A Smart Marketing System and Method for Blood Sugar Control Members Based on Glycemic Index (GI) Dietary Data

Abstract

This application provides a smart marketing system and method for blood sugar control members based on GI dietary data. It includes: a GI dietary data management module for establishing a GI dietary database and generating meal load characterizations; a glycemic response modeling module for correlating meal load characterizations with members' post-meal blood glucose feedback according to meal time, and calculating personalized glycemic coefficients for members under different dietary scenarios based on the correlation results; a blood sugar control profile generation module for generating dynamic and reliable blood sugar control member profiles; a member stratification decision module for determining the blood sugar control stratification type of a member based on the blood sugar control member profile; and a marketing closed-loop optimization module for generating differentiated marketing outreach plans based on candidate content and push constraints, and updating personalized glycemic coefficients, blood sugar control member profiles, and marketing decision parameters based on post-outreach behavioral and blood glucose feedback. This application can improve the accuracy of blood sugar control profiles, enhance product matching accuracy, and strengthen marketing closed-loop optimization capabilities.

Description

Technical Field

[0001] This application relates to the fields of digital health management and intelligent marketing technology, and in particular to an intelligent marketing system and method for blood sugar control members based on GI dietary data. Background Technology

[0002] With the development of health-conscious consumption and digital membership operations, low-GI foods, blood sugar control packages, and nutrition management services targeting people with diabetes are gradually becoming important marketing scenarios for retail platforms and health food companies. GI is used to characterize the relative ability of food to raise blood sugar, while GL further incorporates intake to reflect dietary glycemic load. Therefore, analyzing members' blood sugar control needs based on GI dietary data can provide a data foundation for recommending low-GI products and refining membership operations.

[0003] Existing blood sugar control membership marketing systems typically segment members based on static business data such as age, gender, spending amount, purchase frequency, and membership level, and then push low-GI products, coupons, or dietary content according to preset operating rules. Although some systems can label whether a product is a low-GI food, product recommendations still mainly rely on product labels and historical consumption records, without incorporating a unified analysis of members' daily dietary structure, meal GI composition, GL load level, postprandial blood glucose feedback, and changes in dietary habits.

[0004] Therefore, existing technologies struggle to reflect the individualized glycemic responses of different members to the same food or product, leading to inaccurate matching of low-GI products, insufficient adaptability of dietary plans, and severe homogenization of marketing pushes. Furthermore, members' dietary behavior, blood glucose feedback, purchase conversion, and repurchase results cannot be used to update member profiles and marketing strategies. The lack of a closed-loop linkage mechanism between dietary data, blood glucose control profiles, precise segmentation, and marketing outreach results in increased ineffective pushes, insufficient member loyalty, and limited repurchase growth. Summary of the Invention

[0005] In view of this, this application provides a smart marketing system and method for blood sugar control members based on GI dietary data, in order to solve the problems of crude blood sugar control profiles, inaccurate product matching, and lack of marketing loop in the existing technology.

[0006] The first aspect of this application provides a smart marketing system for blood sugar control members based on GI dietary data, comprising: a GI dietary data management module, used to establish a GI dietary database containing food GI parameters, GL load rules, and product blood sugar control adaptation relationships, and to match food objects and product objects in member dietary behavior data to generate meal dietary load representations; and a glycemic response modeling module, used to correlate meal dietary load representations with member post-meal blood glucose feedback according to meal time, calculate personalized glycemic coefficients for members in different dietary scenarios based on the correlation results, and use personalized glycemic coefficients to correct product blood sugar control adaptation rights. The system comprises four modules: a blood sugar control profile generation module, which integrates corrected product blood sugar control adaptation weights, member dietary habits, historical consumption behavior, and marketing feedback data to generate a dynamically credible blood sugar control member profile; a member segmentation decision module, which determines the blood sugar control segment type of a member based on the blood sugar control member profile and configures candidate low-GI products, candidate dietary plans, candidate member benefits, and candidate push frequencies for each blood sugar control segment type; and a marketing closed-loop optimization module, which generates differentiated marketing outreach plans based on candidate content and push constraints, and updates personalized glycemic indexes, blood sugar control member profiles, and marketing decision parameters based on behavioral and blood sugar feedback after outreach.

[0007] The second aspect of this application provides a smart marketing method for blood sugar control members based on GI dietary data, using the system of the first aspect. The method includes: acquiring member dietary behavior data and calling a GI dietary database to match food objects and product objects in the member dietary behavior data to generate a meal load characterization; associating the meal load characterization with the member's post-meal blood glucose feedback according to meal time, calculating the member's personalized glycemic index in the dietary scenario, and using the personalized glycemic index to correct the product's blood sugar control adaptation weight; integrating the corrected product's blood sugar control adaptation weight, member behavior data, and marketing feedback data to generate a blood sugar control member profile, and determining the member's blood sugar control stratification type based on the blood sugar control member profile; generating a differentiated marketing outreach plan based on the blood sugar control stratification type, and updating the personalized glycemic index, blood sugar control member profile, and marketing decision parameters based on post-outreach behavioral and blood glucose feedback.

[0008] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: The GI dietary data management module is used to establish a GI dietary database containing food GI parameters, GL load rules, and product glycemic control adaptation relationships. It matches food objects and product objects in member dietary behavior data to generate meal dietary load representations. The glycemic response modeling module is used to associate meal dietary load representations with member post-meal blood glucose feedback according to meal time. Based on the association results, it calculates personalized glycemic coefficients for members in different dietary scenarios and uses personalized glycemic coefficients to correct product glycemic control adaptation weights. The glycemic control profile generation module is used to integrate the corrected product glycemic control adaptation weights, member dietary habits, historical consumption behavior, and marketing feedback data to generate a dynamically reliable glycemic control member profile. The member stratification decision module is used to determine the glycemic control stratification type of a member based on the glycemic control member profile and configure candidate low-GI products, candidate dietary plans, candidate member benefits, and candidate push frequencies for the glycemic control stratification type. The marketing closed-loop optimization module is used to generate differentiated marketing outreach plans based on candidate content and push constraints, and update personalized glycemic coefficients, glycemic control member profiles, and marketing decision parameters based on behavioral feedback and blood glucose feedback after outreach. This application can improve the accuracy of blood sugar profiles, enhance the accuracy of product matching, and strengthen the marketing loop optimization capabilities. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of the structural composition of the intelligent marketing system for blood sugar control members based on GI dietary data provided in this application embodiment; Figure 2 This is a flowchart illustrating the intelligent marketing method for blood sugar control members based on GI dietary data provided in this application embodiment. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] With the development of low-GI foods, blood sugar control packages, and health membership services, membership marketing targeting people with blood sugar control is gradually shifting from general consumer recommendations to precise operations that combine dietary health needs. GI dietary data reflects the relative ability of food to raise blood sugar, and GL load reflects the impact of mealtime diets on blood sugar fluctuations when combined with actual intake. Therefore, using GI dietary data for blood sugar control member analysis and marketing outreach has become an important development direction for health food retail, nutrition management platforms, and membership operation systems.

[0013] In existing technologies, blood sugar control membership marketing typically segments members based on static business data such as age, gender, spending amount, purchase frequency, and membership level, and pushes low-GI products, preferential benefits, or blood sugar control content according to manually configured product tags or operational rules. Although some systems can label whether a product is a low-GI food, their recommendation logic still mainly relies on product attributes and historical consumption records, failing to uniformly model members' daily dietary records, meal GI composition, GL load level, postprandial blood glucose feedback, and changes in dietary habits, making it difficult to accurately depict members' true blood sugar control needs.

[0014] Therefore, it is evident that existing technologies suffer from at least the following problems: crude blood sugar profiles, inaccurate product matching, and a lack of a closed-loop marketing system. Specifically, existing member profiles cannot reflect the individualized dietary structure and differences in glycemic response among members, and the suitability of the same low-GI product for different members cannot be effectively differentiated. At the same time, the results of clicks, purchases, repeat purchases, blocking, and post-meal blood sugar feedback after marketing pushes are difficult to use to correct member profiles and recommendation strategies, resulting in homogenized low-GI product recommendations, a large number of ineffective outreach, and limited improvement in member stickiness and repeat purchases.

[0015] To address the aforementioned issues, this application provides a smart marketing system for blood sugar control members based on GI dietary data. The system establishes a GI dietary database containing food GI parameters, GL load rules, and product blood sugar control adaptation relationships through a GI dietary data management module. It matches food and product objects in member dietary behavior data to generate meal load representations. A glycemic response modeling module correlates meal load representations with member post-meal blood glucose feedback according to meal time, calculating personalized glycemic coefficients for members in different dietary scenarios and using these personalized glycemic coefficients to adjust product blood sugar control adaptation weights. A blood sugar control profile generation module integrates the adjusted product blood sugar control adaptation weights, member dietary habits, historical consumption behavior, and marketing feedback data to generate a dynamically credible blood sugar control member profile. A member segmentation decision module determines the member's blood sugar control segment type and configures candidate low-GI products, candidate dietary plans, candidate member benefits, and candidate push frequencies for that segment. Finally, a marketing closed-loop optimization module generates differentiated marketing outreach plans and updates personalized glycemic coefficients, blood sugar control member profiles, and marketing decision parameters based on post-outreach behavioral and blood glucose feedback.

