Personalized nutrition recommendation method and system based on environmental dynamic adaptation

Abstract

The application discloses a kind of personalized nutrition recommendation method and system based on environmental dynamic adaptation, constructs user-environment joint state vector and generates personalized environmental-physiological response baseline based on user multidimensional data;Based on the baseline, the high uncertainty dimension of user nutrition response is identified and comparative nutrition intervention hypothesis is generated, which is further converted into executable diet template, and the adaptive adjustment rule is integrated with environmental dynamic data to form an environmental adaptive nutrition experiment scheme;During the execution of the scheme, substitutes are recommended by a dynamic food replacement engine and environmental compensatory fine tuning is triggered based on real-time environmental monitoring data;Collect user physiological result data and user subjective feedback data during the execution of the scheme and control environmental confounding factors through causal effect estimation module, analyze the influence of comparative nutrition intervention hypothesis to generate personalized nutrition decision rule.The application realizes the accurate causal inference and decision of personalized nutrition intervention effect in dynamic environment.

Description

Technical Field

[0001] This invention relates to the field of health management technology, specifically to a personalized nutrition recommendation method and system based on dynamic environmental adaptation. Background Technology

[0002] Personalized nutrition recommendation technology aims to provide customized dietary guidance based on individual user differences. With the widespread adoption of wearable devices, existing technologies can continuously monitor users' physiological indicators (such as blood sugar and heart rate) and use machine learning models to analyze this time-series data, dynamically generating dietary suggestions to improve the timeliness of recommendations. However, these methods typically focus on fitting a single recommendation strategy from historical data, essentially relying on correlation-based prediction. When the user's response patterns to specific nutritional factors are unclear, the system lacks a mechanism for proactively designing controlled experiments to explore and verify causal relationships. Furthermore, a user's physiological state is simultaneously influenced by multiple factors, including dietary interventions and external dynamic environments (such as weather and seasonal changes). Existing solutions often struggle to effectively isolate the confounding effects of these environmental factors when evaluating recommendation effectiveness, leading to inaccurate attribution of the true effects of nutritional interventions and impacting the credibility and practicality of the resulting personalized nutrition rules. Summary of the Invention

[0003] In view of the above problems, the present invention provides a personalized nutrition recommendation method and system based on dynamic environmental adaptation. By actively generating comparative nutrition experiments and controlling environmental confounding factors to conduct causal effect analysis, the problem of inaccurate causal attribution in personalized nutrition decision-making is solved.

[0004] To achieve the above objectives, in a first aspect, this application provides a personalized nutrition recommendation method based on dynamic environmental adaptation, comprising:

[0005] Acquire multidimensional user data and dynamic environmental data. Multidimensional user data includes static user attribute data, dynamic time-series physiological data, and subjective user feedback data. Dynamic environmental data includes real-time environmental monitoring data and supply chain status data.

[0006] Based on multidimensional user data, a user-environment joint state vector is constructed, and a time series analysis model is used to identify the correlation between user physiological patterns and environmental fluctuations, generating a personalized environmental-physiological response baseline.

[0007] Based on the personalized environmental-physiological response baseline and user multidimensional data, at least one dimension of high uncertainty in the user's nutritional response is identified through the uncertainty assessment module;

[0008] Generate at least two comparative nutritional intervention hypotheses based on the high uncertainty dimension, and transform the comparative nutritional intervention hypotheses into an executable dietary template containing controllable variables;

[0009] By integrating dynamic environmental data with executable dietary templates and adding environmental adaptive adjustment rules to the executable dietary templates, an environmentally adapted nutritional experimental protocol is formed.

[0010] During the execution of the environmentally adapted nutrition experiment, based on the real-time supply chain status data, the dynamic ingredient replacement engine recommends food substitutes that meet the nutritional and availability requirements for the environmentally adapted nutrition experiment. Based on the real-time environmental monitoring data, it triggers environmental compensation fine-tuning of the execution parameters of the environmentally adapted nutrition experiment.

[0011] We collect user physiological outcome data and user subjective feedback data during the implementation of an environment-adaptive nutrition experimental protocol, and combine them with environmental dynamic data acquired concurrently. By controlling environmental confounding factors through a causal effect estimation module, we analyze the impact of comparative nutrition intervention hypotheses on user physiological outcomes and generate personalized nutrition decision-making rules.

[0012] Furthermore, during the execution of the environment-adaptive nutrition experiment, based on real-time supply chain status data, a dynamic ingredient substitution engine recommends food substitutes that meet nutritional and availability requirements. Based on real-time environmental monitoring data, it also triggers environmentally compensated fine-tuning of the experiment's execution parameters, including:

[0013] When the preset ingredients in the environmentally adapted nutrition experiment plan are unavailable, the set of locally available alternative ingredients is queried based on the real-time supply chain status data.

[0014] Calculate the nutritional similarity between each alternative food in the alternative food set and the preset food, and evaluate the environmental adaptability of each alternative food in combination with the environmental context indicated by the current real-time environmental monitoring data.

[0015] Based on nutritional similarity and environmental adaptability, target alternative foods are identified from the set of alternative foods, and the recommended intake of target alternative foods in environmentally adapted nutrition experimental protocols is automatically adjusted.

[0016] Continuously monitor real-time environmental monitoring data. When the monitored environmental parameter values ​​exceed the preset environmental parameter threshold range corresponding to the personalized environment-physiological response baseline, generate an environmental compensatory adjustment instruction for at least one execution parameter in the environmentally adapted nutrition experimental protocol. The execution parameters include at least one of the following: water intake, electrolyte supplementation, and food cooking method.

[0017] The system pushes environmental compensatory adjustment instructions to users. These instructions are used to compensate for the potential interference of environmental changes on users' physiological state while maintaining the effectiveness of the comparative study of environmentally adapted nutrition protocols.

[0018] Furthermore, user physiological outcome data and subjective feedback data were collected during the implementation of the environmentally adapted nutrition experimental protocol. Combined with concurrently acquired environmental dynamic data, and controlling for environmental confounding factors through a causal effect estimation module, the impact of the comparative nutrition intervention hypothesis on user physiological outcomes was analyzed to generate personalized nutrition decision-making rules, including:

[0019] The physiological results data, subjective feedback data, and environmental dynamic data collected during the implementation of the environmentally adapted nutrition experimental protocol were aligned and divided according to the implementation stage corresponding to the comparative nutrition intervention hypothesis to form at least two comparative datasets.

[0020] Based on the comparative dataset, a causal inference model is constructed, in which environmental dynamic data is used as the covariate input of the causal inference model;

[0021] Using a causal inference model, after controlling for the influence of environmental confounding factors represented by environmental dynamic data, the net effect of the comparative nutritional intervention hypothesis on user physiological outcome data is estimated, and the statistical significance of the net effect is calculated.

[0022] Based on net effect and statistical significance, the winning hypothesis with better physiological outcomes is determined among the comparative nutritional intervention hypotheses;

[0023] The core nutritional intervention parameters in the environmentally adapted nutritional experimental protocol corresponding to the winning hypothesis are extracted, and combined with the influence weight of environmental dynamic data in the causal inference model, to generate conditional personalized nutritional decision rules.

[0024] Personalized nutrition decision-making rules include core nutritional intervention parameters, applicable environmental conditions, and additional adjustment suggestions when environmental conditions exceed the applicable environmental conditions.

[0025] Furthermore, a user-environment joint state vector is constructed based on user multidimensional data, including:

[0026] Sliding window sampling and statistical feature extraction are performed on the user's dynamic time-series physiological data to obtain the user's dynamic physiological feature vector;

[0027] Perform one-hot encoding or numerical normalization on the user's static attribute data to obtain the user's static feature vector;

[0028] Perform text sentiment analysis or numerical rating conversion on user subjective feedback data to obtain user subjective state feature vectors;

[0029] The user's dynamic physiological feature vector, static feature vector, and subjective state feature vector are concatenated to generate the user's physiological state feature vector.

[0030] Time alignment and numerical normalization are performed on real-time environmental monitoring data to obtain environmental context feature vectors.

[0031] The user's physiological state feature vector is concatenated with the environmental context feature vector to generate a user-environment joint state vector.

[0032] The user-environment joint state vector is used to characterize the comprehensive system state under the combined influence of the user's physiological state and external environmental pressure at a specific point in time.

[0033] Furthermore, time-series analysis models are used to identify the correlation between user physiological patterns and environmental fluctuations, generating personalized environmental-physiological response baselines, including:

[0034] The user-environment joint state vector is input into the time series analysis model in the form of a time series. The time series analysis model is either a long short-term memory network model or a Transformer time series model.

[0035] Temporal dependency features of the user-environment joint state vector are extracted using a long short-term memory network model or a Transformer temporal model.

[0036] Calculate the cross-attention weights or hidden state correlations between environmental context features and user physiological state features based on temporal dependency features;

[0037] Based on cross-attention weights or hidden state correlation, a gradient analysis algorithm is used to calculate the predicted output change of user physiological state characteristics by a unit change in environmental context features.

[0038] The predicted output change is fitted to a parametric function, which takes environmental dynamic data as input and the baseline predicted value of the user's physiological state characteristics as output, forming a personalized environmental-physiological response baseline.

[0039] Furthermore, based on the personalized environmental-physiological response baseline and user multidimensional data, at least one high-uncertainty dimension of the user's nutritional response is identified through the uncertainty assessment module, including:

[0040] Historical nutrient intake sequences and corresponding historical physiological response sequences are extracted from user multidimensional data. The historical nutrient intake sequences contain intake data for multiple nutrient dimensions, and the historical physiological response sequences contain data for multiple physiological indicators.

[0041] The historical physiological response sequence is input into the personalized environment-physiological response baseline, and combined with the environmental dynamic data of the same period, the environmentally corrected historical physiological response sequence is calculated. The environmentally corrected historical physiological response sequence has removed the baseline influence of environmental fluctuations on physiological indicator data.

[0042] Multivariate association analysis was performed on the environmentally corrected historical physiological response sequence and the historical nutrient intake sequence to calculate the association strength measure between each nutrient dimension and each physiological indicator.

[0043] Based on the correlation strength metric, a nutrition-physiological response uncertainty map is constructed. The nutrition-physiological response uncertainty map uses the nutrition dimension as the horizontal axis, the physiological index as the vertical axis, and the correlation strength metric as the map value.

