CN105631211A - Multi-objective nutrition optimization and decision support system - Google Patents

Abstract

The invention discloses a multi-objective nutrition optimization method and decision support system. The method is based on a fuzzy mathematic membership degree function and is used for analyzing various nutrients. A multi-objective optimization target model with a weight and overall evaluation indexes is put forward. The optimal solution of the overall evaluation indexes is obtained, and the optimal individual and overall nutrient scheme is obtained. Meanwhile, the disclosed decision support system comprises one or more client subsystems and a computer subsystem. Users input initial values of the nutrients through the client subsystems, and the optimal individual and overall nutrient scheme is obtained through computation of the computer subsystem.

Description

A kind of multiple goal nutrition optimum decision supporting system

Technical field

The invention belongs to trophic analysis and guidance technology field, particularly relate to a kind of multiple goal trophic model, optimization method and decision support system (DSS).

Background technology

Dietary nutrition is the basis of human survival, and human body all needs to obtain a certain amount of various required nutritive ingredient from meals every day, to meet the biochemical reaction of a series of complexity in vital process. If certain nutrient substance is too much or not enough in institute's food intake, nutrient imbalance will be produced. Whether the various nutrient substances how instructing and judging human body to take in are reasonable, it is necessary to set up a set of trophic model and be trophic analysis, instruct offer foundation.

Mostly the nutritional support of current clinical a large amount of use is based on experimental formula, these have an appointment more than 160 based on the clinical formula of sex, age, height and body weight, and wherein Harris-Benedict formula (being called H-B formula) is used at most at professional Medical Degree. For the same patient being in the various disease stage, the energy consumption estimated by formula be comparatively speaking one quantitatively, using estimated value as nutritional support according to having problem. These formula all do not consider other factor that may affect energy requirement, such as the impact of endocrine regulation, gland body (such as Tiroidina, suprarenal gland) diacrisis, stress situation (heating, wound, Psychic stress state), moving situation, Body components state and multiple factor such as the antagonism relationship between nutraceutical matrix, nutrient substance, sleep, clinical care, medicine, morbid state, neurohumor and cytokine and environment etc., this makes to there is very big deviation between estimated value and actual demand. There are some researches show clinical upper employing H-B formula add correction factor after estimated value have 1/3 to be quite coarse. So just having more than 1/3 relies on H-B formula to carry out patient's malnutrition or the overnutrition of nutritional supplementation, and this hides serious medical consequences clinically.

The nutrition curve analytical procedure that current clinical maternal nutritional analytical system uses, because combining individual hoc scenario, can more accurately analyze the demand of individual each nutrient substance, but also fail to solve how comprehensive individual (comprising the multiple factor affecting human nutrition demand of morbid state, living environment, food habits etc.) makes the various nutrients comprising energy not reach the optimum regime of individual nutrition demand entirety at present based on nutrition curve analytical procedure.

Summary of the invention

For the problem of prior art, the present invention proposes a kind of multiple goal trophic model and optimization method, on nutrition curve analytical procedure basis, comprehensive Individual Age, height, body weight, sex, moving situation, environment, individual body tissue situation, living habit, food habits, disease condition, metabolic condition, short of money anti-between nutrient substance, proportionlity etc., again the multiple goal trophic model of a belt restraining condition is established, and adopt suitable optimization method, for individuality provides overall Nutritional solutions, further for dietary nutrition science decision lays the first stone.