[0016] Through the above technical solution, this application can form a closed-loop linkage between GI dietary data, GL load calculation, postprandial blood glucose feedback, personalized glycemic index modeling, blood sugar control member profile, and marketing decision-making process. This transforms member segmentation from static consumption segmentation to dynamic segmentation based on dietary behavior and glycemic response, improves the accuracy of blood sugar control profiles and the accuracy of low-GI product matching, reduces homogeneous push notifications and ineffective marketing outreach, and continuously enhances the adaptive optimization capability of marketing strategies through feedback iteration.

[0017] The specific components and functions of the intelligent marketing system for blood sugar control members based on GI dietary data provided in this application will be described in detail below with reference to the accompanying drawings and specific embodiments. Figure 1 This is a schematic diagram of the structural composition of the intelligent marketing system for blood sugar control members based on GI dietary data provided in this application embodiment, as shown below. Figure 1 As shown, the system may specifically include the following components: The GI dietary data management module 101 is used to establish a GI dietary database containing food GI parameters, GL load rules and product sugar control adaptation relationships, and to match food objects and product objects in member dietary behavior data to generate meal dietary load representations. The glycemic response modeling module 102 is used to correlate the meal dietary load characterization with the member's post-meal blood glucose feedback according to the meal time, calculate the member's personalized glycemic coefficient under different dietary scenarios based on the correlation results, and use the personalized glycemic coefficient to correct the product's blood sugar control adaptation weight. The sugar control profile generation module 103 is used to integrate the corrected product sugar control adaptation weight, member dietary habits, historical consumption behavior and marketing feedback data to generate a sugar control member profile with dynamic credibility. The member segmentation decision module 104 is used to determine the blood sugar control segmentation type of a member based on the blood sugar control member profile, and to configure candidate low-GI products, candidate dietary plans, candidate member benefits and candidate push frequency for the blood sugar control segmentation type; The marketing closed-loop optimization module 105 is used to generate differentiated marketing outreach plans based on candidate content and push constraints, and update personalized glycemic index, blood sugar control member profile and marketing decision parameters based on behavioral feedback and blood sugar feedback after outreach.

[0018] In some embodiments, a GI dietary database is established, including food GI parameters, GL load rules, and product glycemic control adaptation relationships, including: Acquire food nutrition data and product nutritional attribute data, and standardize the food nutrition data and product nutritional attribute data to generate unified dietary data; Based on the food composition characterization in the unified dietary data, the food GI parameters corresponding to the food objects are determined, and GL load rules are constructed based on the food GI parameters, available carbohydrate characterization, and intake portion characterization. Based on the semantic matching relationship between product ingredient representation and food objects, generate product sugar control adaptation relationship; Food GI parameters, GL load rules, and product sugar control adaptation relationships are linked and stored according to data reliability to form a dynamically updated GI dietary database.

[0019] Specifically, during the deployment phase of the intelligent marketing platform for blood sugar control members, the system first establishes a basic dietary data access channel, acquiring food nutrition data and product nutritional attribute data from public nutrient composition databases, enterprise product databases, supplier formula systems, product packaging identification systems, and manual verification terminals. Food nutrition data may include food name, food category, standard GI reference value, energy per unit mass, total carbohydrate content, available carbohydrate content, dietary fiber content, protein content, fat content, processing method, and common serving size.

[0020] Product nutritional attribute data can include product name, specifications, ingredient composition, nutrition facts, recommended meal times, sugar control label, sales unit, and information on alternative products. For example, when the platform integrates products such as low-GI whole wheat bread, instant oatmeal, multigrain rice balls, and sugar-free yogurt, it first parses the product ingredients and nutrition facts, and then maps the main carbohydrate sources in the product to the corresponding food objects.

[0021] During data standardization, the system performs unit conversion, name normalization, ingredient field alignment, and outlier validation on data from different sources. For cases where the same food has multiple names, such as oatmeal, instant oats, and pure oats, the system merges them into a unified food object using a food alias dictionary and semantic vector matching. For data expressed per serving, per bag, or per 100 grams in product nutrition facts tables, the system uniformly converts them to standard units of measurement. For data lacking available carbohydrate content, the system estimates it based on total carbohydrate and dietary fiber content, assigning a lower confidence level to the estimation results. After these processes, unified dietary data is generated, with each data item retaining its source identifier, update time, and validation status.

[0022] When determining the GI parameters of food, the system uses food composition characterization in unified dietary data as a basis, combining food category, processing method, and GI reference source to generate the corresponding GI parameters for the food object. For foods with authoritative GI reference values, the corresponding standard values ​​are directly adopted and the reference source is recorded; for foods lacking direct GI values ​​but with similar compositions, the system generates estimated GI parameters based on the GI parameters of similar food objects, differences in processing methods, and differences in carbohydrate structure. For example, ordinary white rice can be assigned a higher GI parameter, mixed grain rice can generate a corrected GI parameter based on the proportion of whole grains and dietary fiber content, and low-GI whole wheat bread can generate corresponding GI parameters based on the proportion of whole wheat flour, sugar content, and dietary fiber content.

[0023] When constructing GL load rules, the system correlates food GI parameters, available carbohydrate characterization, and serving size characterization to form GL load rules applicable to meal calculations. For a single food, the system calculates the GL value per serving based on the standard serving size; for combination or set meals, the system first breaks down the main food items in the product and then calculates the overall GL load according to the proportion of each food item in the product.

[0024] For example, a multigrain rice ball contains brown rice, oats, beans, and a small amount of sauce. The system determines the GI parameters and available carbohydrate content of each component food, and then aggregates them according to the recipe proportions to generate a commercial-grade GL load rule. GL load rules can also be bound to applicable meal times and intake scenarios, making load calculations different for different scenarios such as breakfast, lunch, and snacks.

[0025] When generating sugar control adaptation relationships for products, the system performs semantic matching between product component representations and food objects, and combines the product's carbohydrate source, added sugar level, dietary fiber level, protein supplementation level, and corresponding GL load rules to determine the degree of adaptation between the product and sugar control dietary scenarios. For products suitable as staple food substitutes, the system establishes a substitution relationship between the product and high-GI staple foods; for products suitable as snacks, the system establishes an adaptation relationship between the product and low-load snack scenarios. For example, low-GI whole wheat bread can establish a substitution recommendation relationship with regular sweet bread, sugar-free yogurt can establish a substitution recommendation relationship with sugary dairy beverages, and small packages of nuts can establish a snack substitution relationship with high-sugar snacks. The product sugar control adaptation relationship also records the adaptation level, adaptation reason, contraindications, and recommended scenarios.

[0026] After completing the above processing, the system associates and stores food GI parameters, GL load rules, and product blood sugar control adaptation relationships according to data reliability, forming a GI dietary database. Data reliability can be determined based on the authority of the data source, field completeness, manual verification status, and consistency with historical feedback. When product formulas are updated, nutrition facts tables are changed, member post-meal blood glucose feedback accumulates, or manual verification results change, the system triggers a dynamic database update, recalculates the GI parameters, GL load rules, and blood sugar control adaptation relationships of the affected food and product objects, and retains historical versions.

[0027] Therefore, the GI dietary database can not only support static nutrition queries, but also support subsequent meal load analysis, personalized glycemic index modeling, and accurate matching of low-GI products, thereby improving the data foundation accuracy for blood sugar control member profiling and marketing decisions.

[0028] In some embodiments, food objects and commodity objects in member dietary behavior data are matched to generate a meal load characterization, including: Acquire members' dietary behavior data, and process the dietary records, product consumption records, and intake representations into events according to member identification and meal time to generate meal dietary events; Call the GI dietary database to perform semantic matching on food objects and product objects in meal dietary events, and determine the corresponding food GI parameters, GL load rules and product sugar control adaptation relationship; The dietary load within the same meal is weighted and aggregated based on the matching results, and a meal dietary load characterization is generated by combining the matching confidence.

[0029] Specifically, after a member completes their meal record, purchase of goods, or sugar control check-in, the system retrieves the member's dietary behavior data from the member's mini-program, the POS order system, the barcode scanner, and the dietary record interface. This data can include member ID, meal time, meal type, food name, product code, quantity purchased, consumption ratio, serving size, record source, and upload time. For example, if member A uploads a picture of their breakfast and purchases low-GI whole-wheat bread and sugar-free yogurt during the same time period, the system first identifies food objects such as eggs, bread, and yogurt in the picture, then combines this with the product codes in the order to confirm the corresponding products (bread and yogurt), avoiding incomplete food identification due to relying solely on image recognition.

