Search

CN-120809086-B - AI-based large-client nutrition case rapid evaluation system and method

CN120809086BCN 120809086 BCN120809086 BCN 120809086BCN-120809086-B

Abstract

The invention relates to the technical field of medical information and discloses an AI-based rapid evaluation system and an AI-based rapid evaluation method for large-client nutrition cases, wherein the AI-based rapid evaluation method for large-client nutrition cases comprises the steps of processing crowd statistical feature data by utilizing a meta-learning network to generate a sampling basic model aiming at specific crowds, constructing a Bayesian decision network to calculate nutrition index information value to generate index information cost ratio, constructing a sequential sampling strategy according to the index information cost ratio by adopting a Markov decision process to execute self-adaptive index sampling, calculating information gain ratio to realize dynamic termination decision and avoid over-sampling in the process of executing self-adaptive index sampling, and generating a complete nutrition evaluation result by applying conditional probability inference to realize that only a small number of indexes need to be measured to obtain a high-precision evaluation result.

Inventors

  • XU JIE

Assignees

  • 深圳市前海高新国际医疗管理有限公司

Dates

Publication Date
20260512
Application Date
20250910

Claims (10)

  1. 1. The quick evaluation method of the large-client nutrition case based on the AI is characterized by comprising the following steps: Processing the crowd statistical feature data by using a meta-learning network to generate a sampling basic model for a specific crowd; Based on the sampling basic model, a Bayesian decision network is constructed to calculate the nutritive index information value, an index information cost ratio is generated, the information cost ratio is defined as the ratio of the expected information gain to the product of the economic cost and the time cost, and the information cost ratio is calculated by the following formula: , wherein, Indicating index Is a ratio of the information costs of (a), Indicating the desired information gain of the indicator, Representing measurement indicators Is used for the production of the steel sheet, Representing measurement indicators Time cost of (2); Constructing a sequential sampling strategy by adopting a Markov decision process according to the index information cost ratio, and executing self-adaptive index sampling; in the process of executing the self-adaptive index sampling, calculating the information gain rate to realize dynamic termination decision, so as to avoid over-sampling; Based on the sampling result after the dynamic termination decision, a complete nutrition evaluation result is generated by applying conditional probability inference, and the result reliability estimation is provided.
  2. 2. The AI-based rapid assessment method of large client nutrition cases of claim 1, wherein the step of processing demographic profile data using a meta-learning network comprises: Collecting statistical characteristic data of target crowd, including demographic characteristics, professional characteristics, regional characteristics, historical nutrition status and multidimensional data thereof; Constructing a meta learning network, wherein the meta learning network adopts a double-layer structure, an outer layer network is responsible for learning a mapping function from crowd characteristics to inner layer network parameters, and the inner layer network is responsible for realizing a specific sampling strategy; Training a meta learning network, and adopting a meta learning paradigm based on tasks; And calculating sampling strategy initial parameters of the new crowd through the trained meta-learning network aiming at the new crowd, and generating a sampling basic model suitable for the new crowd.
  3. 3. The AI-based rapid assessment method of large client nutrition cases of claim 1, wherein the step of constructing a bayesian decision network to calculate nutrition index information value comprises: constructing a Bayesian network model to represent probability dependency relationship among nutrition indexes; based on the historical data and expert knowledge, learning the structure and parameters of the Bayesian network; calculating a desired information gain for each potential sampling index to quantify the information contribution of the desired information gain to the nutritional status assessment; the information cost ratio is introduced, and the information value, the economic cost and the time cost of the index are comprehensively considered.
  4. 4. The AI-based rapid assessment method of large client nutrition cases of claim 1, wherein the step of employing a markov decision process to construct a sequential sampling strategy comprises: constructing a Markov decision process model, and forming a sequential sampling problem into a state transition process; an improved Markov blanket information propagation algorithm is realized, and the probability distribution of unmeasured indexes is predicted by effectively utilizing the condition dependency relationship among indexes; Selecting a next optimal measurement index based on the current state and the information cost ratio; Measurements are performed and the status is updated.
  5. 5. The AI-based large client nutrition case rapid assessment method of claim 4, wherein the markov decision process model comprises: a state space representing the state of the current measured index set and its observations; an action space representing a set of indices that may be measured next; A state transition function representing the probability of transition to the next state after selecting a measurement indicator in one state; The reward function, reflecting the net benefit brought by the measurement index, is defined as the information gain of the measurement index minus the measurement cost.
  6. 6. The AI-based rapid assessment method of large client nutrition cases of claim 1, wherein the step of calculating the information gain rate to achieve a dynamic termination decision comprises: defining an information gain rate for measuring marginal information gain caused by newly increasing the measurement index each time; Realizing a dynamic termination algorithm based on a threshold value, setting two termination conditions of an information gain rate threshold value condition and an evaluation accuracy condition, and stopping sampling when any condition is met; Constructing an evaluation accuracy estimation model, and predicting the accuracy of an evaluation result in the current state in real time; and (3) adaptively adjusting a termination threshold value, and dynamically setting termination condition parameters according to the emergency degree of the task and the resource constraint.
  7. 7. The AI-based rapid assessment method of large client nutrition cases of claim 1, wherein the step of applying conditional probability inference to generate complete nutrition assessment results comprises: Deducing the conditional probability distribution of unmeasured indexes by adopting a probability graph model, and deducing complete nutrition index distribution based on the state when sampling is terminated; Combining the conditional probability distribution and expert rules to generate a nutritional status assessment result; Calculating reliability estimation of the evaluation result and quantifying uncertainty of the result; based on the assessment results and the reliability estimate, personalized nutritional advice is generated.
  8. 8. The AI-based rapid assessment method of large-client nutrition cases of claim 7, wherein the reliability estimate of the assessment result is calculated by the following formula: ; wherein, R represents the reliability estimation value of the evaluation result; is shown in the state The normalized conditional entropy of the lower target nutrition state variable Y is used for quantifying the uncertainty of the nutrition state evaluation result in the current sampling state; representing the state at the end of sampling, including all measured metrics and their observations; the normalized conditional entropy scales the original conditional entropy into a [0,1] interval, so that the reliability estimation is more comparable; The value range of R is [0,1], and the larger the value is, the higher the reliability is, namely the lower the uncertainty of the evaluation result is.
  9. 9. The AI-based rapid assessment method of large customer nutrition cases of claim 1, wherein the desired information gain is calculated by the following formula: ; Wherein, the Indicating index Y represents a target nutritional state variable, which is the final nutritional condition to be evaluated; h (Y) represents the entropy of Y, which is used to measure the uncertainty of the target variable; Is shown in the known state Conditional entropy of Y under conditions, representing the index observed Residual uncertainty of the rear target variable Y; Representation of Is the mathematical expectation of (a), i.e. pair of A weighted average of all possible values; indicating by measuring an index A reducible uncertainty, i.e. the amount of information obtained, is expected.
  10. 10. An AI-based rapid assessment system for large client nutrition cases, for performing an AI-based rapid assessment method for large client nutrition cases as claimed in any one of claims 1 to 9, comprising: The crowd characteristic processing module is used for processing the crowd statistical characteristic data by utilizing the meta-learning network and generating a sampling basic model for specific crowd; the information value evaluation module is used for constructing a Bayesian decision network to calculate the nutritive index information value and generate an index information cost ratio; The self-adaptive sampling module is used for constructing a sequential sampling strategy by adopting a Markov decision process and executing self-adaptive index sampling; The dynamic termination decision module is used for calculating the information gain rate to realize the dynamic termination decision of the sampling process; And the evaluation result generation module is used for generating a complete nutrition evaluation result by applying the conditional probability inference and providing a result reliability estimation.

