing interval parameter, after vectorization encoding, the first row of the feature vector in the Astragalus and Ginseng compound dosage gradient is established as: dose 5.0 g, interval 24 hours, encoding bits 1, 0, 0, 0.

[0025] S202: Based on the dosage gradient of the ginseng and astragalus compound, collect the pre-intervention period physiological records of prediabetic subjects in the pre-set sample data, extract the baseline blood glucose value before intervention, obtain the physiological records of prediabetic subjects after the intervention period of the traditional Chinese medicine compound, extract the fasting blood glucose value after intervention, perform cleaning on the fasting blood glucose value after intervention and the baseline blood glucose value before intervention, remove abnormal isolated points, perform pair mapping on the remaining values, and obtain the blood glucose monitoring pairs of the subjects. The parameters in the dosage gradient matrix of the Astragalus and Ginseng compound formula are used as index keys to connect to the database master node of the pre-set sample dataset. A conditional query is initiated to extract the physiological records of prediabetic subjects over a period of seven natural days prior to the aforementioned traditional Chinese medicine compound intervention. The marker position for the morning period in the records is located, and the baseline blood glucose value sequence before intervention is extracted. Similarly, the corresponding physiological records of prediabetic subjects after the completion of four levels of traditional Chinese medicine compound intervention cycles are obtained, and the post-intervention fasting blood glucose value sequence with post-intervention marker attributes is extracted. Data cleaning operations are performed simultaneously on the post-intervention fasting blood glucose value sequence and the pre-intervention baseline blood glucose value sequence, and box-line analysis is called. Figure 4 The quartile range algorithm detects outliers by calculating the upper and lower quartiles. Data nodes exceeding the upper quartile plus 1.5 times the quartile range or falling below the lower quartile minus 1.5 times the quartile range are marked as outliers and erased from memory. The remaining pre-intervention baseline blood glucose values ​​and post-intervention fasting blood glucose values ​​are paired and mapped based on the subject's individual identifier and the relative offset of measurement time. Inner joins are used to merge two sets of values ​​under the same primary key into the same row, resulting in a pair of subject blood glucose monitoring values. For example, if the extracted pre-intervention baseline blood glucose value set contains 5.6 mmol / L, 12.8 mmol / L, and 5.8 mmol / L, the lower quartile is calculated to be 5.6, the upper quartile to be 12.8, and the quartile range = 12.8 - 5.6 = 7.2. The upper bound was set at 12.8 + 1.5 × 7.2 = 23.6. In this dataset, 12.8 was identified as an isolated, abnormally high value that did not exceed the upper bound. Values ​​of 24.0 were removed. After cleaning, the remaining pre-intervention values ​​were 5.6 and 5.8. Post-intervention fasting blood glucose values ​​were extracted as 5.1 and 5.4. Paired mapping was performed to generate the first pair of paired data: pre-intervention 5.6, post-intervention 5.1. The second pair of paired data was generated: pre-intervention 5.8, post-intervention 5.4. These were then combined to obtain the subject blood glucose monitoring pairs.

[0026] S203: Call the subject's blood glucose monitoring group, perform parsing and extract periodic measurement parameters, perform node subtraction operation on the periodic measurement parameters, calculate the node concentration fluctuation deviation, perform timestamp alignment and merging based on the node concentration fluctuation deviation and the dose gradient of the ginseng and astragalus compound, establish dose-effect key-value pairs, perform ascending sorting and reconstruction of the dose-effect key-value pairs according to the dose increasing direction, and generate a blood glucose change difference sequence. The system reads the memory address of the subject's blood glucose monitoring group, iterates through each record, and performs parsing to extract the periodic measurement parameters, which include the pre-intervention baseline sub-parameter and the post-intervention outcome sub-parameter. For the periodic measurement parameters, the CPU's arithmetic logic unit performs node subtraction, calculating the following formula: Node concentration fluctuation deviation = Post-intervention outcome sub-parameter - Pre-intervention baseline sub-parameter. The system extracts the Astragalus and Ginseng compound dosage gradient matrix generated in the previous steps, extracts the absolute timestamp field, and performs timestamp alignment and merging operations with the Astragalus and Ginseng compound dosage gradient, matching records with the same date offset. The successfully matched Astragalus and Ginseng compound dosage parameters are used as keys, and the node concentration fluctuation deviations are used as values ​​to establish dose-effect key-value pairs in a hash dictionary format. A dictionary sorting function is called, and based on the dose value contained in the key, the dose-effect key-value pairs are reconstructed in ascending order along the dose increment direction. The key attributes are stripped, retaining only the reconstructed value attribute sequence, generating a blood glucose change difference sequence. The extrapolation is as follows: Using the first set of paired data mentioned above, the baseline subparameter before intervention is 5.6, and the outcome subparameter after intervention is 5.1. Substituting into the calculation formula, the concentration fluctuation deviation at the first node = 5.1 - 5.6 = -0.5 mmol / L. Using the second set of paired data, the baseline was 5.8 before intervention, and the outcome was 5.4 after intervention. The concentration fluctuation deviation at the second node = 5.4 - 5.8 = -0.4 mmol / L. After timestamp alignment and merging, the dose corresponding to the first deviation is 5.0 g, and the key-value pair is established as: 5.0 g mapped to -0.5 mmol / L. The dose corresponding to the second deviation is 7.5 g, and the key-value pair is established as: 7.5 g mapped to -0.4 mmol / L. Reconstruct the glucose levels by ascending order of 5.0 g and 7.5 g, extracting the value attributes to generate a glucose change difference sequence of -0.5 mmol / L and -0.4 mmol / L.