[0030] During event-based processing, the system uses member ID as the attribution index and meal time as the time sequence boundary to integrate dietary records, product consumption records, and portion sizes into a single meal event. If multiple records exist for the same member within a preset time window, the system merges them based on meal type, time difference, and product consumption status. For products purchased but not yet confirmed for consumption, the system configures a pending confirmation status based on the member's historical consumption habits and check-in records. For example, if member A uploads a breakfast photo at 8:10 AM, scans a code at 8:15 AM to confirm consuming one slice of low-GI whole-wheat bread, and records drinking 200 ml of unsweetened yogurt at 8:20 AM, the system merges these data into a single breakfast meal event and records the food item, product item, and portion size separately.

[0031] After generating a meal event, the system calls the GI diet database to perform semantic matching on food and product objects within the meal event. For standard food objects, the system matches the corresponding food GI parameters based on food name, alias, category, and ingredient characteristics. For packaged product objects, the system matches the corresponding product's sugar control compatibility based on product code, ingredient composition, and nutritional attributes. For dishes or combination meals, the system first breaks down the main carbohydrate sources and then matches the food objects separately. For example, a multigrain rice ball is broken down into components such as brown rice, oats, beans, and sauce. The system determines the GI parameters and GL loading rules for each component and then generates a combination matching result based on the recipe proportions.

[0032] During semantic matching, the system assigns a matching confidence level to each matching result. For product objects directly matched via product codes, the matching confidence level is high. For food objects matched via image recognition and name similarity, the system combines recognition confidence, historical consumption records, and manual correction results to determine the matching confidence level. For ambiguous food names manually entered by members, such as multigrain bread or low-sugar biscuits, the system assigns a medium or low matching confidence level based on ingredient similarity and differences between candidate objects. If multiple candidate matching objects exist, the system prioritizes the object with higher consistency with the member's historical consumption records, portion size, and meal scenario.

[0033] During the dietary load aggregation phase, the system calculates the individual dietary load of each food or product within a meal based on the matched food GI parameters, GL load rules, and intake portion representation, and then weights and aggregates them according to the actual intake proportion. For a breakfast consisting of one slice of low-GI whole-wheat bread, 200 ml of unsweetened yogurt, and one egg, since the egg contributes relatively little to the carbohydrate load, the system primarily calculates the meal GL load based on the available carbohydrate content, GI parameters, and intake portion of the bread and yogurt. Protein-rich foods are also included as a meal structure regulator in the overall representation. For members who purchase only two items but confirm they will only consume one, the system prioritizes the confirmed consumption data rather than directly including the entire purchase amount in the meal load.

[0034] When generating a meal load characterization, the system not only outputs the meal GL load value, but also generates the meal GI composition, main sources of glycemic index, product glycemic control adaptation results, intake portion confidence, and overall matching confidence. The overall matching confidence can be determined comprehensively based on food identification confidence, product code matching results, portion estimation reliability, and data source consistency. For example, if the confidence level of breakfast photo identification is 0.82, the product code matching confidence is 0.98, and the intake portion is manually confirmed by the member, then the meal load characterization has high confidence; if it is generated only from fuzzy text records, the system reduces the sample weight in subsequent glycemic index modeling.

[0035] Through the above processing, the system can transform fragmented dietary records, product consumption records, and portion size data into structured meal-based dietary events, and form a meal-based dietary load characterization that can be used for glycemic response analysis based on a GI dietary database. This approach improves the accuracy of matching food and product objects, reduces the bias caused by judging blood sugar control needs solely based on orders or manual labels, and provides a reliable data foundation for subsequent personalized glycemic index calculations, blood sugar control member profile generation, and differentiated marketing outreach.

[0036] In some embodiments, the glycemic response modeling module is specifically used for: Establish a temporal correspondence between meal dietary load characterization and members' postprandial blood glucose feedback based on meal time, and reduce the credibility weight of data that exceed the effective feedback window to generate glycemic response samples; Based on the glycemic response samples, the blood glucose response deviation information of members in dietary scenarios is extracted, and a personalized glycemic model is trained by combining historical response features to obtain the personalized glycemic coefficient. The product sugar control adaptation relationship in the GI dietary database is individually modified based on the personalized glycemic index, and the modified product sugar control adaptation weight is generated.

[0037] Specifically, during the operation of the blood sugar control member intelligent marketing system, the glycemic response modeling module, based on the meal dietary load characterization, further integrates members' post-meal blood glucose feedback data. This post-meal blood glucose feedback data can come from manual entry by members, synchronization with home blood glucose meters, synchronization with continuous glucose monitoring devices, or upload via a health management interface. The data must include at least the member's identifier, testing time, blood glucose value, data source, device status, and upload time.

[0038] The system first standardizes the meal times in the meal load characterization and then establishes a time-series correspondence based on the post-meal feedback window. For example, if member A completes breakfast recording at 7:50 AM, the meal load characterization shows that breakfast includes low-GI whole-wheat bread, unsweetened yogurt, and eggs, with a comprehensive GL load of low to medium levels. The system retrieves the blood glucose data uploaded by member A between 8:20 AM and 10:20 AM and associates the blood glucose values ​​at 30 minutes, 60 minutes, and 120 minutes post-meal with that breakfast meal.

[0039] When establishing time-series correspondences, the system configures different effective feedback windows based on meal type. For breakfast, lunch, and dinner, the effective feedback window can be set from 30 minutes to 180 minutes after the meal; for snacks, the effective feedback window can be set from 20 minutes to 120 minutes after the meal. If the blood glucose feedback time falls within the effective feedback window, the system configures sample weights according to the time interval and the reliability of the data source. If the blood glucose feedback exceeds the effective feedback window, the system does not discard it directly, but performs reliability reduction processing to reduce the impact of abnormal delayed uploads, missed meal recordings, or blood glucose data across meal times on model training. For example, if member A records their blood glucose value 4 hours after breakfast, the system marks this data as a weakly correlated feedback and reduces its weight when generating glycemic response samples.

[0040] When generating glycemic response samples, the system encapsulates meal load characterization, postprandial blood glucose feedback, and member status data into samples. Sample content may include meal GL load, main sources of glycemic load, food GI parameters, intake portion reliability, pre-meal blood glucose, postprandial peak, 2-hour postprandial blood glucose, blood glucose rise rate, blood glucose fall time, and feedback reliability.

[0041] For samples with records of exercise, staying up late, medication, alcohol consumption, or abnormal eating, the system encodes these factors as interference state representations for subsequent bias correction. For example, if member A briskly walks 30 minutes after breakfast and their post-meal blood glucose peak is lower than historical levels for similar meals, the system will include the exercise state as an interference factor in the glycemic response sample to avoid the model misjudging that low-GI whole wheat bread has no effect on member A's glycemic response.

[0042] When extracting glycemic response bias information, the system first generates an expected glycemic baseline for the corresponding meal load based on the GI dietary database and historical samples of the population. Then, it compares the member's actual postprandial blood glucose response with the expected glycemic baseline. If the actual blood glucose rise is higher than expected, positive bias information is generated, indicating that the member has a high glycemic sensitivity to this type of dietary scenario; if the actual blood glucose rise is lower than expected, negative bias information is generated, indicating that the member has a low glycemic sensitivity to this type of dietary scenario. Bias information can also be categorized according to dietary scenarios such as staple foods, dairy products, grains, and snacks, enabling the model to identify members' differentiated responses to different food categories.

[0043] When training the personalized glycemic index (GI) model, the system performs a time-series fusion of current GI response samples and historical response features. Historical response features include recent similar meal responses, long-term average fluctuation levels, food category sensitivity, meal time sensitivity, and historical sample reliability. The model can be jointly trained using a time-series coding network and individual embedding parameters. This learns both the fundamental relationship between GI parameters, GL load, and blood glucose fluctuations, and the corrective effect of individual member differences on the GI response. When the number of member samples is small, the system uses response parameters from similar member groups as initial constraints; as member samples accumulate, the system gradually increases the weight of individual historical samples, outputting a more stable personalized GI coefficient.

[0044] After obtaining the personalized glycemic index, the system individually adjusts the glycemic control compatibility of products in the GI dietary database. For member A, if multiple breakfast samples show a high glycemic response to bread-based carbohydrates, even if a certain low-GI whole-wheat bread has a high glycemic control compatibility level in the database, the system will lower the glycemic control compatibility weight of that product in member A's profile and prioritize increasing the recommendation weight of oatmeal, nut, or high-protein breakfast products. Conversely, if member A has a stable response to sugar-free yogurt as a snack, the system will increase the compatibility weight of the corresponding product in the snack scenario. The adjusted product glycemic control compatibility weights are stored in association with member ID, dietary scenario, model confidence, and update time.