[0044] Clustering algorithms were used to divide the nutritional-physiological response uncertainty map into regions and identify low-association-strength regions where the association strength metric was below a preset threshold.

[0045] At least one nutritional dimension corresponding to a region with low correlation strength is identified as at least one high-uncertainty dimension of the user's nutritional response.

[0046] Furthermore, at least two comparative nutritional intervention hypotheses are generated based on the high uncertainty dimension, including:

[0047] For each dimension with high uncertainty, at least two different parameter values ​​are selected from the pre-defined nutritional parameter space as candidate intervention values;

[0048] Based on user static attribute data and health goal data in multidimensional user data, a constraint satisfaction algorithm is used to generate at least two combinations of nutritional intervention parameters. Each combination of nutritional intervention parameters uses different candidate intervention values ​​in the high uncertainty dimension, while keeping the parameter values ​​consistent in other nutritional dimensions.

[0049] Each combination of nutritional intervention parameters is transformed into a verifiable comparative nutritional intervention hypothesis, which states that, under the control of other nutritional and environmental variables, the combination of nutritional intervention parameters will produce better physiological outcomes for users.

[0050] Logical consistency checks are performed on at least two generated comparative nutritional intervention hypotheses to ensure that at least two comparative nutritional intervention hypotheses form a valid comparative relationship in the high uncertainty dimension.

[0051] Furthermore, the comparative nutritional intervention hypothesis is transformed into an actionable dietary template that includes controllable variables, including:

[0052] Each comparative nutritional intervention hypothesis is analyzed into specific daily nutrient allocation rules, food category selection rules, or meal timing rules.

[0053] Based on daily nutrient allocation rules, food category selection rules, or meal timing rules, a structured dietary framework is generated. The dietary framework defines the nutrient variables that need to be controlled and their target values, but does not specify specific food items.

[0054] The dietary framework is matched with a general food composition database to generate an initial recipe draft that includes specific foods, portion sizes, and meal schedules.

[0055] The initial recipe draft undergoes safety and feasibility verification. Safety verification includes checking whether it complies with dietary restrictions in the user's multidimensional data, while feasibility verification includes checking whether key ingredients are basically available within the preset geographical and time range.

[0056] The initial recipe draft that passes verification will be identified as an executable diet template.

[0057] Furthermore, environmental dynamic data is integrated with executable dietary templates, and environmental adaptive adjustment rules are added to the executable dietary templates to form an environmentally adapted nutritional experimental protocol, including:

[0058] Based on real-time environmental monitoring data, identify the types and intensities of current and predicted future environmental pressures;

[0059] Based on the type and intensity of environmental stress, at least one general environmental adjustment rule applicable to the executable diet template is matched from the preset environmental-nutrition adjustment rule library;

[0060] General environmental adjustment guidelines include recommendations for adjusting water intake, electrolyte intake, food texture, or cooking methods for specific environmental conditions;

[0061] By logically binding general environmental adjustment rules with executable dietary templates, an environmentally adapted nutritional experimental protocol is generated.

[0062] In environmentally adapted nutrition experimental protocols, the implementation of general environmental adjustment rules is configured to not affect the comparison of core variables between at least two comparative nutritional intervention hypotheses.

[0063] In a second aspect, the present invention also provides a personalized nutrition recommendation system based on dynamic environmental adaptation, applicable to the method described in the first aspect. The system includes a data acquisition module, a user-environment modeling module, an uncertainty assessment module, an experiment generation module, an environment adaptation module, a dynamic execution module, and a result analysis module. The data acquisition module acquires multidimensional user data and dynamic environmental data. The multidimensional user data includes static user attribute data, dynamic time-series physiological data, and subjective user feedback data. The dynamic environmental data includes real-time environmental monitoring data and supply chain status data. The user-environment modeling module constructs a user-environment joint state vector based on the multidimensional user data and uses a time-series analysis model to identify the correlation between user physiological patterns and environmental fluctuations, generating a personalized environment-physiological response baseline. The uncertainty assessment module identifies at least one high-uncertainty dimension of the user's nutritional response based on the personalized environment-physiological response baseline and the multidimensional user data. The experiment generation module generates at least two comparative nutritional recommendations based on at least one high-uncertainty dimension. The experiment proposes a nutritional intervention hypothesis and transforms at least two comparative nutritional intervention hypotheses into an executable dietary template containing controllable variables. An environmental adaptation module integrates dynamic environmental data with the executable dietary template, adding environmental adaptive adjustment rules to form an environmentally adapted nutritional experimental protocol. A dynamic execution module, based on real-time supply chain status data, recommends food substitutes that meet nutritional and availability requirements during the user's execution of the environmentally adapted nutritional experimental protocol using a dynamic food substitution engine. It also triggers environmental compensatory fine-tuning of the execution parameters based on real-time environmental monitoring data. A results analysis module collects user physiological outcome data and subjective feedback data during the execution of the environmentally adapted nutritional experimental protocol. Combined with concurrently acquired dynamic environmental data, and controlling for environmental confounding factors through a causal effect estimation module, it analyzes the impact of at least two comparative nutritional intervention hypotheses on user physiological outcomes, generating personalized nutritional decision-making rules.

[0064] Unlike existing technologies, the above-mentioned technical solution provides a personalized nutrition recommendation method and system based on dynamic environmental adaptation. It constructs a user-environment joint state vector based on multidimensional user data and generates a personalized environment-physiological response baseline. Based on this baseline, it identifies high-uncertainty dimensions of the user's nutritional response and generates comparative nutritional intervention hypotheses, which are further transformed into executable dietary templates. Adaptive adjustment rules are added using dynamic environmental data to form an environment-adaptive nutritional experimental protocol. During protocol execution, a dynamic food replacement engine recommends alternatives and triggers environmental compensatory fine-tuning based on real-time environmental monitoring data. It collects user physiological outcome data and user subjective feedback data during protocol execution and controls environmental confounding factors through a causal effect estimation module to analyze the impact of comparative nutritional intervention hypotheses and generate personalized nutrition decision rules. This invention achieves accurate causal inference and decision-making regarding the effects of personalized nutritional interventions in dynamic environments.

[0065] The above description of the invention is merely an overview of the technical solution of this application. In order to enable those skilled in the art to better understand the technical solution of this application and to implement it based on the description and drawings, and to make the above-mentioned objectives and other objectives, features and advantages of this application easier to understand, the following description is provided in conjunction with the specific embodiments and drawings of this application. Attached Figure Description

[0066] The accompanying drawings are only used to illustrate the principles, implementation methods, applications, features, and effects of specific embodiments of the present invention and other related contents, and should not be considered as limitations on this application.

[0067] In the accompanying drawings of the instruction manual:

[0068] Figure 1 This is a schematic diagram illustrating steps S101 to S107 of the method described in the specific implementation embodiment;

[0069] Figure 2 This is a schematic diagram illustrating steps S201 to S205 of the method described in a specific implementation.

[0070] Figure 3 This is a schematic diagram illustrating steps S301 to S305 of the method described in a specific implementation.

[0071] Figure 4 This is a schematic diagram illustrating steps S401 to S407 of the method described in a specific embodiment;

[0072] Figure 5 This is a schematic diagram of the recommendation system described in a specific implementation.

[0073] The reference numerals used in the above figures are explained as follows:

[0074] 1. Recommendation system;

[0075] 11. Data acquisition module;

[0076] 12. User-Environment Modeling Module;

[0077] 13. Uncertainty Assessment Module;

[0078] 14. Experiment generation module;

[0079] 15. Environment adaptation module;

[0080] 16. Dynamic execution module;

[0081] 17. Results Analysis Module. Detailed Implementation

[0082] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this application in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this application and are therefore intended only as examples, not as limiting the scope of protection of this application.

[0083] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this application, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0084] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit this application.

[0085] In the description of this application, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" logical relationship.

[0086] In this application, terms such as “first” and “second” are used only to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy or order relationship between these entities or operations.

[0087] Without further limitations, the use of terms such as “comprising,” “including,” “having,” or other similar open-ended expressions in this application is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0088] As understood in the Examination Guidelines, in this application, expressions such as "greater than," "less than," and "exceeding" are understood to exclude the stated number; expressions such as "above," "below," and "within" are understood to include the stated number. Furthermore, in the description of the embodiments in this application, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times," unless otherwise explicitly specified.

[0089] In the description of the embodiments of this application, the space-related expressions used, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "vertical," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or drawings. They are only for the purpose of describing the specific embodiments of this application or for the reader's understanding, and do not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0090] The processor described in the embodiments of this application can be implemented by hardware, firmware, software, or a combination thereof. It can be a circuit, one or more of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field-programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, or a microprocessor. It also includes other physical, biological, or chemical structures that can implement the same or equivalent functions as the processors listed above, such as biological neurons, quantum computing units, DNA computing units, etc., so that the processor can execute some or all of the steps in the computer program or method involved in the various embodiments of this application, or any combination of the steps mentioned therein.

[0091] The computer program involved in the embodiments can be stored in a computer device readable storage medium, which includes, but is not limited to, disks, magnetic tapes, magnetic cards, floppy disks, flash memory, optical disks, optical cards, read-only memory (ROM), random access memory (RAM), erasable programmable ROM (EPROM), and electrically erasable programmable ROM (EEPROM), etc., and also includes other biological, physical, or chemical structures that can achieve the same or equivalent functions as the storage media listed above, such as DNA, RNA, proteins, and other units with information storage capabilities. In specific embodiments, the storage medium involved can be one of the above-mentioned media types, or a combination of the above-mentioned media types. In different embodiments, the computer program involved in the embodiments can be centrally stored in a single medium, or distributed and stored in multiple media. The memory containing the computer device readable storage medium can be non-volatile memory or random access memory. These computer device readable storage media can be built into the device, or can be connected to the device involved in the embodiments as an external device or part of an external device. In some embodiments, the memory having a computer device readable storage medium is deployed locally; in other embodiments, the memory may be deployed remotely from the processor, for example, as a network-attached memory accessed via RF circuitry or an external port and a communication network, wherein the communication network may be the Internet, one or more intranets, a local area network (LAN), a wide area network (WLAN), a storage area network (SAN), or a suitable combination thereof, as long as computer device access to the memory is enabled. Furthermore, the computer program involved in the embodiments may be stored in plaintext / ciphertext form, or it may be designed as training data, integrated and recombined through model training and implicitly stored in the parameter states of a deep neural network or other machine learning model.