[0027] Table 2 Dose-Effect Key Pairs Serial Number Align timestamps Ginseng and Astragalus Compound Dosage Parameters Node concentration fluctuation deviation Reconstruct the sorting sequence number 1 Morning of Day 1 5.0 grams -0.5 mmol / L 1 2 Morning of the 2nd day 7.5 grams -0.4 mmol / L 2 3 Morning of the 3rd 10.0 grams -0.8 mmol / L 3 4 Morning of the 4th day 12.5 grams -0.6 mmol / L 4 Please see Figure 4 The specific steps of S3 are as follows: S301: Call the dose gradient of Ginseng and Astragalus Compound and the difference sequence of blood glucose change, set the dose gradient of Ginseng and Astragalus Compound as the independent variable benchmark, set the difference sequence of blood glucose change as the dependent variable target combination to construct a two-dimensional coordinate discrete point matrix, perform piecewise linear regression operation on the two-dimensional coordinate discrete point matrix, calculate the slope value corresponding to the line connecting discrete nodes, extract the intercept parameter and regression coefficient corresponding to the slope change stage, and obtain the regression slope feature parameter set; The process of extracting the intercept parameter and regression coefficient corresponding to the slope variation stage is as follows: First, the local slope sequence is calculated by extracting the ratio of the difference between the ordinate and abscissa of each adjacent node in the two-dimensional coordinate discrete point matrix. Then, the local slope sequence is segmented according to the preset slope tolerance parameter to obtain monotonically changing sub-intervals. Next, the coordinate data of the boundary nodes of the monotonically changing sub-intervals are called to calculate the intercept parameter and regression coefficient corresponding to each sub-interval. Finally, the intercept parameter and regression coefficient corresponding to each sub-interval are arranged in ascending order to obtain the regression slope feature parameter set. The preset slope tolerance parameter is set by extracting the maximum variation range of the historical test local slope sequence and setting the product of the maximum variation range and the inherent error weight as the preset slope tolerance parameter. The inherent error weight is set by calling the percentage value of the allowable error limit in the factory parameters of the test equipment and setting it as the inherent error weight. A virtual planar coordinate system is established by calling the one-dimensional array sequence of the dosage gradient of the Astragalus and Ginseng compound formula and the sequence of blood glucose change differences. The dosage gradient data of the Astragalus and Ginseng compound formula is traversed and filled onto the horizontal axis as the independent variable baseline, and the blood glucose change difference sequence is sequentially mapped onto the vertical axis as the dependent variable target. The corresponding element pairs of the horizontal and vertical axes are combined to construct a two-dimensional coordinate discrete point matrix in two-dimensional space. Piecewise linear regression is performed on the two-dimensional coordinate discrete point matrix. The percentage value of the tolerance limit built into the factory configuration file is read and set as the inherent error weight variable. Historical test local slope sequence data in the previous test period is extracted, and the maximum and minimum values ​​in the sequence are subtracted to obtain the maximum variation range value. The calculation formula is: preset slope tolerance parameter = maximum variation range value multiplied by inherent error weight. Traversing the two-dimensional coordinate discrete point matrix, the difference between the vertical coordinates of each adjacent node is extracted as the numerator, and the difference between the horizontal coordinates is extracted as the denominator. Division is then performed to obtain the local slope sequence. The absolute value of the difference between adjacent slope values ​​in the local slope sequence is calculated. When this absolute value exceeds a preset slope tolerance parameter, a segmentation operation is performed at that node, dividing the overall coordinate matrix into multiple monotonically changing sub-intervals. The coordinate data of the boundary nodes of each monotonically changing sub-interval are retrieved, and the intercept parameter and regression coefficient corresponding to the monotonically changing sub-interval are calculated using the standard least squares formula. Finally, all the obtained intercept parameters and regression coefficients are arranged in ascending order of the initial x-coordinate values ​​of the intervals to obtain the regression slope characteristic parameter set. In the example calculation, the allowable error limit percentage is set to 5%, with an inherent error weight of 0.05. The historical maximum variation range is 2.0. The preset slope tolerance parameter is calculated as 2.0 multiplied by 0.05 = 0.1. Extract the coordinates of the two-dimensional discrete point matrix: node A is 5.0, -0.5; node B is 7.5, -0.4; and node C is 10.0, -0.8. Calculate the local slope of line segment AB: (-0.4 - -0.5) / (7.5 - 5.0) = 0.04. Calculate the local slope of line segment BC: (-0.8 - -0.4) / (10.0 - 7.5) = -0.16. Calculate the absolute difference between adjacent slopes: absolute value (-0.16 - 0.04) = 0.2. Since 0.2 is greater than 0.1, perform a segmentation operation at node B, dividing the first monotonically changing sub-interval from node A to node B. Calculate the intercept parameter of this interval as -0.7 and the regression coefficient as 0.04. Store the regression slope feature parameter set sequentially.