[0045] Through the above implementation methods, the glycemic response modeling module can individually correlate standard GI and GL data with members' actual postprandial blood glucose feedback, avoiding uniform recommendations based solely on product low-GI labels. This approach can more accurately distinguish the differences in glycemic responses among different members to the same product, improve the individualization of product glycemic control adaptation weights, and provide a more reliable model basis for generating glycemic control member profiles and differentiated marketing outreach.

[0046] In some embodiments, blood glucose response deviation information of members in dietary scenarios is extracted based on glycemic response samples, and a personalized glycemic model is trained by combining historical response features to obtain a personalized glycemic coefficient, including: The glycemic response samples were normalized to generate a standard glycemic response characterization corresponding to the member's dietary scenario. Based on the difference between the standard glycemic response characterization and the preset glycemic baseline characterization, the blood glucose response deviation information of members in the corresponding dietary scenarios is extracted; Blood glucose response deviation information and historical response features are fused over time to form individual member response training samples; A personalized glycemic index model is trained based on individual member response training samples, and the output is a personalized glycemic index coefficient that characterizes the member's sensitivity to glycemic index in different dietary scenarios.

[0047] Specifically, during the personalized glycemic index model training phase, the system first extracts valid samples corresponding to the target members from the glycemic index response sample library. Each glycemic index response sample is associated with meal dietary load representation, postprandial blood glucose feedback, meal time, intake quantity reliability, and feedback reliability. To avoid incomparability of model inputs due to different meals, different intake quantities, and different feedback times, the system performs scenario normalization processing on the glycemic index response samples.

[0048] Scenario normalization can include meal type normalization, GL load interval normalization, feedback time normalization, and interference factor normalization. For example, member A consumes low-GI whole-wheat bread and sugar-free yogurt for breakfast and multigrain rice and chicken breast for lunch. The actual intake, post-meal feedback time, and exercise status of the two meals are different. The system converts these into standard glycemic response representations under the medium GL scenario for breakfast and the medium GL scenario for lunch, respectively, so that subsequent models can compare the differences in glycemic response under similar dietary scenarios.

[0049] When generating a standardized glycemic response representation, the system converts postprandial blood glucose feedback into a unified feature scale, including pre-meal baseline blood glucose, postprandial peak increment, 2-hour postprandial decline, area of ​​fluctuation of the blood glucose curve, and feedback reliability. For data collected 45 minutes after a meal, the system estimates its response value at the standard 60-minute node using a time interpolation model; for cases where data from 120 minutes after a meal is missing but continuous blood glucose monitoring curves exist, the system extracts the corresponding node values ​​from the curves; for data with only manually entered single-point blood glucose, the system retains the single-point glycemic amplitude and reduces the weight of curve integrity. Thus, the system can unify blood glucose feedback from different sources and at different time granularities into a standardized glycemic response representation that can be used for training.

[0050] Subsequently, the system extracts blood glucose response deviation information based on a preset glycemic baseline representation. This preset glycemic baseline representation can be determined jointly by food GI parameters, GL load rules, and historical response samples from similar populations within a GI dietary database. It represents the expected standard blood glucose response under the same dietary scenario. The system compares member A's standard glycemic response representation in a medium GL scenario (breakfast) with the corresponding preset glycemic baseline representation. If member A's postprandial peak increase and area of ​​fluctuation are significantly higher than the baseline, positive glycemic response deviation information is generated, indicating that the member is more sensitive to carbohydrate sources in this scenario. If the postprandial peak increase is lower than the baseline and falls back quickly, negative glycemic response deviation information is generated, indicating lower glycemic sensitivity in this scenario.

[0051] During the deviation information extraction process, the system also corrects the deviation amplitude by combining sample credibility and interference status. For example, if member A walks briskly for 30 minutes after a meal and the actual blood glucose peak is lower than the baseline, the system will not simply determine that the member is insensitive to the staple food at breakfast, but will reduce the negative deviation based on the exercise status; if member A has a record of staying up late and having a snack after dinner, the system will mark the high blood glucose deviation in that sample as an anomaly and reduce its weight in model training. After correction, the blood glucose response deviation information can be expressed in a structured manner according to dietary scenario, food category, meal time, and sample credibility.

[0052] Next, the system integrates blood glucose response deviation information with historical response features over time to form individual member response training samples. Historical response features include recent trends in similar meal patterns, long-term average glycemic fluctuations, food category sensitivity, meal time sensitivity, feedback stability, and historical model confidence. The system integrates historical response features with time decay weights, allowing recent changes in dietary habits to more quickly impact model training while preserving long-term stable response patterns. For example, if member A recently replaced regular bread with low-GI whole-wheat bread for breakfast over the past two weeks, resulting in a decrease in post-meal peak glycemic index but still above the baseline for similar groups, the system retains both the improvement trend after the replacement and the member's continued sensitivity to bread in the training samples.

[0053] When training the personalized glycemic index (GI) model, the system inputs individual member response training samples into the model. The model can employ a structure combining temporal feature encoding and individual embedding parameters to output the personalized GI coefficient for each member under different dietary scenarios. The personalized GI coefficient can be used to characterize a member's sensitivity to GI in scenarios such as breakfast staple foods, lunch complex carbohydrates, snack dairy products, and cereal products. For members with insufficient historical samples, the system uses parameters from a similar member group as initial values ​​and gradually increases the weight of individual samples as subsequent samples accumulate. For members with sufficient samples and stable feedback, the system directly updates the personalized GI coefficient based on their historical responses.

[0054] Through the above implementation methods, the system can transform discrete blood glucose feedback and meal load into trainable individual response samples, and distinguish the influence of standard GI / GL on individual member glycemic differences in the model. This approach improves the stability and interpretability of personalized glycemic index, enabling subsequent adjustments to product glycemic control weights to better reflect members' actual dietary responses, thereby enhancing the accuracy of glycemic control member profiles and low-GI product recommendations.

[0055] In some embodiments, the blood sugar control profile generation module is specifically used for: Obtain the corrected product sugar control adaptation weight, member dietary habits, historical consumption behavior and marketing feedback data, and collect the data according to member identification to generate member profile input data; Perform time-series consistency verification and source credibility assessment on the member profile input data, and determine the feature weights corresponding to various profile features; Based on the feature weights, the dietary load status, product preference status and marketing response status of members are integrated to generate a blood sugar control member profile; The dynamic credibility of the blood sugar control member profile is updated based on data update time, feedback completeness, and feature stability.

[0056] Specifically, during the blood sugar control member profile generation phase, the system uses the member identifier as the primary index and retrieves corrected product blood sugar control adaptation weights, member dietary habits, historical consumption behavior, and marketing feedback data from the glycemic response modeling module, member transaction system, dietary record terminal, and marketing outreach platform. The corrected product blood sugar control adaptation weights characterize the individual adaptability of a low-GI product to a specific member and specific dietary scenario. Member dietary habits can include the timing of regular meals, staple food preferences, frequency of snacks, stability of intake portions, and prohibited ingredients. Historical consumption behavior can include the frequency of low-GI product purchases, repurchase intervals, tendency to purchase combined products, and price sensitivity. Marketing feedback data can include push notification clicks, add-to-cart, purchase, repurchase, unfollowing, blocking, and usage of benefits. For example, if member A has recorded bread-type staple foods multiple times for breakfast in the past 30 days, the system can generate profile input data related to the breakfast blood sugar control scenario based on their corrected glycemic response to bread-type products, low-GI whole wheat bread purchase records, and breakfast push notification clicks.

[0057] During data collection, the system merges data from different sources according to member identifiers and correlates dietary records, purchase records, and marketing feedback within the same time period. For members who purchase low-GI products but haven't confirmed their consumption, the system configures this consumption behavior as purchase preference data, rather than directly as dietary load status. For members who add items to their cart or make purchases shortly after a push notification, the system establishes a response association between this behavior and the corresponding marketing outreach event. For members who upload post-meal blood glucose feedback but lack corresponding meal records, the system marks this feedback as pending matching data to avoid directly impacting the blood sugar control profile. After the above data collection, the system generates member profile input data and retains the source type, collection time, associated objects, and completeness identifier for each data category.

[0058] Subsequently, the system performs a time-series consistency check on the input data for member profiles. The system checks whether the time of food recording, product purchase, push notification arrival time, and feedback time conform to business logic. For example, the push notification time should be earlier than the click time, the purchase time should have a reasonable interval with the food record or consumption scenario, and the post-meal feedback time should fall within the corresponding meal feedback window. If a piece of data has a time conflict or is incorrectly associated across scenarios, the system downgrades that data from a strong profile feature to a weak reference feature. Taking member A as an example, if they buy low-GI whole-wheat bread in the evening but only confirm consumption for breakfast the next day, the system categorizes the purchase behavior as a product preference state and the breakfast record for the next day as a dietary load state, avoiding misinterpreting the purchase time as the actual consumption time.