[0092] Please see Figure 1 In a first aspect, this embodiment provides a personalized nutrition recommendation method based on dynamic environmental adaptation, including:

[0093] S101. Acquire multidimensional user data and dynamic environmental data. Multidimensional user data includes static user attribute data, dynamic time-series physiological data, and subjective user feedback data. Dynamic environmental data includes real-time environmental monitoring data and supply chain status data.

[0094] S102. Construct a user-environment joint state vector based on user multidimensional data, and use a time series analysis model to identify the correlation between user physiological patterns and environmental fluctuations, and generate a personalized environmental-physiological response baseline.

[0095] S103. Based on the personalized environmental-physiological response baseline and user multidimensional data, identify at least one high-uncertainty dimension of the user's nutritional response through the uncertainty assessment module;

[0096] S104. Generate at least two comparative nutritional intervention hypotheses based on the high uncertainty dimension, and transform the comparative nutritional intervention hypotheses into an executable diet template containing controllable variables;

[0097] S105. Integrate dynamic environmental data with executable diet templates, add environmental adaptive adjustment rules to the executable diet templates, and form an environmentally adapted nutritional experimental protocol.

[0098] S106. During the execution of the environmentally adapted nutrition experiment plan by the user, based on the real-time supply chain status data, the dynamic food substitution engine recommends food substitutes that meet the nutritional and availability requirements for the environmentally adapted nutrition experiment plan, and based on the real-time environmental monitoring data, triggers environmental compensation fine-tuning of the execution parameters of the environmentally adapted nutrition experiment plan.

[0099] S107. Collect user physiological outcome data and user subjective feedback data during the execution of the environmentally adapted nutrition experimental protocol, and combine them with the environmental dynamic data obtained at the same time. Control the environmental confounding factors through the causal effect estimation module, analyze the impact of the comparative nutrition intervention hypothesis on user physiological outcomes, and generate personalized nutrition decision rules.

[0100] In step S101, the user multidimensional data encompasses various aspects of information describing the user's individual characteristics and real-time status. Static user attribute data refers to relatively stable demographic and basic physiological information, including attributes that do not change frequently over time, such as age, gender, basal metabolic rate, genotype information, and known medical history. Dynamic user time-series physiological data consists of physiological indicators continuously collected by sensor devices and compiled into time series, such as continuous blood glucose monitoring values, heart rate variability, sleep stages, and changes in body composition. User subjective feedback data is collected through the application interface, covering the user's self-reported energy level, hunger, satiety, emotional state, and satisfaction ratings for past diets. Environmental dynamic data reflects the real-time status of the external physical world and resource supply. Real-time environmental monitoring data refers to physical environmental parameters obtained through external data interfaces, including temperature, humidity, ultraviolet radiation intensity, and pollutant concentration. Supply chain status data relates to the availability and circulation information of food ingredients. In practical applications, supply chain status data can be obtained by accessing the inventory query interface of local supermarkets, the application programming interface (API) of fresh food e-commerce platforms, or food availability information shared by the user community.

[0101] In step S102, constructing the user-environment joint state vector aims to integrate the user's internal physiological state and external environmental stress into a unified mathematical representation, enabling the model to analyze the interaction between the two. The user-environment joint state vector is typically constructed by converting the various data obtained in step S101 into numerical features and concatenating them using feature engineering methods. The time-series analysis model is used to uncover the patterns of changes in user physiological indicators over time and to assess how these patterns are affected by environmental fluctuations. The personalized environment-physiological response baseline is a function or set of rules learned by the time-series analysis model. It quantifies the expected baseline impact of unit changes in environmental parameters on various user physiological indicators in the absence of specific nutritional intervention. This baseline allows for the removal of background noise from environmental fluctuations in subsequent analyses, enabling a purer observation of the effects of nutritional intervention itself. For example, the personalized environment-physiological response baseline might reveal that the user's resting heart rate exhibits a regular increase in a high-temperature environment. This step establishes a personalized model of the impact of the environment on user physiology, providing a crucial reference for distinguishing between the effects of nutritional intervention and confounding environmental effects.

[0102] In step S103, the uncertainty assessment module is an analytical unit used to evaluate the clarity of a user's response patterns to different nutritional factors. Based on the personalized environment-physiological response baseline generated in step S102, it re-analyzes the user's historical multidimensional data (especially nutritional intake and physiological response data) to identify nutritional dimensions where, even after removing the influence of the environmental baseline, the correlation between the user's historical data and physiological outcomes remains unclear, unstable, or statistically insignificant—these are high-uncertainty dimensions. Identifying high-uncertainty dimensions is a prerequisite for subsequent targeted scientific exploration, ensuring that limited experimental resources are used to address the most critical cognitive blind spots, rather than repeatedly verifying known patterns.

[0103] In step S104, based on the identified high-uncertainty dimensions, the system proactively constructs a scientific experiment. The comparative nutritional intervention hypothesis is a mutually exclusive and verifiable proposition addressing the same high-uncertainty dimension, such as "increasing the proportion of carbohydrates at breakfast is better than increasing the proportion of carbohydrates at dinner." The executable dietary template transforms the abstract intervention hypothesis into a concrete dietary framework that users can follow in their daily lives. It clarifies the core nutritional variables (i.e., controllable variables) that need to be strictly controlled during the experiment and their target range, but may retain some flexibility in specific food choices to ensure the feasibility of the plan and user compliance.

[0104] In step S105, the environmental adaptive adjustment rule is a set of predefined logical conditions and execution actions that adaptively modify the non-comparative execution parameters in the executable diet template based on real-time or predicted environmental dynamic data. The environmentally adaptive nutrition experimental protocol is the final, complete experimental execution plan capable of self-adjusting according to the environmental context. It ensures that during experimental execution, certain non-core experimental parameters (such as hydration recommendations) can be automatically adjusted based on real-time environmental conditions (such as sudden temperature increases) to counteract the potential interference of environmental changes on the user's physiological state, thereby maintaining relative consistency of experimental conditions across different control groups outside of core variables. This step enhances the robustness and effectiveness of the experimental protocol in the dynamic real world.

[0105] In step S106, the dynamic ingredient replacement engine is a real-time decision-making function that recommends alternative ingredients that meet the nutritional requirements of the original plan to the user based on real-time changes in supply chain status data during the execution of the experimental protocol. When the ingredients specified in the original plan are unavailable, the engine is activated to find alternative options with similar nutritional properties that are currently available. Environmental compensatory fine-tuning is an immediate, small-scale adaptive adjustment of certain execution parameters (such as fluid intake and supplement recommendations) in the plan during the execution of the experimental protocol, based on real-time environmental monitoring data. Triggering fine-tuning aims to quickly compensate for the instantaneous stress on the user's physiology caused by sudden environmental changes (such as increased air pollution) to maintain the relative stability of the user's baseline state within the experimental observation window. This step, by introducing dynamic adaptability during the execution phase, solves the two major challenges of resource availability and environmental abrupt changes faced by experimental protocols in actual implementation.

[0106] In step S107, user physiological outcome data refers to objective indicators (such as average blood glucose and sleep heart rate variability) obtained through device monitoring during the experiment; user subjective feedback data refers to results reported by the user (such as energy scores and digestive feelings). The causal effect estimation module uses statistical or machine learning methods to perform causal inference, incorporating concurrently acquired environmental dynamic data as covariates into the analysis model to control for environmental confounding factors—variables that affect both environmental dynamic data and user physiological outcomes but are unrelated to nutritional intervention. By controlling for environmental confounding factors, the causal effect estimation module can more purely estimate the net effect of comparative nutritional intervention hypotheses on user physiological outcomes, thereby analyzing which comparative nutritional intervention hypothesis is superior under specific conditions. The final personalized nutrition decision-making rule is a conditionally applicable personalized dietary guidance principle derived from the above causal analysis conclusions. This step completes a closed loop from experimental execution and data collection to causal identification and knowledge extraction, realizing personalized nutrition decision-making based on controlled experimental evidence.

[0107] This embodiment quantifies environmental disturbances by constructing a personalized environment-physiological response baseline and proactively generates comparative nutrition experiments based on uncertainty assessment, transforming nutrition recommendations from static fitting to dynamic scientific exploration. By integrating environmental adaptive rules and real-time dynamic adjustment mechanisms into the experimental protocol execution, the feasibility and validity of the experiment under complex real-world conditions are effectively guaranteed. Finally, by introducing dynamic environmental data as covariates for causal effect estimation, a pure attribution of the true effects of nutritional intervention is achieved while controlling for the influence of confounding environmental factors, thereby generating highly reliable and adaptive personalized nutrition decision-making rules.

[0108] Please see Figure 2 In some embodiments, during the execution of an environment-adaptive nutrition experiment by a user, based on real-time supply chain status data, a dynamic ingredient replacement engine recommends food substitutes that meet nutritional and availability requirements. Furthermore, based on real-time environmental monitoring data, it triggers environmentally compensated fine-tuning of the execution parameters of the environment-adaptive nutrition experiment, including:

[0109] S201. When the preset ingredients in the environmentally adapted nutrition experiment plan are unavailable, query the set of locally available alternative ingredients based on the real-time supply chain status data.

[0110] S202. Calculate the similarity of nutritional components between each alternative food in the alternative food set and the preset food, and evaluate the environmental adaptability of each alternative food in combination with the environmental context indicated by the current real-time environmental monitoring data.

[0111] S203. Based on the similarity of nutritional components and environmental adaptability, target alternative ingredients are determined from the set of alternative ingredients, and the recommended intake of target alternative ingredients in the environmentally adapted nutrition experimental protocol is automatically adjusted.

[0112] S204. Continuously monitor real-time environmental monitoring data. When the monitored environmental parameter values ​​exceed the preset environmental parameter threshold range corresponding to the personalized environment-physiological response baseline, generate an environmental compensatory adjustment instruction for at least one execution parameter in the environmentally adapted nutrition experimental protocol. The execution parameters include at least one of the following: water intake, electrolyte supplementation, and food cooking method.

[0113] S205. Push the environmental compensatory adjustment instruction to the user. The environmental compensatory adjustment instruction is used to compensate for the potential interference of environmental changes on the user's physiological state while maintaining the effectiveness of the comparative study of environmentally adapted nutrition protocols.