[0028] The regression slope feature parameter set is read into the memory processing queue, and a critical state detection mechanism is executed based on the regression slope feature parameter set. The regression coefficient fields associated with each monotonically changing sub-interval within the parameter set are traversed. The regression coefficients of two adjacent monotonically changing sub-intervals are extracted, subtracted, and their absolute values ​​are taken to generate a sequence of absolute difference values. The global extreme value retrieval function is called on the generated sequence of absolute difference values. The initial extreme value pointer variable is set to 0, and the value of the current sequence element is compared with the value recorded in the extreme value pointer variable through iterative looping. The maximum value is updated, and its position sequence number in the sequence is recorded simultaneously to obtain the index of the sequence node containing the maximum difference. This sequence node index is extracted, and a coordinate mapping addressing operation is performed on the one-dimensional array memory address where the dose gradient of the Astragalus compound is located. The sequence node index is incremented by 1 as an offset to locate the specific array cell, and the scale value stored at the corresponding position is extracted. Finally, this scale value is output as the inflection point of the dose response feature. In the actual calculation scenario, the regression slope feature parameter set contains three monotonically changing sub-intervals with regression coefficients of 0.04, -0.16, and -0.02, respectively. The first step is to extract adjacent regression coefficients, subtract them, and take the absolute value. The first absolute difference is calculated as (-0.16 - 0.04) = 0.20, and the second is calculated as (-0.02 - -0.16) = 0.14. This generates a sequence of absolute differences of 0.20 and 0.14. A global extreme value search is performed; 0.20 is greater than 0.14, so the maximum difference is locked at 0.20. Its position in the sequence is recorded, and the node index of the sequence containing the maximum difference is obtained as 1. This node index 1 is used to address the dosage gradient array of the Astragalus and Ginseng Compound Formula, which contains 5.0g, 7.5g, 10.0g, and 12.5g. Locate the physical position corresponding to index 1 plus 1 equals 2, and extract the corresponding scale value as 7.5 grams. Thus, the dose response characteristic inflection point is obtained as 7.5 grams.

[0029] S302: Based on the regression slope feature parameter set, perform critical state detection, extract the absolute value of the difference between adjacent regression coefficients in the regression slope feature parameter set, perform global extreme value retrieval on the absolute value difference sequence, obtain the node index of the sequence where the maximum difference is located, call the node index to perform coordinate mapping addressing to the dose gradient of the ginseng and astragalus compound, extract the scale value at the corresponding position, and obtain the dose response feature inflection point; S303: Based on the inflection point of the dose response characteristics, threshold segmentation is performed on the difference sequence between the dose gradient of the ginseng and astragalus compound and the change in blood glucose. Data subsets of independent intervals on both sides of the inflection point are extracted. The associated parameters in the data subsets are reorganized according to the ascending order rule. The adjacent and overlapping intervals after reorganization are fused and spliced. The spliced ​​dose values ​​and the magnitude of blood glucose change are used to construct associated key-value pairs to establish a dose interval mapping relationship. The dose response feature inflection point value is received from the memory control bus. The dose gradient array of the Astragalus and Ginseng compound formula and the blood glucose change difference sequence are loaded respectively. A threshold segmentation operation is performed using the dose response feature inflection point value as the boundary. The Astragalus and Ginseng compound formula dose gradient array is traversed. Elements with values ​​less than or equal to the feature inflection point, along with their corresponding values ​​in the blood glucose change difference sequence, are allocated to the first memory block. Elements with values ​​greater than the feature inflection point and their corresponding values ​​are allocated to the second memory block, thereby extracting independent interval data subsets located on both sides of the inflection point. For each record within the independent interval data subset, the position is reorganized according to the dose value in the associated parameters using an insertion sort algorithm in ascending order. After reorganization, a fusion and splicing operation is performed on adjacent intervals between the two blocks with overlapping numerical boundaries to eliminate redundant boundary nodes. All complete dose values ​​after splicing are extracted as key fields, and the corresponding blood glucose change amplitude is extracted as value fields. A key-value mapping mechanism is called to construct associated key-value pairs, ultimately establishing a dose interval mapping relationship network across the entire dosage range. The specific calculation example is as follows: the inflection point of the dose response feature obtained above is 7.5 grams. The dose gradient array of the Astragalus and Ginseng compound is 5.0 grams, 7.5 grams, 10.0 grams, and 12.5 grams, and the corresponding blood glucose change difference sequence is -0.5 mmol / L, -0.4 mmol / L, -0.8 mmol / L, and -0.6 mmol / L. A threshold segmentation operation is performed: 5.0 grams and 7.5 grams are determined to be less than or equal to 7.5 grams, and are allocated to the first independent interval data subset containing data pairs: 5.0 and -0.5, 7.5 and -0.4. 10.0 grams and 12.5 grams are determined to be greater than 7.5 grams, and are allocated to the second independent interval data subset containing data pairs: 10.0 and -0.8, 12.5 and -0.6. The doses within both subsets already conform to the ascending order rule, and positional recombination is directly completed. No duplicate data was found at the boundary, so the fusion and splicing were performed directly in sequence. The spliced ​​dose values ​​and blood glucose change amplitudes were extracted to construct associated key-value pairs, and the final dose interval mapping relationship was established as follows: 5.0 maps to -0.5, 7.5 maps to -0.4, 10.0 maps to -0.8, and 12.5 maps to -0.6.