[0059] During the source credibility assessment process, the system determines the feature weights corresponding to various profile features based on data source reliability, collection method, field completeness, and historical consistency. Consumption data directly matched by product codes has high source credibility; dietary records manually entered by members require weighting based on historical habits and portion completeness; food objects generated by image recognition require weighting based on recognition confidence; clicks and purchase feedback recorded by the marketing system can directly serve as important inputs for marketing response status. If member A repeatedly confirms their breakfast portion size manually, the system increases their dietary load status weight; if member B only occasionally vaguely fills in low-sugar foods, the system decreases the weight of the relevant profile features.

[0060] During the profile fusion phase, the system integrates members' dietary load status, product preference status, and marketing response status based on feature weights to generate a blood sugar control member profile. Dietary load status reflects the member's daily meal GI composition, GL load level, and glycemic sensitivity scenarios; product preference status reflects the member's purchase and repurchase tendency for low-GI staple foods, blood sugar control snacks, and sugar-free beverages; and marketing response status reflects the member's acceptance of different content, benefit formats, and push frequencies. For example, if member A has a consistently high breakfast GL load but a high click-through rate and stable purchase conversion rate for low-GI whole-wheat bread push notifications, the system can generate a blood sugar control member profile indicating a high demand for breakfast staple food substitutes, a strong preference for low-GI bread products, and moderate sensitivity to benefits.

[0061] During the dynamic credibility update process, the system periodically modifies the blood sugar control member profile based on data update time, feedback completeness, and feature stability. If a certain profile feature has not been updated for a long time, the system reduces its dynamic credibility according to the time decay rule; if a member recently supplements complete dietary records and post-meal blood glucose feedback, the system increases the dynamic credibility of the corresponding dietary load status; if a member's recent purchasing behavior differs significantly from historical preferences, the system identifies profile drift and reduces the weight of old preference labels. For example, if member A changes from bread-based breakfasts to oatmeal-based breakfasts for two consecutive weeks, the system gradually reduces the credibility of the bread preference label and increases the credibility of the oatmeal staple food substitution label.

[0062] Through the above implementation methods, the blood sugar control profile generation module can transform product matching weights, dietary habits, consumption behavior, and marketing feedback into blood sugar control member profiles with time constraints and source credibility. This approach can reduce profile bias caused by static consumption segmentation, enabling member profiles to reflect actual dietary load, product preference status, and marketing response status, providing a more accurate data foundation for subsequent member segmentation decisions and differentiated marketing outreach.

[0063] In some embodiments, the membership tier decision-making module is specifically used for: Obtain profiles of members who are in blood sugar control, and determine the effective profile representations that can participate in hierarchical calculations based on the dynamic credibility corresponding to the profile features; The effective profile representation is input into the multidimensional member segmentation model, and combined with the intensity of blood sugar control demand, low GI substitution potential and marketing response status, the blood sugar control segmentation type of the member is determined. Based on the blood sugar control stratification type, the stratification strategy rules are invoked to determine candidate low-GI products, candidate dietary plans, and candidate membership benefits that are suitable for the blood sugar control needs of members; Based on the reach response status and push fatigue status in the sugar control member profile, the candidate push frequency is configured for the sugar control stratification type.

[0064] Specifically, during the membership segmentation decision-making stage, the system obtains the target member's blood sugar control profile from the blood sugar control profile generation module, and reads the dynamic credibility, update time, and data source corresponding to various features in the profile. The blood sugar control member profile may include profile features such as dietary load status, product preference status, marketing response status, push notification fatigue status, and rights sensitivity status.

[0065] The system first filters the blood sugar control member profiles for validity, excluding profile features with dynamic reliability below a preset threshold, update times exceeding the valid period, or significant temporal conflicts from the stratified calculation. The remaining profile features are then encoded as valid profile representations. For example, member A has continuously recorded breakfast and lunch for the past 7 days, with complete post-meal blood glucose feedback. Their breakfast GL load is high, and their click rate for low-GI staple foods is high, indicating that these features have high dynamic reliability. However, since they purchased sugar-free beverages 3 months ago and have not exhibited related behavior recently, the system downweights or removes this old preference feature.

[0066] Furthermore, after forming an effective profile representation, the system converts dietary load status, product preference status, and marketing response status into computable hierarchical features. Dietary load status is used to represent a member's GL load level at different meal times, glycemic sensitivity scenarios, and blood sugar control stability; product preference status is used to represent a member's acceptance of product categories such as low-GI staple foods, blood sugar-controlled snacks, and sugar-free drinks; and marketing response status is used to represent a member's clicks, purchases, repurchases, and blocking behavior for different content and benefit forms. For example, taking member A as an example, if breakfast meals repeatedly show medium-to-high GL loads, post-meal blood sugar fluctuations are high, and there is a high click and purchase behavior for low-GI whole-wheat bread and oat products, the system encodes this as an effective profile representation with strong breakfast staple food substitution demand, high low-GI substitution potential, and a good marketing response status.

[0067] Furthermore, after receiving effective profile representations, the multi-dimensional member stratification model calculates the intensity of blood sugar control demand, the potential for low-GI substitution, and the marketing response status. The intensity of blood sugar control demand can be determined based on meal GL load levels, personalized glycemic index, and blood glucose fluctuation deviation information; the potential for low-GI substitution can be determined based on the current proportion of high-GI or high-GL intake, the coverage of alternative products, and member product preferences; the marketing response status can be determined based on recent outreach conversions, benefit usage, repurchase intervals, and blocking status.

[0068] The system then combines the dynamic credibility corresponding to the effective profile representation, and performs weighted fusion of multiple stratified evaluation results to generate a member stratification judgment representation. If a member has complete blood glucose feedback samples and stable dietary records, the weight related to blood glucose control needs is relatively high; if a member has less blood glucose feedback but complete consumption data, the weight related to product preferences and marketing responses is relatively higher.

[0069] The system matches member segmentation characteristics with preset segmentation boundaries to determine a member's sugar control segmentation type. Segmentation types can be configured based on a combination of sugar control needs and marketing responses, such as members with high sugar control needs and high substitution potential, members focused on sugar control awareness development, members maintaining stable repeat purchases, members sensitive to benefits and conversions, and members protected against push notification fatigue. For member A, if the model determines that they are sensitive to glycemic spikes at breakfast and have a high click-through and purchase tendency towards low-GI staple foods, the system will classify member A as a member with high sugar control needs and high substitution potential. For member B, if they frequently purchase low-GI products but recently experience a decrease in push notification clicks and an increase in blocking rates, the system can classify them as a member with stable repeat purchases but experiencing push notification fatigue to avoid over-reach.

[0070] Furthermore, after determining the blood sugar control stratification type, the system invokes the stratification strategy rules to configure candidate low-GI products, candidate dietary plans, and candidate membership benefits that match the member's blood sugar control needs. For members with high blood sugar control needs and high substitution potential, the system prioritizes staple food substitutes with lower personalized glycemic coefficients and higher product blood sugar control suitability weights, and configures breakfast or lunch substitute dietary plans; for members who are cultivating blood sugar control awareness, the system prioritizes configuring basic low-GI products and content-based blood sugar control plans; for members who are sensitive to conversion due to benefits, the system prioritizes configuring bundled discounts, repurchase coupons, and membership points benefits. The candidate content is not a fixed manual list, but is dynamically generated based on the blood sugar control stratification type, product blood sugar control suitability weights, inventory status, and member preferences.

[0071] During the configuration of candidate push frequency, the system further reads the reach response status and push fatigue status in the sugar control member profile. If the member has a high click rate for sugar control content recently, stable purchase conversion, and no blocking behavior, the system can configure a higher push frequency; if the member has not clicked for a continuous period of time recently or has blocked behavior, the system reduces the push frequency and prioritizes low-interference content or benefit reminders.

[0072] Through the above implementation methods, the member segmentation decision module can transform the blood sugar control member profile into actionable segmentation results and candidate marketing resources, ensuring that low-GI products, dietary plans, member benefits, and push frequency are consistent with the actual blood sugar control needs of members, thereby improving the accuracy of member segmentation and the suitability of marketing reach.

[0073] In some embodiments, effective profile representations are input into a multidimensional member stratification model, and combined with the intensity of blood sugar control demand, low-GI substitution potential, and marketing response status, the member's blood sugar control stratification type is determined, including: Feature encoding is performed on the effective profile representation to generate a hierarchical member input representation that includes features of blood sugar control needs, low-GI substitution features, and marketing response features; The member stratification input is input into the member multidimensional stratification model, and the stratification evaluation values ​​corresponding to the intensity of sugar control demand, low GI substitution potential and marketing response status are calculated respectively. Based on dynamic credibility, the evaluation values ​​of each level are weighted and integrated to generate a member level determination representation; The member stratification criteria are matched with the preset stratification boundaries to determine the blood sugar control stratification type to which the member belongs.