Description

AI-based large-client nutrition case rapid evaluation system and method Technical Field The invention relates to the technical field of medical information, in particular to an AI-based rapid evaluation system and method for nutrition cases of large clients. Background In the nutrition management of large customer groups such as large enterprises, schools and the like, medical institutions need to efficiently evaluate the nutrition condition of the groups, and scientific basis is provided for health management. Traditional nutrition assessment methods require that all detection projects be completed for each member, typically including tens of indicators, resulting in high detection costs, long time consumption and resource limitations. The existing nutrition evaluation technology mainly has the following problems that high-precision nutrition state evaluation cannot be achieved under limited detection indexes, and particularly, individual difference adaptability aiming at different crowds is lacking, the existing sampling method cannot accurately quantify the balance relation between index information value and sampling cost, so that resource allocation is unreasonable, the existing sparse sampling method lacks the adaptability to group differences, the sampling strategy cannot be quickly adjusted according to different crowd characteristics, and a scientific decision method for guiding resource allocation cannot be used for maximizing limited resource value in emergency. Therefore, a method for rapidly evaluating the nutrition cases of large clients, which can overcome the technical problems, is needed, and efficient and accurate nutrition condition evaluation is realized under the condition of limited resources. Disclosure of Invention The invention provides an AI-based rapid evaluation system and method for nutrition cases of large clients, which solve the technical problem that high-precision nutrition evaluation cannot be realized under the condition of limited resources in the related technology. The invention provides an AI-based quick evaluation method for large client nutrition cases, which comprises the following steps: Processing the crowd statistical feature data by using a meta-learning network to generate a sampling basic model for a specific crowd; Based on the sampling basic model, constructing a Bayesian decision network to calculate the nutritive index information value, generating an index information cost ratio, wherein the information cost ratio is defined as the ratio of the expected information gain to the product of the economic cost and the time cost; Constructing a sequential sampling strategy by adopting a Markov decision process according to the index information cost ratio, and executing self-adaptive index sampling; in the process of executing the self-adaptive index sampling, calculating the information gain rate to realize dynamic termination decision, so as to avoid over-sampling; Based on the sampling result after the dynamic termination decision, a complete nutrition evaluation result is generated by applying conditional probability inference, and the result reliability estimation is provided. In a preferred embodiment, the step of processing the demographic data using the meta-learning network comprises: Collecting statistical characteristic data of target crowd, including demographic characteristics, professional characteristics, regional characteristics, historical nutrition status and multidimensional data thereof; Constructing a meta learning network, wherein the meta learning network adopts a double-layer structure, an outer layer network is responsible for learning a mapping function from crowd characteristics to inner layer network parameters, and the inner layer network is responsible for realizing a specific sampling strategy; Training a meta learning network, and adopting a meta learning paradigm based on tasks; And calculating sampling strategy initial parameters of the new crowd through the trained meta-learning network aiming at the new crowd, and generating a sampling basic model suitable for the new crowd. In a preferred embodiment, the step of constructing a bayesian decision network to calculate the nutritional index information value comprises: constructing a Bayesian network model to represent probability dependency relationship among nutrition indexes; based on the historical data and expert knowledge, learning the structure and parameters of the Bayesian network; calculating a desired information gain for each potential sampling index to quantify the information contribution of the desired information gain to the nutritional status assessment; the information cost ratio is introduced, and the information value, the economic cost and the time cost of the index are comprehensively considered. In a preferred embodiment, the step of constructing a sequential sampling strategy using a markov decision process comprises: constructing a Markov decision proce