[0030] Please see Figure 5 The specific steps of S4 are as follows: S401: Obtain the glucose tolerance test values ​​corresponding to multiple measurement nodes of prediabetic subjects in the authorized public historical test dataset, perform time node concentration correlation comparison and value traversal for the glucose tolerance test values, extract the global maximum concentration extreme value node in the concentration sequence, call the global maximum concentration extreme value node mapping timestamp identifier, and obtain the blood glucose peak time parameter. The application programming interface (API) connects to the publicly available historical test dataset. By inputting the subject identifier and the standard glucose tolerance test item code, it obtains the glucose tolerance test (GTT) blood glucose measurement sequences at multiple specific time points after multiple prediabetic subjects ingest a standard glucose solution. For the GTT blood glucose measurement sequences, a bidirectional hash table of timestamps and concentration measurements is established to perform time-node concentration correlation comparison. The value traversal program is started, declaring a global maximum concentration extreme value node variable and initializing it to a negative infinity value, while simultaneously declaring a corresponding timestamp temporary variable. Concentration value items in the hash table are read one by one in the order of collection time. The currently read concentration value is compared with the global maximum concentration extreme value node variable. If the current concentration value is greater than the value in the global maximum concentration extreme value node variable, a memory overwrite operation is performed to update the extreme value variable, and the timestamp associated with the current concentration is simultaneously assigned to the timestamp temporary variable. The traversal process continues until the last node element of the hash table is read. At this point, the global maximum concentration extreme value node stored in memory is extracted, the reverse-mapping timestamp identifier of that node is called, and finally the extracted timestamp temporary variable content is output to obtain the blood glucose peak time parameter. Substituting actual monitoring data into the calculations, five specific time points were obtained for a subject during an oral glucose tolerance test. The concentration at time point 30 minutes was 8.4 mmol / L, at time point 60 minutes it was 11.2 mmol / L, at time point 90 minutes it was 9.8 mmol / L, and at time point 120 minutes it was 8.1 mmol / L. After establishing a hash association, the traversal was initiated. In the first comparison, 8.4 is greater than negative infinity, so the maximum value was updated to 8.4, and the value was temporarily stored for 30 minutes. In the second comparison, 11.2 is greater than 8.4, so the maximum value was updated to 11.2, and the value was temporarily stored for 60 minutes. In the third comparison, 9.8 is less than 11.2, so the value remained unchanged. In the fourth comparison, 8.1 is less than 11.2, so the value remained unchanged. The traversal ended, and the global maximum concentration extreme value node of 11.2 mmol / L was successfully extracted. The timestamp identifier of 60 minutes was extracted by mapping this extreme value node, thus obtaining the blood glucose peak time parameter as 60 minutes.

[0031] S402: Based on the blood glucose peak time parameter, the glucose tolerance test values ​​are divided into time axis segments. The concentration values ​​corresponding to the interval before the blood glucose peak time parameter are extracted and combined to form a pre-sequence. The concentration values ​​corresponding to the interval after the blood glucose peak time parameter are extracted and combined to form a post-sequence, thus obtaining the rising segment time data sequence and the falling segment time data sequence. The system receives the peak blood glucose time parameter from the receiving process and uses it as a benchmark to initiate a search and comparison operation on the time axis of the glucose tolerance test values. It then performs a time axis segmentation task, comparing the time node attributes of all measurement records with the peak blood glucose time parameter. Records with time node attribute values ​​strictly less than the peak blood glucose time parameter are selected, and their preceding records' time parameters and corresponding concentration values ​​are extracted. These are then combined and arranged in chronological order to create a preceding sequence. Similarly, records with time node attribute values ​​strictly greater than the peak blood glucose time parameter are selected, and their following records' time parameters and corresponding concentration values ​​are extracted and arranged in chronological order to create a following sequence. Finally, the completed preceding sequence is defined as equivalent to the rising segment of the time data sequence, and the completed following sequence is defined as equivalent to the falling segment of the time data sequence. The actual test dataset from the previous steps is further broken down to obtain four nodes for the glucose tolerance test values. The data pairs are: 30 minutes associated with 8.4 mmol / L, 60 minutes associated with 11.2 mmol / L, 90 minutes associated with 9.8 mmol / L, and 120 minutes associated with 8.1 mmol / L. The previously obtained peak blood glucose time parameter is input as 60 minutes. Each time node is compared with 60 minutes to segment it. For the 30-minute record, 30 is less than 60, and this node is determined as the preceding interval. For the 90-minute record, 90 is greater than 60, and this node is determined as the following interval. For the 120-minute record, 120 is greater than 60, and this node is also determined as the following interval. The time and concentration values ​​of the corresponding records in the preceding intervals are extracted and combined to establish a pre-sequence containing the element pairs: 30 minutes and 8.4 mmol / L. This yields the rising segment time data sequence. Extract the corresponding records of the subsequent intervals and perform combinations to create a subsequent sequence containing element pairs: 90 minutes and 9.8 mmol / L, 120 minutes and 8.1 mmol / L.

[0032] S403: Perform a quotient operation by dividing the difference between the first and last concentrations in the rising segment of the time data sequence by the time span to obtain the positive climbing rate parameter; perform a quotient operation by dividing the difference between the first and last concentrations in the falling segment of the time data sequence by the time span to obtain the negative falling rate parameter; establish a feature pair between the positive climbing rate parameter and the negative falling rate parameter to generate the rising segment slope feature parameter and the falling segment slope feature parameter; The rising and falling time data sequences, along with the global maximum concentration extreme node, are retrieved from the memory control center for subsequent arithmetic processing. For the rising time data sequence, the first concentration value and its associated first time node are extracted, and the formula is: Positive rise rate parameter = (Value corresponding to the global maximum concentration extreme node minus the first concentration value of the rising segment) divided by (Blood glucose peak time parameter minus the value of the first time node of the rising segment). The quotient operation is performed to obtain the positive rise rate parameter. Similarly, for the falling time data sequence, the last concentration value and its associated last time node are extracted, and the formula is: Negative fall rate parameter = (Last concentration value of the falling segment minus the value corresponding to the global maximum concentration extreme node) divided by (Last time node value of the falling segment minus the blood glucose peak time parameter). The quotient operation corresponding to the time span is performed to obtain the negative fall rate parameter. The obtained positive ascent rate parameter and negative descent rate parameter are stored in a tuple structure to establish feature pairs, thus generating the ascent and descent slope feature parameters. Substituting these parameters into specific numerical calculations, the aforementioned calculation node data is extracted. The ascent time data sequence only contains the first node data pair: time 30 minutes, concentration 8.4 mmol / L. The global maximum concentration extreme node data pair: time 60 minutes, concentration 11.2 mmol / L. Substituting into the calculation formula, the positive ascent rate parameter = (11.2 - 8.4) / (60 - 30) = 2.8 / 30 = 0.093 mmol / L / min. The extracted descent time data sequence includes the last node data pair: time 120 minutes, concentration 8.1 mmol / L. Substituting into the calculation formula, the negative descent rate parameter = (8.1 minus 11.2) divided by (120 minus 60) = -3.1 divided by 60 = -0.051 millimoles per liter per minute. Establishing the feature pair structure, the slope characteristic parameter for the ascending segment is 0.093, and the slope characteristic parameter for the descending segment is -0.051.