[0074] Specifically, when the member stratification decision module performs stratification calculations, the system first reads the effective profile representations of the target members and encodes these representations according to stratification dimensions. Effective profile representations can come from profile features in the blood sugar control member profile that meet dynamic credibility conditions, including recent meal load status, personalized glycemic index, low-GI product preference, historical repurchase behavior, benefit usage behavior, and outreach response behavior. The system does not directly use the original labels for stratification; instead, it converts the profile features into a unified member stratification input representation. For example, if member A has consistently high breakfast GL load for the past 14 days, post-meal blood glucose peaks are higher than the benchmark for similar members, and they have repeatedly clicked on low-GI staple food recommendations and completed purchases, the system encodes the relevant data as blood sugar control demand features, low-GI substitution features, and marketing response features.

[0075] During feature encoding, the blood sugar control demand feature is used to characterize a member's current blood sugar control intervention needs. The system can calculate this feature based on meal GL load levels, blood glucose response deviation information, glycemic sensitivity scenarios, and the duration of dietary load. If member A experiences multiple medium-to-high GL loads at breakfast, and their personalized glycemic index indicates sensitivity to bread-type carbohydrates, the system increases the blood sugar control demand feature value for breakfast staple foods. The low-GI substitution feature characterizes the feasibility of a member replacing their existing high-load diet with low-GI products or dietary plans. The system can calculate this feature based on the current proportion of high-GI food intake, the availability of low-GI products, the product's blood sugar control suitability weight, and the member's purchasing preferences. If member A frequently buys regular bread but also clicks on and adds low-GI whole-wheat bread to their cart, the system determines that they have high low-GI substitution potential.

[0076] Marketing response features are used to characterize a member's acceptance of marketing outreach and conversion stability. The system can encode these features based on push click-through rate, add-to-cart rate, purchase conversion rate, repurchase interval, benefit usage rate, unfollowing behavior, and blocking behavior. For member A, if they click on breakfast sugar control content multiple times within the past 30 days and use repurchase coupons to buy low-GI staple foods, the system generates a high marketing response feature. For member B, if they do not click on push notifications multiple times consecutively and block the push notification, the system generates a low marketing response feature and identifies push fatigue in subsequent stratification. Through the above encoding, the system forms a member stratification input representation, allowing data from different sources and with different dimensions to be included in the same stratification model calculation.

[0077] Subsequently, the system inputs the member stratification representation into the member multidimensional stratification model, calculating the stratification evaluation values ​​corresponding to the intensity of sugar control demand, low-GI substitution potential, and marketing response status. The member multidimensional stratification model can employ a structure combining rule constraints and machine learning scoring. First, it filters out obviously unsuitable stratification results based on business rules, and then uses the stratification scoring model to calculate the evaluation values ​​for each dimension. For example, the system calculates that member A has a high intensity of sugar control demand, a high low-GI substitution potential, and a medium-high marketing response status; for member B, it calculates a medium intensity of sugar control demand, a low low-GI substitution potential, and a low marketing response status. The stratification evaluation values ​​not only reflect individual consumption capacity but also reflect the member's actual dietary load and the likelihood of conversion to sugar control products.

[0078] During the stratified evaluation value fusion stage, the system weights and fuses the intensity of blood sugar control demand, low-GI substitution potential, and marketing response status based on the dynamic credibility of each profile feature in the blood sugar control member profile. If a member has complete post-meal blood glucose feedback and stable recent dietary records, the intensity of blood sugar control demand has a higher weight; if a member has insufficient blood glucose data but complete purchase and outreach feedback, the low-GI substitution potential and marketing response status have higher weights; if a certain profile feature has not been updated for a long time or has low source credibility, the system reduces the weight of the relevant evaluation value. For example, if member A has relatively complete breakfast records and blood glucose feedback, the system increases the weight of the intensity of blood sugar control demand during fusion, making the stratification results more biased towards dietary intervention needs; if member C lacks blood glucose feedback but has stable repurchase of low-GI products, the system increases the weight of product preference and marketing response during fusion.

[0079] The system generates a member segmentation determination representation based on the weighted fusion result and matches this representation with preset segmentation boundaries. These preset boundaries can be configured based on a combination of factors including sugar control demand intensity, low-GI substitution potential, and marketing response status, and can also be periodically adjusted based on historical marketing performance. For example, when the member segmentation determination representation shows high sugar control demand intensity and high low-GI substitution potential, and the marketing response status is not lower than a preset level, the system classifies the member into the high sugar control demand, high substitution potential type; when the sugar control demand intensity is relatively high but the marketing response status is low, the system classifies the member into the sugar control awareness cultivation type; and when a member has stable low-GI product purchase behavior but recent push notification responses have declined, the system classifies the member into the repeat purchase maintenance and push notification protection type.

[0080] Through the above implementation methods, the multi-dimensional membership stratification model can form a calculable, verifiable, and updatable stratification process based on effective profile representation, avoiding coarse stratification based solely on age, spending amount, or membership level. This approach can improve the accuracy of blood sugar control membership stratification, enabling blood sugar control needs, low-GI substitution potential, and marketing response status to work synergistically within the same decision-making framework. This provides a more precise stratification basis for subsequent selection of candidate low-GI products, dietary plans, membership benefits, and push notification frequency configurations.

[0081] In some embodiments, the marketing closed-loop optimization module is specifically used for: Obtain candidate content and push constraints, and combine them with the member's current sugar control stratification type, product sugar control suitability weight, and historical outreach response to generate marketing decision inputs; The marketing decision is input into the marketing push decision model, and the target content, timing and intensity of the push are determined under the condition of meeting the push constraints, so as to generate a differentiated marketing reach plan. Collect behavioral and blood glucose feedback after the implementation of differentiated marketing outreach programs, and establish feedback correlations between behavioral and blood glucose feedback and the corresponding target outreach content; Based on feedback correlation, the reach bias and blood sugar control adaptation bias are determined, and the personalized glycemic index, blood sugar control member profile and marketing decision parameters are updated according to the reach bias and blood sugar control adaptation bias.

[0082] Specifically, in the marketing closed-loop optimization phase, the system first obtains candidate low-GI products, candidate dietary plans, candidate membership benefits, and candidate push frequencies corresponding to the target members from the membership segmentation decision module, and reads the push constraints configured on the marketing operations side. Push constraints may include the daily reach limit, the interval for repeated reach of similar products, benefit budget, inventory status, member do-not-disturb periods, and sensitive content blocking rules.

[0083] The system combines a member's current sugar control stratification type, the sugar control suitability weight of products, and historical outreach responses to generate marketing decision inputs. For example, if member A is categorized as having high sugar control needs and high substitution potential, low-GI oat products have a high sugar control suitability weight in the breakfast scenario. In the past 7 days, member A has consistently clicked on sugar control content related to breakfast and has not exhibited any blocking behavior. The system integrates this member's breakfast substitution needs, product suitability weight, benefit sensitivity, and reach frequency into marketing decision inputs.

[0084] When generating marketing decision inputs, the system also performs feasibility filtering on candidate content. If a candidate low-GI product is out of stock or outside the member's delivery range, the system removes it from the list. If a candidate benefit exceeds the current budget or the member has already claimed a similar benefit, the system lowers the priority of the corresponding benefit. If a member has not responded to the same product recently, the system lowers the reach weight of that product and increases the candidate priority of dietary content or alternative solutions. Therefore, the marketing decision inputs not only reflect the member's blood sugar control needs but also the product's availability and the feasibility of reaching them.

[0085] Subsequently, the system inputs marketing decisions into the marketing push decision model, determining the target content, timing, and intensity of reach under the constraints of the push. The marketing push decision model can employ a combination of constraint optimization and contextual exploration strategies to comprehensively score the sugar control adaptation benefits, conversion probability, disturbance risk, and feedback value of different candidate content. Target content can include low-GI product recommendations, breakfast alternative meal plans, combo meals, repurchase coupons, or sugar control check-in incentives; timing can be determined based on the member's historical opening time and meal time; intensity can be determined based on push frequency, incentive strength, and content display priority. For example, for member A, the system chooses to push a low-GI oatmeal breakfast combo one hour before their usual breakfast, along with a small coupon and sugar control check-in reminder, instead of pushing ordinary promotional information at night.

[0086] Furthermore, after the differentiated marketing outreach plan is implemented, the system collects behavioral and blood glucose feedback from the marketing platform, transaction system, dietary record terminal, and blood glucose feedback interface. Behavioral feedback can include exposure, clicks, browsing duration, adding to cart, purchasing, repeat purchase, claiming benefits, unfollowing, and blocking; blood glucose feedback can include changes in post-meal blood glucose levels after a member uses the recommended product or follows the recommended dietary plan. The system establishes feedback associations between behavioral and blood glucose feedback and the corresponding target outreach content, forming outreach feedback records. For example, if member A clicks on a low-GI oatmeal breakfast combo and confirms consumption in their breakfast record the next day, the system will associate the purchase behavior, consumption record, and post-meal blood glucose feedback with the breakfast outreach content for that breakfast.