[0114] In step S201, the scenario where the pre-set ingredients are unavailable refers to determining, based on the query results of real-time supply chain status data, that the specific food items specified in the original environment-adaptive nutrition experiment plan are out of stock, in limited supply, or have prices exceeding a preset tolerance range in the current user's geographical location or designated procurement channel. The construction of a set of locally available alternative ingredients relies on the accessed supply chain database, which must contain ingredient inventory, geographical location, and real-time updated information. The query process is a real-time data retrieval and filtering operation based on constraints (such as ingredient category and nutritional attributes).

[0115] In step S202, nutrient similarity is used to measure the degree of similarity between the substitute ingredient and the preset ingredient in terms of key nutrient content. Calculating nutrient similarity typically involves selecting a set of core nutrient indicators (such as protein, fat, carbohydrates, dietary fiber, specific vitamins, or minerals) and using methods such as vector space models or weighted Euclidean distance for calculation. Environmental adaptability is another evaluation dimension, considering whether the physicochemical properties of the substitute ingredient (such as moisture content, traditional understanding of its heat / cooling properties, and whether it is conducive to digestion and absorption under specific environmental conditions) or the recommended cooking method are suitable for the environmental conditions under the specific environmental context reflected by current real-time environmental monitoring data (such as high temperature, high humidity, dryness, and pollution). Environmental adaptability can be evaluated based on a preset environment-ingredient adaptability knowledge base or rule base.

[0116] In step S203, determining the target alternative food is a multi-objective decision-making process that comprehensively weighs nutritional similarity and environmental suitability. Preferably, a weighted scoring method or multi-criteria decision algorithm is used to assign weights to the two factors and calculate a comprehensive score, selecting the alternative food with the highest score. Automatically adjusting the recommended intake ensures that after replacement, the total amount of core nutrients ingested by the user remains consistent with the original plan's preset target. Specifically, this can be achieved by converting the nutritional equivalents based on the difference in unit nutrient density between the target alternative food and the preset food.

[0117] In step S204, the preset environmental parameter threshold range is set based on the personalized environment-physiological response baseline. The personalized environment-physiological response baseline quantifies the normal range of influence of environmental parameters on the user's physiological indicators. The preset environmental parameter threshold range defines an acceptable range of environmental fluctuations that do not require additional compensation. It can be set based on the personalized environment-physiological response baseline by calculating the statistical distribution of environmental parameters during historical stable periods (e.g., mean ± 2 standard deviations), or it can be directly set as the environmental input range corresponding to when the baseline function output remains stable, used to define the boundary of normal environmental fluctuations that do not require additional compensation. When the real-time monitoring data continuously or momentarily exceeds this threshold range, it indicates that environmental changes may have a significant impact on the user's physiology that exceeds the baseline expectations.

[0118] The logic for generating environmental compensatory adjustment instructions stems from a pre-defined rule base mapping environmental anomalies to compensatory actions. For example, when the temperature exceeds a threshold and humidity is low, instructions to increase water intake and recommend cooking methods for juicy ingredients may be generated. The selection of execution parameters aims to directly address physiological needs or risks that may be exacerbated by specific environmental stresses.

[0119] In step S205, the timing and manner of sending environmental compensatory adjustment instructions take into account user experience and compliance. The instructions clearly state their compensatory nature and emphasize that they do not affect the core variables of the ongoing comparative experiment (i.e., the differences between different nutritional intervention hypotheses), thus ensuring that users understand that the adjustment is used to maintain the stability of experimental conditions, rather than changing the experiment itself. It should be noted that the execution parameters targeted by the environmental compensatory adjustment instructions, such as water intake and electrolyte supplementation, are not core nutritional variables to be compared in the comparative nutritional intervention hypotheses (such as macronutrient ratios, feeding times, etc.). Such adjustments aim to provide a more stable physiological baseline environment for all control groups, and their rules apply consistently to all experimental groups, therefore they will not interfere with the comparison of core variables between different nutritional intervention hypotheses.

[0120] This embodiment introduces a dual assessment mechanism of nutritional similarity and environmental adaptability, ensuring that food substitution not only meets nutritional equivalence but also considers environmental suitability, thus enhancing the scientific rigor and personalization of substitution decisions. By linking compensatory environmental adjustments to a personalized environmental-physiological response baseline and setting clear trigger thresholds, fine-tuning actions are upgraded from empirical, general responses to precise, predictive interventions based on individual physiological models. Together, these embodiments enhance the robustness, feasibility, and scientific validity of environmentally adaptable nutrition experimental protocols in the face of resource fluctuations and environmental abrupt changes, ensuring the effectiveness of experimental data collection and the reliability of subsequent causal inferences.

[0121] Please see Figure 3 In some embodiments, user physiological outcome data and user subjective feedback data are collected during the execution of an environment-adaptive nutrition experimental protocol. Combined with concurrently acquired environmental dynamic data, and using a causal effect estimation module to control for environmental confounding factors, the impact of the comparative nutrition intervention hypothesis on user physiological outcomes is analyzed to generate personalized nutrition decision-making rules, including:

[0122] S301. The user physiological results data, user subjective feedback data and environmental dynamic data collected during the execution of the environmentally adapted nutrition experimental protocol are aligned and divided according to the execution stage corresponding to the comparative nutrition intervention hypothesis to form at least two comparative datasets.

[0123] S302. Based on the comparative dataset, construct a causal inference model, in which environmental dynamic data is used as the covariate input of the causal inference model;

[0124] S303. Using a causal inference model, after controlling for the influence of environmental confounding factors represented by the dynamic environmental data, estimate the net effect of the comparative nutritional intervention hypothesis on the user's physiological outcome data, and calculate the statistical significance of the net effect.

[0125] S304. Based on net effect and statistical significance, determine the winning hypothesis among comparative nutritional intervention hypotheses that has better physiological outcomes.

[0126] S305. Extract the core nutritional intervention parameters from the environmentally adapted nutritional experimental protocol corresponding to the winning hypothesis, and combine them with the influence weight of environmental dynamic data in the causal inference model to generate conditional personalized nutritional decision rules.

[0127] Personalized nutrition decision-making rules include core nutritional intervention parameters, applicable environmental conditions, and additional adjustment suggestions when environmental conditions exceed the applicable environmental conditions.

[0128] In step S301, the alignment and partitioning operations ensure that the data corresponding to each comparative nutritional intervention hypothesis are independent and complete in the time dimension. The alignment process matches user physiological outcome data, user subjective feedback data, and concurrent environmental dynamic data according to a unified time benchmark; the partitioning process divides the matched overall data stream into mutually independent subsets according to the start and end times of the user's actual implementation of different nutritional intervention hypotheses. Each subset constitutes a comparative dataset, which contains all observation results and corresponding environmental condition records during the implementation of a specific hypothesis.

[0129] In step S302, the causal inference model is constructed based on a comparative dataset, using identifiers representing different nutritional intervention hypotheses as primary analytical variables and user physiological outcome data as variables to be explained. Environmental dynamic data are input into the model as a set of auxiliary variables, which are used in the statistical model to correct for their interference with the estimation of the relationship between the primary analytical variables and the variables to be explained.

[0130] In step S303, the net effect is estimated using a causal inference model. This process statistically simulates the logic of "comparing the differences in outcomes from different interventions under the same conditions (especially environmental conditions)." The model uses mathematical methods to remove the portion of outcome variation represented by dynamic environmental data when calculating the intervention effect. The calculation of statistical significance typically involves hypothesis testing of the estimated net effect value to assess whether the degree to which the effect value deviates from zero (i.e., no effect) exceeds the range explainable by random fluctuations.

[0131] In step S304, the winning hypothesis is determined based on a combination of the direction, magnitude, and statistical significance of the net effect. For example, a rule can be set: when the net effect corresponding to a hypothesis points to a beneficial physiological change and its statistical significance measure is better than a preset statistical standard, then the hypothesis is determined to be the winning hypothesis. If none of the hypotheses meet this standard, it may be concluded that no clear winning conclusion has been found under the current experimental conditions.

[0132] In step S305, the core nutritional intervention parameters are essential nutritional variables extracted from the experimental protocols corresponding to the selected hypothesis, distinguishing them from other hypotheses. The influence weights of environmental dynamic data in the model quantify the relative contributions of different environmental factors to the model's explanation of physiological outcome variability. Conditionalized personalized nutritional decision rules are guiding principles with environmental suitability statements, clarifying the range of environmental conditions under which the recommended core parameters are validated, and providing additional adjustment suggestions based on model weight inference or safety and conservatism principles for situations where environmental conditions exceed this range. The determination of the applicable environmental condition range can refer to the distribution characteristics of environmental data during model training, or be defined based on the model's predictive stability across different environmental subsets.

[0133] This embodiment provides a technical means to control for environmental confounding factors in observational studies by systematically incorporating dynamic environmental data as covariates into a causal inference model, thereby enabling a more reliable estimation of the net effect of nutritional interventions. By extracting core parameters and combining them with the influence weights of environmental factors to generate conditional rules, the final nutritional decision is no longer a static plan, but a contextualized strategy that dynamically adapts to environmental changes, thus improving the accuracy, robustness, and practical value of personalized nutritional recommendations.

[0134] Please see Figure 4 In some embodiments, a user-environment joint state vector is constructed based on user multidimensional data, including:

[0135] S401. Perform sliding window sampling and statistical feature extraction on the user's dynamic time-series physiological data to obtain the user's dynamic physiological feature vector;

[0136] S402. Perform one-hot encoding or numerical normalization on the user's static attribute data to obtain the user's static feature vector.

[0137] S403. Perform text sentiment analysis or numerical rating conversion on the user's subjective feedback data to obtain the user's subjective state feature vector.

[0138] S404. Concatenate the user's dynamic physiological feature vector, user's static feature vector, and user's subjective state feature vector to generate the user's physiological state feature vector.

[0139] S405. Perform time alignment and numerical normalization operations on the real-time environmental monitoring data to obtain the environmental context feature vector.

[0140] S406. Concatenate the user's physiological state feature vector with the environmental context feature vector to generate a user-environment joint state vector.

[0141] S407. The user-environment joint state vector is used to characterize the comprehensive system state under the combined influence of the user's physiological state and external environmental pressure at a specific point in time.