[0033] Table 3 Comparison of Slope Characteristic Parameters Subject number Blood glucose peak time parameter Concentration gradient in the rising segment Positive climb rate parameter negative fall rate parameter 001A 60 minutes 2.8 mmol / L 0.093 ﹣0.051 002B 45 minutes 3.5 mmol / L 0.116 ﹣0.062 003C 90 minutes 1.9 mmol / L 0.042 ﹣0.035 Please see Figure 6 The specific steps of S5 are as follows: S501: Obtain the standard peak time parameter within the authorized clinical medical knowledge graph node, call the blood glucose peak time parameter and the standard peak time parameter to perform time-domain difference calculation on the time axis, calibrate the time delay span index, perform linear normalization mapping on the time delay span index and the preset physiological extreme value, and generate the peak time deviation vector. The standard peak time parameter is obtained by acquiring continuous blood glucose monitoring data of healthy individuals with normal glucose tolerance characteristics from a pre-set sample dataset, extracting the time span records from the glucose load start node to the global maximum extreme value node of blood glucose concentration from the continuous blood glucose monitoring data, performing central trend statistical calculations on the entire time span record set, obtaining the time mean or median parameter, and setting the time node parameter that characterizes the normal physiological metabolic absorption rhythm as the standard peak time parameter. Connect to and retrieve from an authorized clinical medical knowledge graph database, requesting continuous blood glucose monitoring data packages from 500 healthy individuals with normal glucose tolerance characteristics, selected through rigorous screening. For each individual's test record within the continuous blood glucose monitoring data package, extract the timestamp from the oral glucose load initiation node to the node where the system detected the global maximum blood glucose concentration. Perform a subtraction operation between the two timestamps to collect time-span record sequences. Perform central trend statistics on all 500 time-span records, extracting the median value after ascending sequence sorting as the median time parameter. This median time parameter, representing the normal physiological metabolic absorption rhythm, is directly set as the standard peak time parameter. Obtain the previously calculated individual peak blood glucose time parameter and perform a time-domain difference calculation with this standard peak time parameter. The calculation formula is: Time Delay Span Index = Individual Peak Blood Glucose Time Parameter minus Standard Peak Time Parameter. Take the absolute value of the result to calibrate the time delay span index. The pre-defined physiological extreme value parameter, reflecting the maximum physiological tolerance time of the human body, is extracted. A linear normalization mapping operation is performed on the time delay span index, and the calculation formula is: Peak time deviation vector = Time delay span index divided by the pre-defined physiological extreme value. In this assignment example, the median time parameter is obtained by sorting the continuous time span records of 500 people and taking the median, which is 45 minutes. The standard peak time parameter is set to 45 minutes. The peak blood glucose time parameter of an individual obtained earlier is 60 minutes. Substituting into the calculation formula, the time delay span index = absolute value (60 minus 45) = 15 minutes, completing the calibration. The pre-defined physiological extreme value parameter is retrieved and set to the maximum delay tolerance limit of 120 minutes. Substituting into the formula again for normalization, the peak time deviation vector = 15 divided by 120 = 0.125. The peak time deviation vector value is 0.125.

[0034] S502: Call the preset recovery threshold parameter, perform bidirectional numerical boundary comparison detection between the descent slope feature parameter and the preset recovery threshold parameter, identify the attenuation fall compliance status, perform discrete category variable encoding assignment based on the attenuation fall compliance status, extract the scalar value of the blockage risk, and establish the glucose clearance blockage coefficient. The system initiates authorized access to the clinical database center node, collecting blood glucose test records from 300 individuals in a normal metabolic control group during the decline phase of an oral glucose tolerance test. For this complete set of decline-phase blood glucose test records, the concentration decline difference for each sample is extracted and divided by the corresponding decline time span, and all quotients are aggregated to construct a baseline decline slope set. A normal distribution fitting probability density analysis tool is used to perform statistical distribution analysis on the baseline decline slope set, extracting the lower limit critical slope value of the cumulative probability distribution curve corresponding to the 95% confidence interval. This rate index, representing the boundary of the normal body's glucose clearance compensation capacity, is explicitly set as the preset recovery threshold parameter. The system requests the individual decline slope feature parameters generated in the aforementioned steps from the register and performs a bidirectional numerical boundary comparison with the preset recovery threshold parameter. The system determines whether the absolute value of the individual decline slope feature parameter is greater than the absolute value of the preset recovery threshold parameter. If it is greater, the decline is identified as a normal decline; if it is less, the decline is identified as a delayed or blocked decline. Based on the assessment of the compliance status of the decline, discrete categorical variable encoding and assignment are performed. If the decline is normal, the scalar value of the blocking risk is assigned as 0; if the decline is delayed, the scalar value of the blocking risk is assigned as 1. This scalar value is encapsulated and stored in a memory pool to establish the glucose clearance blocking coefficient. The calculation process is shown below: Based on the statistical distribution analysis of 300 healthy individuals, the lower limit critical slope value of the corresponding 95% confidence interval is calculated to be -0.080 mmol / L / min, which is set as the preset recovery threshold parameter. The individual's decline slope characteristic parameter calculated in the previous steps is -0.051 mmol / L / min. A two-way numerical limit comparison is performed, and the absolute value of the two values ​​is taken; 0.051 is less than 0.080. The decline compliance status is identified as delayed. The encoding rule is triggered, and the discrete categorical variable encoding and assignment force the blocking risk scalar value to be set to 1. Thus, the glucose clearance blocking coefficient for this individual is successfully established as 1.