[0087] During the deviation analysis process, the system identifies reach deviation and glycemic control adaptation deviation based on feedback correlation. Reach deviation reflects the difference between the expected marketing response and the actual behavioral feedback. For example, if the model predicts that member A will purchase a combo meal, but they only browse and do not buy, a conversion deviation occurs. Glycemic control adaptation deviation reflects the difference between the expected glycemic control effect of the recommended content and the actual blood sugar feedback. For example, a low-GI oatmeal combo is expected to reduce breakfast fluctuations, but member A's post-meal peak is still significantly high, resulting in glycemic control adaptation deviation. The system further distinguishes the sources of deviation, determining that the deviation stems from insufficient product adaptation, insufficient incentives, poor timing of the push notification, recent dietary changes in the member, or abnormal blood sugar feedback.

[0088] Based on reach deviations and blood sugar control adaptation deviations, the system updates personalized glycemic index, blood sugar control member profiles, and marketing decision parameters. If a product is purchased multiple times but post-meal blood sugar feedback is unsatisfactory, the system lowers the product's blood sugar control adaptation weight in the corresponding member's dietary scenario and corrects the personalized glycemic index. If a member has high click-through rates for content-based outreach but low purchase conversion rates, the system updates the marketing response status and adjusts subsequent benefit formats. If a member continuously blocks similar push notifications, the system increases push fatigue status and reduces the frequency of candidate push notifications. Through these processes, the system forms a closed-loop mechanism from candidate content decision-making, differentiated outreach, feedback correlation to model parameter updates, which can continuously reduce ineffective outreach, improve the accuracy of blood sugar control product matching, and enhance the adaptive optimization capability of member marketing strategies.

[0089] The above embodiments have described in detail the specific components and functions of the intelligent marketing system for blood sugar control members based on GI dietary data of this application. The implementation process of the intelligent marketing method for blood sugar control members based on GI dietary data of this application will be described in detail below with reference to specific embodiments. Figure 2 This is a flowchart illustrating the intelligent marketing method for blood sugar control members based on GI dietary data provided in this application embodiment, as follows: Figure 2 As shown, the method may specifically include the following steps: S201, Obtain member dietary behavior data, and call the GI dietary database to match food objects and commodity objects in the member dietary behavior data to generate a meal dietary load characterization; S202, the meal load characterization and the member's post-meal blood glucose feedback are correlated according to the meal time, the personalized glycemic index of the member in the meal scenario is calculated, and the personalized glycemic index is used to correct the product blood sugar control adaptation weight. S203 integrates the revised product sugar control adaptation weights, member behavior data, and marketing feedback data to generate a sugar control member profile, and determines the sugar control stratification type of the member based on the sugar control member profile; S204 generates differentiated marketing outreach plans based on blood sugar control stratification types, and updates personalized glycemic index, blood sugar control member profiles, and marketing decision parameters based on behavioral and blood sugar feedback after outreach.

[0090] Specifically, in step S201, the system first obtains member dietary behavior data through the member-end mini-program, the cashier order system, the product barcode scanning interface, the food image recognition interface, and the manual recording interface. Member dietary behavior data may include member identification, meal time, meal type, food name, product code, quantity purchased, actual consumption ratio, and portion size.

[0091] The system uses member IDs as data indexes and meal times as event boundaries. It processes food records, product consumption records, and portion sizes within the same time window into events, generating meal events. For example, if a member uploads a picture of their breakfast and purchases low-GI whole-wheat bread and sugar-free yogurt, the system merges the food objects identified in the image with the product objects in the order, forming a single breakfast meal event.

[0092] Subsequently, the system invokes the GI dietary database to match food objects and product objects within the meal event. The GI dietary database pre-associates and stores food GI parameters, GL load rules, and product sugar control adaptation relationships. For standard food objects, the system determines the corresponding food GI parameters based on name normalization, category identification, and component characterization; for packaged product objects, the system determines the corresponding product sugar control adaptation relationship based on product code, ingredient composition, and nutritional attributes; for combination meals, the system first breaks down the main food components and then performs GI and GL matching separately.

[0093] Furthermore, after matching is completed, the system performs weighted aggregation of dietary loads within the same meal based on the GI parameters, available carbohydrate content, intake amount, and matching confidence of each food or commodity, generating a meal dietary load characterization. This characterization may include meal GL load, major sources of glycemic index, meal GI composition, commodity glycemic control adaptation results, and overall matching confidence.

[0094] In step S202, the system correlates the meal load representation with the member's post-meal blood glucose feedback in a time series. Post-meal blood glucose feedback can come from a blood glucose meter, continuous glucose monitoring device, health management interface, or be manually entered by the member. The system sets an effective feedback window based on the meal time and retrieves the corresponding member's post-meal blood glucose data within the effective feedback window, extracting response features such as pre-meal blood glucose, post-meal peak, 2-hour post-meal blood glucose, blood glucose rise rate, fall time, and area of ​​fluctuation. For data exceeding the effective feedback window, the system does not directly delete it but configures a lower confidence level to avoid excessive impact on model training due to delayed uploads or missed meal recordings.

[0095] After obtaining glycemic response samples, the system further extracts information on members' blood glucose response deviations in different dietary scenarios. The system can generate a preset glycemic baseline representation based on GI parameters and GL loading rules in the GI dietary database, and compare the member's actual postprandial blood glucose response with the preset glycemic baseline representation to determine whether the member's glycemic response is high or low in different dietary scenarios such as breakfast staples, lunch complex carbohydrates, and snacks. The system then trains a personalized glycemic model using historical response features to obtain a personalized glycemic coefficient that characterizes the member's sensitivity to glycemic responses in different dietary scenarios. If a member's postprandial peak blood glucose level remains significantly high after repeatedly consuming low-GI whole-wheat bread, the system can determine that they are sensitive to glycemic responses in this type of staple food scenario and adjust the glycemic control adaptation weight of the corresponding product in that member's breakfast scenario using the personalized glycemic coefficient.

[0096] In step S203, the system integrates the corrected product sugar control adaptation weights, member behavior data, and marketing feedback data to generate a sugar control member profile. Member behavior data may include dietary habits, frequency of low-GI product purchases, repurchase intervals, preferred product categories, stability of intake portions, and usage of benefits; marketing feedback data may include push notification clicks, views, add-to-cart actions, purchases, repurchases, blocking, and unfollowing responses. The system aggregates the above data according to member identifiers and performs time-series consistency checks and source credibility assessments to determine the feature weights of different profile characteristics.

[0097] During the profile generation process, the system integrates a member's dietary load status, product preference status, and marketing response status based on feature weights to form a blood sugar control member profile. For example, if a member has a consistently high GL load for breakfast, a high click-through rate for low-GI oat products, and stable purchase conversion, the system can generate a blood sugar control member profile indicating a strong demand for breakfast staple food alternatives, a high preference for low-GI oat products, and a good marketing response status. Simultaneously, the system updates the dynamic credibility of the blood sugar control member profile based on data update time, feedback completeness, and feature stability. If a certain type of data has not been updated for a long time, the system reduces the credibility of the corresponding profile features; if recent dietary records and blood sugar feedback are complete, the system increases the credibility of the corresponding dietary load features.

[0098] Subsequently, the system determines the member's blood sugar control stratification type based on the member's profile. The system first filters effective profile representations based on dynamic credibility, then inputs these representations into a multi-dimensional member stratification model to calculate the intensity of blood sugar control demand, low-GI substitution potential, and marketing response status. The system weights and merges the evaluation values ​​for each stratification and matches them with preset stratification boundaries to determine whether the member belongs to a stratification type such as high blood sugar control demand and high substitution potential, blood sugar control awareness cultivation, stable repurchase maintenance, or push fatigue protection. This stratification result is used for subsequent configuration of candidate low-GI products, candidate dietary plans, candidate member benefits, and candidate push notification frequencies.

[0099] In step S204, the system generates differentiated marketing outreach plans based on the member's sugar control stratification type. The system generates marketing decision inputs based on the member's current sugar control stratification type, product sugar control suitability weight, historical outreach responses, and push constraints. Push constraints may include a daily outreach limit, outreach intervals for similar products, inventory status, benefit budget, and member's do-not-disturb periods. Under the condition of satisfying push constraints, the marketing push decision model determines the target outreach content, timing, and intensity. For example, for members with high demand for breakfast staple food replacements and a good response to oat products, the system can push low-GI oat combinations and mild benefits before breakfast; for members experiencing fatigue, the system reduces the push frequency and prioritizes content-based sugar control suggestions.