[0142] In step S401, sliding window sampling refers to the process of extracting subsequences from the continuous time series of user dynamic physiological data according to a fixed time length or number of data points. Adjacent windows may overlap. The statistical feature extraction operation calculates a series of statistics for each subsequence within the sliding window, such as mean, standard deviation, maximum, minimum, skewness, kurtosis, or more complex time-domain and frequency-domain features. The user dynamic physiological feature vector is composed of the statistics extracted from multiple windows arranged in sequence, thereby converting the time series information into a fixed-dimensional feature representation.

[0143] In step S402, one-hot encoding is a method to convert discrete categorical variables into binary vectors. Each dimension of the vector corresponds to a possible category; if the category appears, the corresponding dimension is 1, and otherwise 0. Numerical normalization is the process of scaling numerical user static attribute data (such as age, height, and weight) with different dimensions or ranges to a uniform interval (such as between 0 and 1) to eliminate the influence of dimensions. The user static feature vector is formed by concatenating all static attribute values ​​after one-hot encoding or numerical normalization.

[0144] In step S403, text sentiment analysis refers to processing the text content (such as dietary log descriptions) in user subjective feedback data, using natural language processing technology to identify the sentiment tendency or emotional state expressed in the text, and quantifying it into a numerical score. The numerical score conversion operation involves standardizing or directly adopting the numerical subjective scores provided by the user. The user subjective state feature vector is composed of these analyzed or converted numerical features.

[0145] In step S404, the user's physiological state feature vector comprehensively represents the user's overall physiological state at a specific moment, which is jointly determined by their own attributes, dynamic physiological signals, and subjective feelings.

[0146] In step S405, the time alignment operation ensures that the timestamps of the real-time environmental monitoring data are consistent with the time base of the user's multidimensional data, enabling the environmental data to establish a correspondence with the user's status data at the same point in time. The numerical normalization operation is also applied to the environmental monitoring data, transforming the raw values ​​of different environmental parameters (such as temperature, humidity, and PM2.5 concentration) to the same scale range. The environmental context feature vector consists of multiple environmental parameter values ​​arranged in sequence after alignment and normalization.

[0147] In step S406, the splicing results in the final, higher-dimensional user-environment joint state vector, which realizes the fusion of user internal state and external environment information in terms of data structure.

[0148] In step S407, each dimension of the user-environment joint state vector corresponds to the quantified state of the user-environment system in a specific aspect. The user-environment joint state vector serves as input to the subsequent time-series analysis model, used to model and predict the dynamic interaction between environmental stress and the user's physiological state.

[0149] This embodiment transforms time-series physiological data into fixed-dimensional features through sliding window sampling and statistical feature extraction, resolving the issue of inconsistent model input dimensions. Specific data preprocessing techniques, such as one-hot encoding, numerical normalization, and text sentiment analysis, convert heterogeneous multi-source data into standardized numerical features that the model can process. Through two concatenation operations, a hierarchical feature construction is achieved, from raw data to user physiological state feature vectors, and then to user-environment joint state vectors incorporating environmental information. This embodiment constitutes a key technical means for generating high-quality, structured model inputs from raw multimodal data, providing a reliable data foundation for accurately establishing personalized environment-physiological response baselines.

[0150] In some embodiments, a time-series analysis model is used to identify the correlation between user physiological patterns and environmental fluctuations, generating a personalized environmental-physiological response baseline, including:

[0151] The user-environment joint state vector is input into the time series analysis model in the form of a time series. The time series analysis model is either a long short-term memory network model or a Transformer time series model.

[0152] Temporal dependency features of the user-environment joint state vector are extracted using a long short-term memory network model or a Transformer temporal model.

[0153] Calculate the cross-attention weights or hidden state correlations between environmental context features and user physiological state features based on temporal dependency features;

[0154] Based on cross-attention weights or hidden state correlation, a gradient analysis algorithm is used to calculate the predicted output change of user physiological state characteristics by a unit change in environmental context features.

[0155] The predicted output change is fitted to a parametric function, which takes environmental dynamic data as input and the baseline predicted value of the user's physiological state characteristics as output, forming a personalized environmental-physiological response baseline.

[0156] In this embodiment, the Long Short-Term Memory (LSTM) network model, through its internal gating mechanisms (input gate, forget gate, output gate) and cell states, can learn and remember long-term dependencies in the user-environment joint state vector sequence; the Transformer temporal model, on the other hand, utilizes a self-attention mechanism to calculate the correlation strength between features at any two time steps in the sequence, thereby capturing global temporal dependency patterns. Both models can effectively process sequence data and extract temporal dependency features containing dynamic patterns.

[0157] After extracting temporal dependency features, the temporal analysis model further analyzes the interaction between environmental context features and user physiological state features. For the Transformer temporal model, this analysis is achieved by calculating cross-attention weights, which quantify the model's attention to environmental context features at different times when predicting the user's physiological state at a certain moment. For the Long Short-Term Memory network model, the correlation between hidden states and physiological features can be calculated by analyzing the activation patterns of its hidden states at different time steps or by using specific network layers (such as attention layers).

[0158] Based on the aforementioned cross-attention weights or hidden state correlations, a gradient analysis algorithm is used to calculate the change in the predicted output of user physiological state features caused by a unit change in environmental context features. The gradient analysis algorithm is implemented by calculating the partial derivative of the model output with respect to specific environmental input features. The change in the predicted output reflects the expected strength and direction of the effect of small perturbations in environmental factors on the predicted user physiological state, under the patterns learned by the model.

[0159] The calculated series of predicted output changes are fitted to different environmental feature dimensions using parametric functions. These parametric functions can be linear, polynomial, or other learnable functions. The fitting process aims to establish a mathematical relationship that takes dynamic environmental data (after the same preprocessing as when constructing the user-environment joint state vector) as input and directly outputs the baseline predicted values ​​of the user's physiological state characteristics. This relationship function is the personalized environment-physiological response baseline, encapsulating the individualized mapping from environmental fluctuations to the user's physiological baseline response. Alternatively, the parametric function can be fitted by training a simplified, rapidly computed surrogate function on the training data of the time-series analysis model, using environmental features as input and the model's baseline predicted values ​​of physiological state as output, employing linear regression or ridge regression methods.

[0160] This embodiment specifies a Long Short-Term Memory (LSTM) network model and a Transformer temporal model as specific implementations of the temporal analysis model. It utilizes the internal mechanisms of these models (gating, self-attention) to extract temporal dependency features and quantifies the association between environmental and physiological characteristics through cross-attention weights or hidden state correlations. Furthermore, it introduces a gradient analysis algorithm to calculate the sensitivity of environmental changes to physiological predictions and fits this sensitivity relationship to a parameterized function, thereby transforming the complex neural network black-box model into an interpretable and computable personalized environmental-physiological response baseline. This embodiment achieves the transformation from data-driven modeling to generating a quantitative model of explicit causal relationships, providing a crucial technical foundation for accurately removing environmental confounding effects in subsequent steps.

[0161] In some embodiments, based on a personalized environmental-physiological response baseline and user multidimensional data, an uncertainty assessment module identifies at least one dimension of high uncertainty in the user's nutritional response, including:

[0162] Historical nutrient intake sequences and corresponding historical physiological response sequences are extracted from user multidimensional data. The historical nutrient intake sequences contain intake data for multiple nutrient dimensions, and the historical physiological response sequences contain data for multiple physiological indicators.

[0163] The historical physiological response sequence is input into the personalized environment-physiological response baseline, and combined with the environmental dynamic data of the same period, the environmentally corrected historical physiological response sequence is calculated. The environmentally corrected historical physiological response sequence has removed the baseline influence of environmental fluctuations on physiological indicator data.

[0164] Multivariate association analysis was performed on the environmentally corrected historical physiological response sequence and the historical nutrient intake sequence to calculate the association strength measure between each nutrient dimension and each physiological indicator.

[0165] Based on the correlation strength metric, a nutrition-physiological response uncertainty map is constructed. The nutrition-physiological response uncertainty map uses the nutrition dimension as the horizontal axis, the physiological index as the vertical axis, and the correlation strength metric as the map value.

[0166] Clustering algorithms were used to divide the nutritional-physiological response uncertainty map into regions and identify low-association-strength regions where the association strength metric was below a preset threshold.

[0167] At least one nutritional dimension corresponding to a region with low correlation strength is identified as at least one high-uncertainty dimension of the user's nutritional response.

[0168] In this embodiment, the extraction of historical nutrient intake sequences involves organizing the daily or per-meal intake data (such as total calories, carbohydrates, protein, fat, dietary fiber, and specific trace elements) from the user's dietary records in chronological order. The historical physiological response sequence extracts physiological indicators (such as the area under the postprandial blood glucose curve, average heart rate, and sleep depth and duration) corresponding to the nutrient intake time from the user's dynamic time-series physiological data.

[0169] When calculating the environmentally corrected historical physiological response sequence, preferably, a baseline function is used to predict the expected baseline value of the user's physiological indicators under the actual environmental data at each time point. The predicted baseline value is then subtracted (or divided) from the observed historical physiological response sequence to obtain the residual sequence. This residual sequence is the environmentally corrected historical physiological response sequence, and its fluctuations mainly reflect the influence of non-environmental factors (such as nutritional intake).

[0170] Multivariate association analysis was performed on the historical physiological response sequences after environmental correction and the historical nutritional intake sequences. The strength of the association can be measured using methods such as Pearson correlation coefficient, Spearman rank correlation coefficient, or standardized coefficients based on regression models to quantify the strength of the linear or monotonic relationship between each nutritional dimension and each physiological indicator. Taking the Pearson correlation coefficient as an example, the specific calculation process is as follows: For a specific nutritional dimension (such as carbohydrate intake) and a specific physiological indicator (such as postprandial blood glucose fluctuation), firstly, obtain the environmentally corrected value sequences of these two variables at the same time point; then calculate the covariance of these two sequences and divide them by their respective standard deviations. The ratio obtained is the Pearson correlation coefficient. The closer the absolute value is to 1, the stronger the linear association; the closer it is to 0, the weaker the linear association.

[0171] The uncertainty map of nutrition-physiological response can be in the form of a matrix, where rows correspond to physiological indicators, columns correspond to nutritional dimensions, and each element in the matrix is ​​the corresponding correlation strength metric.