[0035] S503: Perform multi-dimensional feature splicing and combination based on the peak time deviation vector and glucose clearance blockade coefficient to reconstruct the individual metabolic baseline tensor. Call the rising slope feature parameter and the individual metabolic baseline tensor to perform matrix dot product operation to calibrate the basal absorption and metabolism score parameter. Perform addressing mapping on the basal absorption and metabolism score parameter within the dose interval mapping relationship to match the drug reversal estimate. Call the drug reversal estimate to perform unsupervised spatial clustering to define the intervention response echelon of the subjects and obtain efficacy grading evaluation indicators. The recovery threshold parameter is obtained by collecting blood glucose test records of the normal metabolic control group in the authorized clinical database during the decline phase. A baseline decline slope set is constructed by performing a quotient operation between the concentration decline difference and the corresponding time span for the blood glucose test records during the decline phase. Statistical distribution analysis is performed on the baseline decline slope set to extract the lower limit critical slope value of the preset probability confidence interval. The rate index that characterizes the boundary of the normal body's glucose clearance compensation ability is set as the recovery threshold parameter.

[0036] The peak time deviation vector and glucose clearance blockade coefficient generated earlier are extracted from the feature storage stack. A one-dimensional array structure is initialized in memory, with the peak time deviation vector as the first element and the glucose clearance blockade coefficient as the second element. A multi-dimensional feature concatenation and combination operation is performed to reconstruct an individual metabolic baseline tensor containing two dimensions. The aforementioned rising slope feature parameter is requested and expanded into a diagonal matrix structure consistent with the tensor dimension. The matrix processing function library is called to perform a standard matrix-dot product operation on the diagonal matrix formed by the rising slope feature parameter and the individual metabolic baseline tensor. The internal elements of the operation output are summed to obtain the basal absorption and metabolism score parameter. This score parameter is used as a query factor to initiate a hash addressing mapping operation on the dose interval mapping relationship network established in step S303, finding the matching node with the closest absolute difference, and extracting the parameter value stored in the corresponding node to match the medication reversal estimate. The medication reversal estimates of this subject and other subjects in the same batch are uniformly input into the K-means unsupervised spatial clustering algorithm engine. The number of cluster centers was set to 3. Euclidean distance was used to calculate the deviation between the sample points and each cluster center. The cluster centers were iteratively updated until the centroid position change was less than 0.001. Based on the clustering results, the current subjects were defined as belonging to high-response, medium-response, or low-response intervention response tiers. The tier labels were ultimately obtained as the output of efficacy grading evaluation indicators. The entire calculation process was as follows: the peak time deviation vector was extracted to a value of 0.125, and the glucose clearance blockade coefficient was set to 1. After splicing and combining, the reconstructed individual metabolic baseline tensor was: 0.125, 1. The slope feature parameter of the rising segment was extracted to be 0.093. This was expanded into a diagonal matrix, with all diagonal elements being 0.093. The matrix dot product operation was performed, and the formula was: Basal absorption metabolic score parameter = (0.125 x 0.093) + (1 x 0.093) = 0.011625 + 0.093 = 0.104625. Carrying 0.104625, we addressed the corresponding gradient within the dose interval mapping relationship. The mapping matching yielded a medication reversal estimate of 0.12. Integrating data from multiple subjects, we performed K-means unsupervised spatial clustering, iteratively calculating the Euclidean distance. This value of 0.12 was assigned to the second cluster. The subjects' intervention response echelon was defined as the intermediate response echelon, and the final efficacy grading index was determined to be an intermediate response.

[0037] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the technical solution.

Claims

1. A stratified randomized controlled trial method for evaluating traditional Chinese medicine compound interventions in prediabetes, characterized in that: Includes the following steps: S1: Collect fasting blood glucose measurements from prediabetic subjects, obtain the maximum fasting blood glucose difference, compare the maximum fasting blood glucose difference with the blood glucose stability interval parameter, obtain the average fasting blood glucose value, compare the average fasting blood glucose value with the preset threshold boundary parameter, and construct subject stratified coding. S2: Based on the subject stratification coding, the dosage gradient of ginseng and astragalus compound was extracted for the traditional Chinese medicine compound intervention for prediabetic subjects. The fasting blood glucose value after the intervention cycle and the baseline blood glucose value before the intervention cycle were collected. The difference was calculated and combined with the dosage gradient of ginseng and astragalus compound to construct a blood glucose change difference sequence. S3: Input the dose gradient of the ginseng and astragalus compound and the difference sequence of blood glucose changes into a piecewise linear regression function to obtain the inflection point of the dose response characteristics and merge them to construct a dose interval mapping relationship; S4: Collect glucose tolerance test values ​​of prediabetic subjects to obtain peak blood glucose time parameters. Based on the peak blood glucose time parameters, obtain the rising segment time data sequence and the falling segment time data sequence, and calculate the slope to obtain the rising segment slope characteristic parameters and the falling segment slope characteristic parameters. S5: Compare the blood glucose peak time parameter with the standard peak time parameter, the descending slope characteristic parameter with the preset recovery threshold parameter, and based on the comparison results, the ascending slope characteristic parameter and the dose interval mapping relationship, classify the prediabetic subjects and construct efficacy grading evaluation indicators.

2. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 1, characterized in that, The subject stratified coding includes a steady-state grading label, a blood glucose fluctuation penalty term, and an intervention matching index. The blood glucose change difference sequence includes single-dose attenuation magnitude, cumulative net response value, and baseline drift constant. The dose interval mapping relationship includes an effective onset threshold, an effect saturation boundary, and a preferred treatment window. The rising slope characteristic parameters include glucose absorption burst rate, early impedance calculus, and pre-peak secretory kinetic energy. The falling slope characteristic parameters include late fallback lag rate, islet compensatory half-life, and glucose clearance constant. The efficacy grading evaluation indicators include physiological reversal grading, relapse risk weighting, and long-term steady-state score.

3. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Collect fasting blood glucose measurement values ​​of prediabetic subjects, sort the fasting blood glucose measurement values ​​by time series, establish a blood glucose time detection sequence, perform difference calculation based on adjacent nodes of the blood glucose time detection sequence, extract the single detection difference sequence, call the single detection difference sequence to perform extreme value screening and comparison, select the upper bound span extreme value parameter, and obtain the maximum fasting blood glucose difference. S102: Collect the blood glucose stability interval parameters from the standard configuration library, compare the difference limit between the maximum fasting blood glucose difference and the blood glucose stability interval parameters, determine the blood glucose fluctuation trend, and if the blood glucose fluctuation trend meets the baseline stability condition, call the fasting blood glucose measurement value, perform span value accumulation on the fasting blood glucose measurement value, and perform equal division conversion in combination with the total number of recorded nodes to obtain the average fasting blood glucose value. S103: Obtain preset threshold boundary parameters, perform segmented interval value mapping comparison based on the average fasting blood glucose value and the preset threshold boundary parameters, determine the risk tier, associate the corresponding preset stratification code with the risk tier and extract the basic identity identifier of the subject, perform segment character sequence concatenation of the stratification code and identity identifier, combine the risk feature file, and establish subject stratification coding.

4. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 3, characterized in that, The specific process for setting the threshold boundary parameter involves: collecting fasting blood glucose baseline recording sequences from individuals diagnosed with prediabetes in a clinical historical database; performing kernel density estimation and cumulative probability distribution calculation on the fasting blood glucose baseline recording sequences; extracting morphological inflection points and discrete feature nodes corresponding to preset percentiles in the probability density distribution curve; extracting the theoretical boundary interval between impaired glucose tolerance and impaired fasting blood glucose in standard clinical disease diagnosis guidelines; performing an intersection mapping comparison between the extracted discrete feature nodes and the theoretical boundary interval; and extracting node data that are within the theoretical boundary interval and close to the trough of the probability density distribution curve as the critical judgment criterion for risk stratification to obtain the threshold boundary parameter.

5. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 1, characterized in that, The specific steps of S2 are as follows: S201: Perform clinical record retrieval for the subject stratified coding, extract the list of traditional Chinese medicine compound preparations, divide the list of traditional Chinese medicine compound preparations into concentration increments, determine the dosage parameters, perform sequence combination of the dosage parameters according to the time sequence arrangement rules, perform vectorized coding in combination with the interval parameter, and establish the dosage gradient of ginseng and astragalus compound preparations. S202: Based on the dosage gradient of the ginseng and astragalus compound, collect the pre-intervention period physiological records of prediabetic subjects in the pre-set sample data, extract the baseline blood glucose value before intervention, obtain the fasting blood glucose value after intervention from the physiological records of prediabetic subjects after the intervention period of the traditional Chinese medicine compound, perform cleaning on the fasting blood glucose value after intervention and the baseline blood glucose value before intervention, remove abnormal isolated points, perform pair mapping on the remaining values, and obtain the blood glucose monitoring pairs of the subjects. S203: Call the subject's blood glucose monitoring pair, perform parsing and extract periodic measurement parameters, perform node subtraction operation on the periodic measurement parameters, calculate the node concentration fluctuation deviation, perform timestamp alignment and merging based on the node concentration fluctuation deviation and the dose gradient of the ginseng and astragalus compound, establish dose-effect key-value pairs, perform ascending sorting and reconstruction of the dose-effect key-value pairs according to the dose increment direction, and generate a blood glucose change difference sequence.

6. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 1, characterized in that, The specific steps for S3 are as follows: S301: Call the dose gradient of the ginseng and astragalus compound and the difference sequence of blood glucose changes, set the dose gradient of the ginseng and astragalus compound as the independent variable benchmark, set the difference sequence of blood glucose changes as the dependent variable target combination to construct a two-dimensional coordinate discrete point matrix, perform piecewise linear regression operation on the two-dimensional coordinate discrete point matrix, calculate the slope value corresponding to the line connecting the discrete nodes, extract the intercept parameter and regression coefficient corresponding to the slope change stage, and obtain the regression slope feature parameter set; S302: Based on the regression slope feature parameter set, perform critical state detection, extract the absolute value of the difference between adjacent regression coefficients in the regression slope feature parameter set, perform global extreme value retrieval on the absolute value difference sequence, obtain the node index of the sequence where the maximum difference is located, call the node index to perform coordinate mapping addressing to the dosage gradient of the ginseng and astragalus compound, extract the scale value at the corresponding position, and obtain the inflection point of the dose response feature. S303: Based on the inflection point of the dose response feature, perform threshold segmentation on the difference sequence between the dose gradient of the ginseng and astragalus compound and the change in blood glucose, extract the data subset of independent intervals on both sides of the inflection point, perform position recombination on the associated parameters within the data subset according to the ascending order rule, perform fusion splicing on adjacent and overlapping intervals after recombination, construct the associated key-value pairs between the spliced ​​dose values ​​and the magnitude of blood glucose change, and establish a dose interval mapping relationship.

7. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 6, characterized in that, The process of extracting the intercept parameter and regression coefficient corresponding to the slope variation stage is as follows: First, the local slope sequence is calculated by extracting the ratio of the difference between the ordinate and abscissa of each adjacent node in the two-dimensional coordinate discrete point matrix. Second, the local slope sequence is segmented according to a preset slope tolerance parameter to obtain monotonically changing sub-intervals. Third, the intercept parameter and regression coefficient corresponding to the monotonically changing sub-interval are calculated by calling the coordinate data of the boundary nodes of the monotonically changing sub-interval. Fourth, the intercept parameter and regression coefficient corresponding to the monotonically changing sub-interval are arranged in ascending order to obtain the regression slope feature parameter set. The preset slope tolerance parameter is set by extracting the maximum variation range of the historical test local slope sequence and setting the product of the maximum variation range and the inherent error weight as the preset slope tolerance parameter. The inherent error weight is set by calling the permissible error limit percentage value in the factory parameters of the test equipment as the inherent error weight.

8. The stratified randomized controlled trial method for evaluating the intervention of traditional Chinese medicine compound prescriptions in prediabetes according to claim 1, wherein the specific steps of S4 are as follows: S401: Obtain the glucose tolerance test values ​​corresponding to multiple measurement nodes of prediabetic subjects in the authorized public historical test dataset, perform time node concentration correlation comparison and value traversal for the glucose tolerance test values, extract the global maximum concentration extreme value node in the concentration sequence, call the global maximum concentration extreme value node mapping timestamp identifier, and obtain the blood glucose peak time parameter. S402: Based on the blood glucose peak time parameter, the glucose tolerance test values ​​are divided into time axes, and the concentration values ​​corresponding to the preceding interval of the blood glucose peak time parameter are extracted and combined to establish a pre-sequence. The concentration values ​​corresponding to the following interval of the blood glucose peak time parameter are extracted and combined to establish a post-sequence, thus obtaining the rising segment time data sequence and the falling segment time data sequence. S403: Perform a quotient operation by dividing the first and last concentration difference in the rising segment time data sequence by the time span to obtain the positive climbing rate parameter; perform a quotient operation by dividing the first and last concentration deviation in the falling segment time data sequence by the time span to obtain the negative falling rate parameter; establish a feature pair between the positive climbing rate parameter and the negative falling rate parameter to generate the rising segment slope feature parameter and the falling segment slope feature parameter.

9. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 1, characterized in that, The specific steps of S5 are as follows: S501: Obtain the standard peak time parameter within the authorized clinical medical knowledge graph node, call the blood glucose peak time parameter and the standard peak time parameter to perform time axis time domain difference calculation, calibrate the time delay span index, perform linear normalization mapping between the time delay span index and the preset physiological extreme value, and generate the peak time deviation vector. S502: Call the preset recovery threshold parameter, perform bidirectional numerical boundary comparison and detection between the descending slope feature parameter and the preset recovery threshold parameter, identify the attenuation fallback compliance status, perform discrete category variable encoding and assignment based on the attenuation fallback compliance status, extract the scalar value of the blockage risk, and establish the glucose clearance blockage coefficient. S503: Perform multi-dimensional feature splicing and combination based on the peak time deviation vector and glucose clearance blockade coefficient to reconstruct the individual metabolic baseline tensor. Call the rising slope feature parameter and the individual metabolic baseline tensor to perform matrix-dot product operation to calibrate the basal absorption and metabolism score parameter. Perform addressing mapping on the basal absorption and metabolism score parameter within the dose interval mapping relationship to match the drug reversal estimate. Call the drug reversal estimate to perform unsupervised spatial clustering to define the intervention response echelon of the subjects and obtain efficacy grading evaluation indicators.

10. The stratified randomized controlled trial method for evaluating traditional Chinese medicine compound intervention in prediabetes according to claim 9, characterized in that, The standard peak time parameter is obtained by acquiring continuous blood glucose monitoring data of healthy individuals with normal glucose tolerance characteristics from a pre-set sample dataset, extracting the time span records from the glucose load start node to the global maximum extreme value node of blood glucose concentration from the continuous blood glucose monitoring data, performing central trend statistical calculations on the entire time span record set, obtaining the time mean or median parameter, and setting the time node parameter that characterizes the normal physiological metabolic absorption rhythm as the standard peak time parameter. The recovery threshold parameter is obtained by collecting blood glucose test records of the normal metabolic control group in the authorized clinical database during the decline phase. A baseline decline slope set is constructed by performing a quotient operation between the concentration decline difference and the corresponding time span for the blood glucose test records during the decline phase. Statistical distribution analysis is performed on the baseline decline slope set to extract the lower limit critical slope value of the preset probability confidence interval. The rate index that characterizes the boundary of the normal body's glucose clearance compensation ability is set as the recovery threshold parameter.

Images