[0100] After the differentiated marketing outreach plan is implemented, the system continues to collect behavioral and blood glucose feedback after the outreach, and establishes a feedback correlation between the feedback data and the target outreach content. If a member clicks on and purchases a recommended product, and their post-meal blood glucose fluctuations subsequently decrease, the system increases the product's blood glucose control suitability weight in the corresponding scenario and enhances the credibility of relevant profile features; if the member's post-meal blood glucose fluctuations remain high after purchase, the system identifies blood glucose control suitability deviations and corrects the personalized glycemic index; if a member does not click on or blocks similar push notifications consecutively, the system identifies outreach deviations and adjusts marketing decision parameters and candidate push frequency.

[0101] Through the above process, the system forms a closed-loop process of dietary behavior collection, GI data analysis, personalized glycemic index modeling, member segmentation, differentiated marketing outreach, and feedback-based iterative optimization, thereby improving the accuracy of blood sugar control member profiling, the accuracy of low-GI product matching, and the adaptive optimization capability of marketing strategies. It should be understood that the sequence number of each step in the above method embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0102] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although the technical solutions of this application have been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A smart marketing system for blood sugar control members based on GI dietary data, characterized in that, include: The GI dietary data management module is used to establish a GI dietary database that includes food GI parameters, GL load rules, and product sugar control adaptation relationships, and to match food objects and product objects in member dietary behavior data to generate meal dietary load representations. The glycemic response modeling module is used to correlate the meal dietary load characterization with the member's post-meal blood glucose feedback according to the meal time, calculate the member's personalized glycemic coefficient under different dietary scenarios based on the correlation results, and use the personalized glycemic coefficient to correct the product's blood sugar control adaptation weight. The blood sugar control profile generation module is used to integrate the corrected product blood sugar control adaptation weights, member dietary habits, historical consumption behavior and marketing feedback data to generate a blood sugar control member profile with dynamic credibility. The member stratification decision module is used to determine the stratification type of a member based on the blood sugar control member profile, and to configure candidate low-GI products, candidate dietary plans, candidate member benefits and candidate push frequency for the blood sugar control stratification type; The marketing closed-loop optimization module is used to generate differentiated marketing outreach plans based on candidate content and push constraints, and update the personalized glycemic index, the blood sugar control member profile, and marketing decision parameters based on behavioral feedback and blood sugar feedback after outreach.

2. The system according to claim 1, characterized in that, The establishment of a GI dietary database, including food GI parameters, GL load rules, and product glycemic control adaptation relationships, includes: Acquire food nutrition data and product nutritional attribute data, and standardize the food nutrition data and product nutritional attribute data to generate unified dietary data; Based on the food component characterization in the unified dietary data, the food GI parameters corresponding to the food objects are determined, and GL load rules are constructed according to the food GI parameters, available carbohydrate characterization, and intake portion characterization. Based on the semantic matching relationship between the product ingredient representation and the food object, a product sugar control adaptation relationship is generated; The food GI parameters, the GL load rules, and the product sugar control adaptation relationships are associated and stored according to data reliability to form a dynamically updatable GI dietary database.

3. The system according to claim 2, characterized in that, The process of matching food and product objects in member dietary behavior data to generate a meal load representation includes: Acquire members' dietary behavior data, and process the dietary records, product consumption records, and intake representations into events according to member identification and meal time to generate meal dietary events; The GI dietary database is invoked to perform semantic matching on food objects and product objects in the meal dietary events, and to determine the corresponding food GI parameters, GL load rules and product sugar control adaptation relationships. The dietary load within the same meal is weighted and aggregated based on the matching results, and a meal dietary load characterization is generated by combining the matching confidence.

4. The system according to claim 1, characterized in that, The glycemic response modeling module is specifically used for: Based on the meal time, a temporal correspondence is established between the dietary load characterization of the meal and the post-meal blood glucose feedback of the members, and the data that exceeds the effective feedback window is weighted down in terms of credibility to generate blood glucose response samples. Based on the glycemic response samples, the blood glucose response deviation information of members in dietary scenarios is extracted, and a personalized glycemic model is trained by combining historical response features to obtain a personalized glycemic coefficient. The personalized glycemic index is used to individually modify the product glycemic control fit relationship in the GI dietary database, generating modified product glycemic control fit weights.

5. The system according to claim 4, characterized in that, The process involves extracting members' blood glucose response deviation information under dietary scenarios based on the glycemic response samples, and training a personalized glycemic model by combining historical response features to obtain a personalized glycemic coefficient, including: The glycemic response samples are subjected to scenario normalization processing to generate a standard glycemic response characterization corresponding to the member's dietary scenario; Based on the difference between the standard glycemic response characterization and the preset glycemic baseline characterization, the blood glucose response deviation information of members in the corresponding dietary scenarios is extracted; The blood glucose response deviation information is fused with historical response features over time to form individual member response training samples. A personalized glycemic index model is trained based on the individual member response training samples, and the output is a personalized glycemic index coefficient that characterizes the member's sensitivity to glycemic index in different dietary scenarios.

6. The system according to claim 1, characterized in that, The blood sugar control profile generation module is specifically used for: Obtain the corrected product sugar control adaptation weight, member dietary habits, historical consumption behavior and marketing feedback data, and collect the data according to member identification to generate member profile input data; The time-series consistency verification and source credibility assessment are performed on the member profile input data to determine the feature weights corresponding to various profile features; Based on the aforementioned feature weights, the member's dietary load status, product preference status, and marketing response status are integrated to generate a blood sugar control member profile; The dynamic credibility of the blood sugar control member profile is updated based on data update time, feedback completeness, and feature stability.

7. The system according to claim 1, characterized in that, The membership tier decision-making module is specifically used for: Obtain the profile of the blood sugar control member, and determine the effective profile representation that can participate in the hierarchical calculation based on the dynamic credibility corresponding to the profile features; The effective profile representation is input into the member multidimensional stratification model, and combined with the intensity of blood sugar control demand, low GI substitution potential and marketing response status, the blood sugar control stratification type of the member is determined. Based on the blood sugar control stratification type, the stratification strategy rules are invoked to determine candidate low-GI products, candidate dietary plans, and candidate membership benefits that are suitable for the member's blood sugar control needs; Based on the reach response status and push fatigue status in the sugar control member profile, candidate push frequencies are configured for the sugar control stratification type.

8. The system according to claim 7, characterized in that, The effective profile representation is input into the member multidimensional stratification model, and combined with the intensity of blood sugar control demand, low-GI substitution potential, and marketing response status, the blood sugar control stratification type of the member is determined, including: The effective profile representation is feature-encoded to generate a member-level input representation that includes features of sugar control needs, low-GI substitution features, and marketing response features; The member stratification input representation is input into the member multidimensional stratification model, and the stratification evaluation values ​​corresponding to the intensity of sugar control demand, low GI substitution potential and marketing response status are calculated respectively. Based on the dynamic credibility, the evaluation values ​​of each stratification are weighted and fused to generate a member stratification determination representation. The member stratification determination characteristics are matched with preset stratification boundaries to determine the blood sugar control stratification type to which the member belongs.

9. The system according to claim 1, characterized in that, The marketing closed-loop optimization module is specifically used for: Obtain candidate content and push constraints, and combine them with the member's current sugar control stratification type, product sugar control suitability weight, and historical outreach response to generate marketing decision inputs; The marketing decisions are input into the marketing push decision model, and the target content, timing, and intensity of the push are determined under the conditions of satisfying the push constraints, thereby generating a differentiated marketing outreach plan. Collect behavioral feedback and blood glucose feedback after the implementation of the differentiated marketing outreach plan, and establish feedback associations between the behavioral feedback and the blood glucose feedback and the corresponding target outreach content; Based on the feedback correlation, the reach deviation and the blood sugar control adaptation deviation are determined, and the personalized blood sugar increase coefficient, the blood sugar control member profile, and the marketing decision parameters are updated according to the reach deviation and the blood sugar control adaptation deviation.

10. A smart marketing method for blood sugar control based on GI dietary data, using a system as described in any one of claims 1 to 9, characterized in that, include: Acquire member dietary behavior data, and call the GI dietary database to match food objects and product objects in the member dietary behavior data to generate a meal dietary load characterization; The meal load characterization and the member's post-meal blood glucose feedback are correlated according to the meal time to calculate the member's personalized glycemic index in the dietary scenario, and the personalized glycemic index is used to correct the product blood sugar control adaptation weight. By integrating the corrected product sugar control adaptation weights, member behavior data, and marketing feedback data, a sugar control member profile is generated, and the sugar control stratification type to which the member belongs is determined based on the sugar control member profile; Based on the blood sugar control stratification type, a differentiated marketing outreach plan is generated, and the personalized glycemic index, the blood sugar control member profile, and the marketing decision parameters are updated according to the behavioral feedback and blood sugar feedback after the outreach.

Images