[0172] Clustering algorithms are used to divide the uncertainty map of nutrition-physiological response into regions. Optionally, a density-based clustering algorithm is used to group points in the map whose association strength metric values ​​are spatially adjacent (i.e., adjacent in the matrix) and numerically below a preset threshold into clusters. Each such cluster constitutes a low association strength region.

[0173] The columns (nutritional dimensions) corresponding to the identified low-correlation regions are extracted, which are then identified as at least one high-uncertainty dimension of the user's nutritional response. The identified high-uncertainty dimensions indicate that even if users have historical data and environmental interference has been removed, the regular relationship between their intake and physiological outcomes remains unclear or very weak, thus becoming targets that need to be actively explored and verified through subsequent comparative experiments.

[0174] This embodiment introduces an environmental correction step to pre-extract environmental confounding effects before correlation analysis, ensuring that the uncertainties identified subsequently truly originate from the nutritional response itself rather than environmental interference. By constructing a nutritional-physiological response uncertainty map and using a clustering algorithm to divide regions, high-dimensional and sparse correlations are visualized and structured, thereby systematically and automatically locating nutritional dimensions with weak correlations. This embodiment provides a data-driven, objective method for identifying dimensions with high uncertainty, offering precise guidance for generating targeted comparative nutritional intervention hypotheses.

[0175] In some embodiments, at least two comparative nutritional intervention hypotheses are generated based on the high uncertainty dimension, including:

[0176] For each dimension with high uncertainty, at least two different parameter values ​​are selected from the pre-defined nutritional parameter space as candidate intervention values;

[0177] Based on user static attribute data and health goal data in multidimensional user data, a constraint satisfaction algorithm is used to generate at least two combinations of nutritional intervention parameters. Each combination of nutritional intervention parameters uses different candidate intervention values ​​in the high uncertainty dimension, while keeping the parameter values ​​consistent in other nutritional dimensions.

[0178] Each combination of nutritional intervention parameters is transformed into a verifiable comparative nutritional intervention hypothesis, which states that, under the control of other nutritional and environmental variables, the combination of nutritional intervention parameters will produce better physiological outcomes for users.

[0179] Logical consistency checks are performed on at least two generated comparative nutritional intervention hypotheses to ensure that at least two comparative nutritional intervention hypotheses form a valid comparative relationship in the high uncertainty dimension.

[0180] In this embodiment, the preset nutritional parameter space defines a reasonable range of values ​​or a set of discrete options that can be explored for each nutritional dimension (such as carbohydrate intake ratio, fat type, and eating time window). For the identified high-uncertainty dimension, at least two different parameter values ​​are selected from the parameter space of that dimension as candidate intervention values.

[0181] Constraint satisfaction algorithms construct problems by defining a set of variables (each nutritional dimension), the domain of each variable (its parameter space), and constraints (restrictions derived from user attributes and objectives, such as the range of total daily calories, minimum protein intake, and contraindications to specific nutrients). Taking the backtracking search algorithm as an example, the specific generation process is as follows: First, assign initial values ​​to all nutritional dimensions (including high-uncertainty dimensions and other dimensions that need to be kept consistent) (e.g., using the user's historical average or the median recommended intake); prioritize trying different combinations of candidate intervention values ​​for high-uncertainty dimensions; for each set of values ​​assigned to high-uncertainty dimensions, try assigning values ​​to other nutritional dimensions, while checking whether all constraints are violated (e.g., whether total calories are exceeded or protein intake is sufficient); if a set of values ​​is found that satisfies all constraints, an effective combination of nutritional intervention parameters is successfully generated. In this process, "keeping parameter values ​​consistent across other nutritional dimensions" means that, among the multiple combinations of nutritional intervention parameters generated, the parameter values ​​of each nutritional dimension, except for the high-uncertainty dimension, are the same across different combinations; continue searching until a specified number (at least two) of effective parameter combinations with different values ​​on the high-uncertainty dimensions are generated, or the search space is exhausted.

[0182] When each generated combination of nutritional intervention parameters is transformed into a verifiable comparative nutritional intervention hypothesis, the transformation process involves expressing the numerical settings in the parameter combination in natural language or structured logic as a proposition with causal inference, stating that, under the control of other variables (including other nutritional variables and environmental variables controlled by experimental design), using this particular combination of parameters will lead to better physiological outcomes for the user.

[0183] The logical consistency verification of comparative nutritional intervention hypotheses includes: ensuring that all hypotheses do indeed use different candidate intervention values ​​in the high-uncertainty dimension, thus forming a basis for comparison; checking whether the parameter values ​​of other nutritional dimensions, excluding the high-uncertainty dimension, are strictly consistent among different hypotheses to ensure the univariate comparison principle of the experiment; and verifying that the statements of all hypotheses are logically consistent and based on the same control conditions. The set of hypotheses that pass the verification constitutes a set of internally rigorous comparative nutritional intervention hypotheses that can be used for subsequent experiments.

[0184] This embodiment achieves personalized customization of the comparative experimental protocol by selecting candidate values ​​from a preset parameter space and automatically generating a complete combination of nutritional intervention parameters that meets the user's individual conditions and health goals using a constraint satisfaction algorithm. By transforming the parameter combination into a clear scientific hypothesis and performing logical consistency verification, the generated comparative experiment is ensured to possess scientific rigor and feasibility. This embodiment transforms the exploration requirements with high uncertainty into specific, safe, and operable comparative nutritional intervention hypotheses, providing core input for the construction of subsequent experimental protocols.

[0185] In some embodiments, the comparative nutritional intervention hypothesis is transformed into an executable dietary template containing controllable variables, including:

[0186] Each comparative nutritional intervention hypothesis is analyzed into specific daily nutrient allocation rules, food category selection rules, or meal timing rules.

[0187] Based on daily nutrient allocation rules, food category selection rules, or meal timing rules, a structured dietary framework is generated. The dietary framework defines the nutrient variables that need to be controlled and their target values, but does not specify specific food items.

[0188] The dietary framework is matched with a general food composition database to generate an initial recipe draft that includes specific foods, portion sizes, and meal schedules.

[0189] The initial recipe draft undergoes safety and feasibility verification. Safety verification includes checking whether it complies with dietary restrictions in the user's multidimensional data, while feasibility verification includes checking whether key ingredients are basically available within the preset geographical and time range.

[0190] The initial recipe draft that passes verification will be identified as an executable diet template.

[0191] In this embodiment, analyzing the combination of nutritional intervention parameters involves translating numerical targets into operational instructions that can guide daily dietary behaviors. Daily nutrient allocation rules specify the proportions of total daily calories and macronutrients (carbohydrates, protein, fat), and can be refined to recommended intake ranges for specific micronutrients. Food category selection rules specify the categories of primary food sources (e.g., emphasizing whole grains, limiting red meat, and increasing deep-sea fish). Meal timing rules set meal frequency, time windows, or the allocation of specific nutrients between meals (e.g., carbohydrates later in the meal cycle).

[0192] When generating a structured dietary framework, the framework organizes the above rules in a machine-readable data structure (such as JSON or a specific schema), clearly listing all nutritional variables that need to be controlled (such as total calories, protein grams, and the proportion of carbohydrates at breakfast), their target values ​​or ranges, as well as related non-nutritive rules (such as meal times). At the same time, it marks the core comparison variables (i.e., controllable variables) of this experiment, as well as the background variables that need to be kept consistent.

[0193] The process of matching a dietary framework with a general food composition database can be viewed as a constraint satisfaction and optimization problem. Taking linear programming as an example, the specific matching process is as follows: The target values ​​of nutritional variables in the dietary framework are used as linear constraints; the nutrient content per unit weight of each food in the food composition database is used as coefficients; the decision variable is defined as the grams of each food consumed in the recipe; optimization objectives are set, such as minimizing the deviation from user taste preferences or minimizing total cost; the linear programming problem is solved to obtain a set of foods and their specific portions that satisfy all nutritional constraints and have a relatively good optimization objective; the food portions are allocated to each meal according to meal rules (such as the ratio of three meals), forming an initial recipe draft containing specific foods, precise portions, and meal arrangements.

[0194] When performing safety checks on the initial draft recipe, the system compares the food list in the recipe with dietary restrictions (such as allergens, religious restrictions, and disease-related restrictions) recorded in the user's multidimensional data, marking and removing or replacing foods containing prohibited ingredients. Feasibility checks may call interfaces for supply chain status data or, based on a pre-set geographical and seasonal knowledge base, determine whether key ingredients in the recipe are routinely available in the user's current geographical location and during the experiment's execution time. If an ingredient is determined to be extremely difficult to obtain, it is marked as infeasible.

[0195] The initial draft diet plan that passes validation becomes the final executable diet template, providing users with a daily diet plan that they can directly follow without having to calculate or combine the food themselves. At the same time, it ensures the consistency of the plan with the original comparative nutritional intervention hypothesis on core nutritional variables.

[0196] This embodiment transforms scientific hypotheses into specific rules and constructs a structured dietary framework, achieving the conversion from scientific hypotheses to dietary guidelines. By introducing optimization algorithms such as linear programming for recipe generation, specific food plans are automatically generated while meeting complex nutritional constraints, improving efficiency and scientific rigor. Rigorous safety and feasibility checks ensure the personalized adaptability and practical operability of the generated dietary templates, laying a solid foundation for the successful execution of environment-adaptive nutritional experimental protocols.

[0197] In some embodiments, environmental dynamic data is integrated with an executable diet template, and environmental adaptive adjustment rules are added to the executable diet template to form an environmentally adapted nutritional experimental protocol, including:

[0198] Based on real-time environmental monitoring data, identify the types and intensities of current and predicted future environmental pressures;

[0199] Based on the type and intensity of environmental stress, at least one general environmental adjustment rule applicable to the executable diet template is matched from the preset environmental-nutrition adjustment rule library;

[0200] General environmental adjustment guidelines include recommendations for adjusting water intake, electrolyte intake, food texture, or cooking methods for specific environmental conditions;

[0201] By logically binding general environmental adjustment rules with executable dietary templates, an environmentally adapted nutritional experimental protocol is generated.

[0202] In environmentally adapted nutrition experimental protocols, the implementation of general environmental adjustment rules is configured to not affect the comparison of core variables between at least two comparative nutritional intervention hypotheses.

[0203] In this embodiment, identifying the type and intensity of environmental stress involves parsing and classifying real-time environmental monitoring data streams. Environmental stress types can be categorized based on monitoring parameters; for example, high temperature and humidity belong to heat stress, low temperature and dryness belong to cold stress, and high concentrations of particulate matter belong to pollution stress. Intensity is determined by comparing monitored values ​​with predefined grading thresholds, such as classifying temperatures into mild, moderate, and severe high-temperature levels.

[0204] The pre-defined environment-nutrition adjustment rule base is a structured knowledge base or database that stores multiple rules. Each rule typically includes the correspondence between triggering conditions and environmental stress types and intensity ranges, as well as the corresponding adjustment actions. The adjustment actions specifically describe modification suggestions for water intake, electrolyte intake, food texture, or cooking methods.

[0205] Matching applicable general environmental adjustment rules from the rule base can be viewed as a rule-based reasoning process. Taking a production system as an example, the specific matching process is as follows: the identified current and predicted environmental stress types and intensities are matched with the triggering conditions of each rule in the rule base; if the current environmental condition meets the triggering conditions of a rule, that rule is activated; all activated rules are collected, and the set of adjustment actions of these rules constitutes the matched general environmental adjustment rules applicable to the current executable dietary template.

[0206] The system logically binds general environmental adjustment rules to executable diet templates. Within the experimental protocol's data structure, one or more condition-action pairs are associated with each diet template. As the protocol is executed, the system continuously monitors environmental data. Once environmental conditions meet the criteria of a certain rule, the corresponding adjustment action is automatically triggered, and the adjustment suggestions are integrated into the final guidance provided to the user that day.

[0207] All general environmental adjustment rules are limited to adjusting only those parameters that are not part of the core nutrient variables or eating time rules defined by the comparative nutritional intervention hypothesis. This ensures that the implementation of general environmental adjustment rules does not affect the comparison of core variables. For example, if the core variable of the comparison hypothesis is the carbohydrate distribution ratio between breakfast and dinner, the environmental adjustment rules may suggest increasing daily water intake or adjusting the way vegetables are cooked at lunch, but will not change the total amount or ratio of carbohydrates at breakfast or dinner.

[0208] This embodiment injects dynamic environmental responsiveness into static dietary templates by establishing an environment-nutrition adjustment rule base and achieving real-time rule matching and binding based on environmental states. By clearly defining the scope of application of the environmental adjustment rules, it ensures that the comparative validity of the experiment is not affected by the environmental adaptation process. This embodiment enables nutritional experimental protocols to intelligently adapt to external environmental fluctuations, maintaining the scientific rigor of the experiment while improving the feasibility of the protocol in real and complex environments and the user's ability to maintain physiological homeostasis.

[0209] Please see Figure 5In a second aspect, this embodiment also provides a personalized nutrition recommendation system 1 based on dynamic environmental adaptation, applicable to the method described in the first aspect. The system includes a data acquisition module 11, a user-environment modeling module 12, an uncertainty assessment module 13, an experiment generation module 14, an environment adaptation module 15, a dynamic execution module 16, and a result analysis module 17. The data acquisition module 11 is used to acquire multidimensional user data and dynamic environmental data. The multidimensional user data includes static user attribute data, dynamic time-series physiological data, and subjective user feedback data. The dynamic environmental data includes real-time environmental monitoring data and supply chain status data. The user-environment modeling module 12 is used to construct a user-environment joint state vector based on the multidimensional user data and use a time-series analysis model to identify the correlation between user physiological patterns and environmental fluctuations, generating a personalized environment-physiological response baseline. The uncertainty assessment module 13 is used to identify at least one high-uncertainty dimension of the user's nutritional response based on the personalized environment-physiological response baseline and the multidimensional user data. The experiment generation module 14 is used to generate an experiment based on at least one high-uncertainty dimension. The experiment involves formulating at least two comparative nutritional intervention hypotheses and transforming them into executable dietary templates containing controllable variables. An environmental adaptation module 15 integrates dynamic environmental data with the executable dietary templates, adding environmental adaptive adjustment rules to form an environmentally adapted nutritional experimental protocol. A dynamic execution module 16, during the user's execution of the environmentally adapted nutritional experimental protocol, recommends food substitutes that meet nutritional and availability requirements based on real-time supply chain status data via a dynamic food substitution engine. It also triggers environmental compensatory fine-tuning of the execution parameters of the environmentally adapted nutritional experimental protocol based on real-time environmental monitoring data. A results analysis module 17 collects user physiological outcome data and user subjective feedback data during the execution of the environmentally adapted nutritional experimental protocol. Combined with concurrently acquired dynamic environmental data, it controls environmental confounding factors through a causal effect estimation module, analyzes the impact of at least two comparative nutritional intervention hypotheses on user physiological outcomes, and generates personalized nutritional decision-making rules.

[0210] In this embodiment, the data acquisition module 11 connects to sensors, user input interfaces, and external data sources via an application programming interface (API), and is responsible for data collection and initial formatting. The user-environment modeling module 12 integrates a time-series analysis model to train and infer the fused data. The uncertainty assessment module 13, experiment generation module 14, environment adaptation module 15, dynamic execution module 16, and result analysis module 17 are sequentially connected, forming an automated processing pipeline from state perception, experiment design, dynamic execution to effect attribution. The dynamic ingredient replacement engine and causal effect estimation module, as core sub-units of the dynamic execution module 16 and the result analysis module 17, respectively handle the dynamic interference of the real-time supply chain and environment, and the causal effect estimation after controlling for confounding factors. This system achieves full-process automation from data perception, intelligent experiment design, dynamic adaptive execution to accurate causal inference, solving the problems of difficulty in removing environmental confounding effects and static rigidity of recommendation strategies in personalized nutrition recommendations.

[0211] By adopting the above technical solutions, this invention differs from existing technologies and possesses the following beneficial effects: It quantifies the individualized impact of the environment on user physiology by constructing a personalized environment-physiological response baseline, and based on this, identifies the high uncertainty dimension of user nutritional responses, proactively generating comparative nutritional intervention hypotheses and executable dietary templates; by integrating dynamic environmental data with dietary templates to form an environment-adaptive nutritional experimental plan, and utilizing a dynamic ingredient replacement engine to address supply chain fluctuations and triggering environmental compensatory fine-tuning to cope with external environmental mutations during plan execution, it effectively ensures the feasibility and data quality of the experiment under real and complex conditions; finally, through a causal effect estimation module, it uses dynamic environmental data as a covariate to control environmental confounding factors, achieving a pure estimate of the net effect of the comparative nutritional intervention hypothesis, and generating conditional personalized nutritional decision-making rules. The above technical solutions transform personalized nutritional recommendations from static fitting based on historical correlations to a proactive, dynamically adaptable scientific experiment and causal inference process capable of controlling environmental confounding, significantly improving the accuracy, reliability, and practical value of personalized nutritional decisions in dynamic environments.

[0212] Finally, it should be noted that although the above embodiments have been described in the text and drawings of this application, this should not limit the scope of patent protection of this application. Any technical solutions that are based on the essential concept of this application and utilize the content described in the text and drawings of this application, resulting in equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this application.

Claims

1. A method of personalized nutrition recommendation based on environmental dynamic adaptation, characterized in that, include: Acquire multidimensional user data and dynamic environmental data. The multidimensional user data includes static user attribute data, dynamic time-series physiological data, and subjective user feedback data. The dynamic environmental data includes real-time environmental monitoring data and supply chain status data. Based on the user's multidimensional data, a user-environment joint state vector is constructed, and a time series analysis model is used to identify the correlation between the user's physiological patterns and environmental fluctuations, generating a personalized environment-physiological response baseline. The user-environment joint state vector is input into the time series analysis model in a time series manner. The time series analysis model is a long short-term memory network model or a Transformer time series model. Based on the personalized environmental-physiological response baseline and the user's multidimensional data, at least one dimension of high uncertainty in the user's nutritional response is identified through the uncertainty assessment module; Based on the aforementioned high uncertainty dimension, at least two comparative nutritional intervention hypotheses are generated, and the comparative nutritional intervention hypotheses are transformed into an executable dietary template containing controllable variables. The environmental dynamic data is integrated with the executable diet template, and environmental adaptive adjustment rules are added to the executable diet template to form an environmentally adapted nutritional experimental scheme. During the execution of the environmentally adapted nutrition experiment, based on the real-time acquired supply chain status data, a dynamic ingredient replacement engine recommends food substitutes that meet the nutritional and availability requirements for the environmentally adapted nutrition experiment. Based on the real-time acquired environmental monitoring data, environmental compensation fine-tuning of the execution parameters of the environmentally adapted nutrition experiment is triggered. Collect user physiological results data and user subjective feedback data during the execution of the environmentally adapted nutrition experimental protocol, and combine them with the environmental dynamic data obtained at the same time. Control the environmental confounding factors through the causal effect estimation module, analyze the impact of the comparative nutrition intervention hypothesis on user physiological results, and generate personalized nutrition decision rules. Specifically, based on the personalized environment-physiological response baseline and the user's multidimensional data, at least one high-uncertainty dimension of the user's nutritional response is identified through the uncertainty assessment module, including: Historical nutrient intake sequences and corresponding historical physiological response sequences are extracted from the user's multidimensional data. The historical nutrient intake sequences contain intake data for multiple nutrient dimensions, and the historical physiological response sequences contain data for multiple physiological indicators. The historical physiological response sequence is input into the personalized environment-physiological response baseline, and combined with the environmental dynamic data of the same period, the environmentally corrected historical physiological response sequence is calculated. The environmentally corrected historical physiological response sequence has removed the baseline influence of environmental fluctuations on physiological indicator data. Multivariate association analysis was performed on the environmentally corrected historical physiological response sequence and the historical nutrient intake sequence to calculate the association strength measure between each nutrient dimension and each physiological indicator. Based on the correlation strength metric, a nutrition-physiological response uncertainty map is constructed, wherein the nutrition-physiological response uncertainty map is plotted with the nutrition dimension as the horizontal axis, the physiological index as the vertical axis, and the correlation strength metric as the map value. Clustering algorithms are used to divide the nutritional-physiological response uncertainty map into regions, and regions with low association strength whose association strength metric values ​​are below a preset threshold are identified. At least one nutritional dimension corresponding to the region with low correlation strength is identified as at least one high uncertainty dimension of the user's nutritional response.

2. The method for personalized nutrition recommendations based on environmental dynamics adaptation according to claim 1, characterized in that, During the user's execution of the environment-adaptive nutrition experiment, based on the real-time acquired supply chain status data, a dynamic ingredient replacement engine recommends ingredient substitutes that meet the nutritional and availability requirements for the environment-adaptive nutrition experiment. Furthermore, based on the real-time acquired environmental monitoring data, it triggers environmentally compensated fine-tuning of the execution parameters of the environment-adaptive nutrition experiment, including: When the preset ingredients in the environmentally adapted nutrition experiment plan are unavailable, the set of locally available alternative ingredients is queried based on the real-time acquired supply chain status data. Calculate the nutritional similarity between each alternative food in the set of alternative food and the preset food, and evaluate the environmental adaptability of each alternative food in combination with the environmental context indicated by the current real-time environmental monitoring data; Based on the nutritional similarity and environmental adaptability, a target alternative food is determined from the set of alternative food ingredients, and the recommended intake of the target alternative food in the environmentally adapted nutrition experimental protocol is automatically adjusted. The system continuously monitors the real-time environmental monitoring data. When the monitored environmental parameter value exceeds the preset environmental parameter threshold range corresponding to the personalized environment-physiological response baseline, it generates an environmental compensatory adjustment instruction for at least one execution parameter in the environmentally adapted nutrition experimental protocol. The execution parameter includes at least one of water intake, electrolyte supplementation, and food cooking method. The environmental compensatory adjustment instruction is pushed to the user. The environmental compensatory adjustment instruction is used to compensate for the potential interference of environmental changes on the user's physiological state while maintaining the comparative effectiveness of the environmentally adapted nutrition experimental protocol.

3. The method for personalized nutrition recommendations based on environmental dynamics adaptation according to claim 1, characterized in that, During the implementation of the environmentally adapted nutrition experimental protocol, user physiological outcome data and user subjective feedback data are collected and combined with the environmental dynamic data acquired concurrently. By controlling for environmental confounding factors through a causal effect estimation module, the impact of the comparative nutritional intervention hypothesis on user physiological outcomes is analyzed, and personalized nutrition decision-making rules are generated, including: The user physiological results data, user subjective feedback data, and environmental dynamic data collected during the execution of the environmentally adapted nutrition experimental protocol are aligned and divided according to the execution stage corresponding to the comparative nutrition intervention hypothesis to form at least two comparative datasets. Based on the aforementioned comparison dataset, a causal inference model is constructed, wherein the environmental dynamic data serves as the covariate input to the causal inference model. Using the causal inference model, after controlling for the influence of environmental confounding factors represented by the environmental dynamics data, the net effect of the comparative nutritional intervention hypothesis on the user's physiological outcome data is estimated, and the statistical significance of the net effect is calculated. Based on the net effect and the statistical significance, the superior hypothesis with better physiological results is determined among the comparative nutritional intervention hypotheses. The superior hypothesis is the hypothesis whose net effect points to a beneficial physiological change and whose statistical significance measure is better than a preset statistical standard. Extract the core nutritional intervention parameters from the environmentally adapted nutritional experimental protocol corresponding to the winning hypothesis, and combine them with the influence weights of the environmental dynamic data in the causal inference model to generate the conditional personalized nutritional decision rules. The personalized nutrition decision-making rules include core nutrition intervention parameters, applicable environmental conditions range, and additional adjustment suggestions when the environmental conditions exceed the applicable environmental conditions range.

4. The method for personalized nutrition recommendations based on environmental dynamics adaptation according to claim 1, characterized in that, Based on the aforementioned multidimensional user data, a user-environment joint state vector is constructed, including: Sliding window sampling and statistical feature extraction are performed on the user's dynamic time-series physiological data to obtain the user's dynamic physiological feature vector; Perform one-hot encoding or numerical normalization on the user static attribute data to obtain the user static feature vector; Perform text sentiment analysis or numerical rating conversion on the user subjective feedback data to obtain the user subjective state feature vector; The user's dynamic physiological feature vector, the user's static feature vector, and the user's subjective state feature vector are concatenated to generate the user's physiological state feature vector. The real-time environmental monitoring data is time-aligned and numerically normalized to obtain an environmental context feature vector. The user's physiological state feature vector is concatenated with the environmental context feature vector to generate the user-environment joint state vector; The user-environment joint state vector is used to characterize the comprehensive system state under the combined influence of the user's physiological state and external environmental pressure at a specific point in time.

5. The method of claim 4, wherein the method further comprises: Using time-series analysis models, the correlation between user physiological patterns and environmental fluctuations is identified, generating personalized environmental-physiological response baselines, including: The temporal dependency features of the user-environment joint state vector are extracted using the Long Short-Term Memory Network model or the Transformer temporal model. Calculate the cross-attention weight or hidden state correlation degree between the environmental context features and the user physiological state features based on the temporal dependency features; Based on the cross-attention weights or hidden state correlation, a gradient analysis algorithm is used to calculate the change in the predicted output of the user's physiological state features caused by a unit change in the environmental context features. The predicted output change is fitted in the form of a parameterized function, which takes environmental dynamic data as input and the baseline predicted value of the user's physiological state characteristics as output, to form the personalized environment-physiological response baseline.

6. The method for personalized nutrition recommendations based on environmental dynamics adaptation according to claim 1, characterized in that, Based on the aforementioned high uncertainty dimension, at least two comparative nutritional intervention hypotheses are generated, including: For each dimension with high uncertainty, at least two different parameter values ​​are selected from the pre-defined nutritional parameter space as candidate intervention values; Based on the user static attribute data and health goal data in the user multidimensional data, at least two nutritional intervention parameter combinations are generated using a constraint satisfaction algorithm. Each nutritional intervention parameter combination uses different candidate intervention values ​​in the high uncertainty dimension, while keeping the parameter values ​​consistent in other nutritional dimensions. Each combination of nutritional intervention parameters is transformed into a verifiable comparative nutritional intervention hypothesis, which states that, under the control of other nutritional and environmental variables, the combination of nutritional intervention parameters will produce better physiological outcomes for users. Logical consistency checks are performed on at least two generated comparative nutritional intervention hypotheses to ensure that the at least two comparative nutritional intervention hypotheses form a valid comparative relationship on the high uncertainty dimension.

7. The method for personalized nutrition recommendations based on environmental dynamics adaptation according to claim 1, characterized in that, Transforming the comparative nutritional intervention hypothesis into an executable dietary template containing controllable variables includes: Each of the comparative nutritional intervention hypotheses is combined with its corresponding nutritional intervention parameters and analyzed into specific daily nutrient allocation rules, food category selection rules, or eating time rules. Based on the daily nutrient allocation rules, food category selection rules, or eating time rules, a structured dietary framework is generated. The dietary framework defines the nutrient variables to be controlled and their target values, but does not specify specific food items. The dietary framework is matched with a general food composition database to generate an initial recipe draft that includes specific foods, portion sizes, and meal schedules; The initial recipe draft is subjected to safety and feasibility verification. The safety verification includes checking whether it complies with the dietary restrictions in the user's multidimensional data. The feasibility verification includes checking whether the key ingredients are basically available within a preset geographical and time range. The basic availability refers to whether the key ingredients in the recipe are among the commonly available categories in the user's current geographical location and the time period of the experiment. The initial recipe draft that has passed verification will be determined as the executable diet template.

8. The method for personalized nutrition recommendations based on environmental dynamics adaptation according to claim 1, characterized in that, The environmental dynamic data is integrated with the executable diet template, and environmental adaptive adjustment rules are added to the executable diet template to form an environmentally adapted nutritional experimental protocol, including: Based on the real-time environmental monitoring data, identify the types and intensities of current and predicted future environmental pressures; Based on the type and intensity of environmental stress, at least one general environmental adjustment rule applicable to the executable diet template is matched from the preset environmental-nutrition adjustment rule library; The general environmental adjustment rules include recommendations for adjusting water intake, electrolyte intake, food texture, or cooking methods for specific environmental conditions; The general environmental adjustment rules are logically bound to the executable diet template to generate the environmentally adapted nutritional experimental protocol. In the aforementioned environment-adaptive nutrition experimental protocol, the execution of general environmental adjustment rules is configured to not affect the comparison of core variables between the at least two comparative nutritional intervention hypotheses.

9. A personalized nutrition recommendation system based on environmental dynamic adaptation, characterized in that, The system applicable to the method of any one of claims 1 to 8, the system comprising: The data acquisition module is used to acquire multidimensional user data and dynamic environmental data. The multidimensional user data includes static user attribute data, dynamic time-series physiological data, and subjective user feedback data. The dynamic environmental data includes real-time environmental monitoring data and supply chain status data. The user-environment modeling module is used to construct a user-environment joint state vector based on the user's multidimensional data, and to use a time series analysis model to identify the correlation between the user's physiological patterns and environmental fluctuations, thereby generating a personalized environmental-physiological response baseline. An uncertainty assessment module is used to identify at least one dimension of high uncertainty in the user's nutritional response based on the personalized environmental-physiological response baseline and the user's multidimensional data. An experiment generation module is used to generate at least two comparative nutritional intervention hypotheses based on the at least one high uncertainty dimension, and to transform the at least two comparative nutritional intervention hypotheses into an executable dietary template containing controllable variables. An environment adaptation module is used to integrate the dynamic environmental data with the executable diet template, and to add environmental adaptive adjustment rules to the executable diet template to form an environment-adaptive nutritional experimental plan. The dynamic execution module is used to recommend food substitutes that meet the nutritional and availability requirements of the environmentally adapted nutrition experiment plan based on the real-time acquired supply chain status data during the user's execution of the environmentally adapted nutrition experiment plan, and to trigger environmental compensation fine-tuning of the execution parameters of the environmentally adapted nutrition experiment plan based on the real-time acquired environmental monitoring data. The results analysis module is used to collect user physiological results data and user subjective feedback data during the execution of the environment-adaptive nutrition experimental protocol, and combine them with the environmental dynamic data acquired at the same time. By controlling environmental confounding factors through the causal effect estimation module, the module analyzes the impact of the at least two comparative nutrition intervention hypotheses on user physiological results and generates personalized nutrition decision rules.

Images