For obtaining individual nutrition demand global optimization scheme better, the present invention establishes a multiple objective function model based on the multifactor constraint condition of nutrition curve analytical procedure. When setting up the multiple objective function model of multifactor constraint condition, not only consider the proportionlity between the effect between nutrient substance (synergy, antagonistic action), nutrient substance, also fully consider living habit, take medicine, environmental factors, individual primary metabolic condition, body tissue's composition, morbid state, different physiological periods (as: old age, pregnancy period, teenager etc.) etc. are on the special impact of some nutrient substance. All nutritions involved in model in principle are all the targets needing to optimize, such as total energy, protein, fat, carbohydrate, calcium, potassium, iron, zinc, phosphorus, sodium, various VITAMIN, food fibre etc.; Synergy between proportionlity between nutrient substance, nutrient substance, antagonistic action, living habit, take medicine, environmental factors etc. to the special impact of some nutrient substance, and different physiological period is to the specific demand of some nutrient substance. Individual nutrition needs of problems is real be complexity, a huge multiple factor constraints condition and Multi-objective Decision Model. Need multiple factor and the constraint condition of fully considering to affect individual nutrition demand, and set up a multiple objective function and model, the overall optimal solution of the individual nutrition obtained by suitable optimization algorithm.

Based on above-mentioned model and algorithm, the present invention also develops a set of multiple goal nutrition optimum decision supporting system, user by input system according to sex, age, meals situation, motion conditions, living habit, take medicine, environmental factors, different physiological period, measure in conjunction with metabolism monitoring, body tissue, the multiple objective function model parameter such as monitoring of motion and disease, by systems analysis, user finally obtains the nutritional programs after global optimization.

Accompanying drawing explanation

Fig. 1: the structure block diagram of decision support system (DSS) of the present invention.

Embodiment

Nutrition curve function and mathematical model, based on existing nutrition curve analytical procedure, have been improved by the present invention, it is proposed that model is optimized in the multiple goal nutrition based on fuzzy mathematics membership function belt restraining condition. This model comprises 25 kinds of nutrient substances of energy, adopts the membership function in fuzzy mathematics to express the corresponding relation between the intake of various nutrient substance and individual health situation, if x_iIt is the daily intaking amount (1��i��25) of i-th kind of nutrient substance, it is subordinate to curvilinear function to represent this corresponding relation with following L-R type:

Wherein, R is real number field, function fuzzy permission interval (0, be divided into three sections on c), namely (0, a), [a, b), [and b, c), a and b is predetermined segmentation value, it is preferable that, for different x_i, there is different a/b/c values according to metabolism monitoring. Function F_LAnd F_RBeing left piecewise function and right piecewise function, both need not to be symmetrical, and those skilled in the art can according to concrete nutrient substance and individual health situation design F_LAnd F_R, in this area, more existing concrete method of design, repeat no more herein. The described curvilinear function that is subordinate to is a regular convex function.

It is subordinate to curve by above-mentioned, establishing multiple goal nutrition and optimize model, the n kind nutrient substance of the individual overall nutritional status of impact can regard n the objective function (mainly consider current Chinese food here and become the 25 kinds of macronutrient comprising energy listed in submeter 2004) of multiobjectives decision as. All the other affect multifactor (short of money anti-between age, height, body weight, sex, moving situation, environment, individual body tissue situation, living habit, food habits, disease condition, metabolic condition, nutrient substance of individual nutrition changes in demand, proportionlity etc.) constraint condition of then conversion cost model, when all nutrients be subordinate to curvilinear function and after constraint condition determines, what the acquisition of nutritional needs optimum value had just changed into the optimal value of multiple goal asks for problem.

In order to ask for the optimal value of multiple goal, it is possible to ask for its efficient solution, i.e. Pareto optimum solution. Model is optimized in multiple goal nutrition for the present invention, will find its efficient solution, it is necessary to set an evaluation index to efficient solution, and along with the virtual value of nutrient substance meeting constraint condition is the closer to optimum value, this index is also closest to maximum value or minimum value. According to first embodiment of the invention, adopt P below as the total appraisal index of nutritional status, that is:

$P=\max \[{\mathrm{\φ}}_{{\stackrel{\‾}{A}}_i}\left(x_i\right)\]/\[\frac{1}{n-1}\Σ\frac{1}{{\mathrm{\φ}}_{{\stackrel{\‾}{A}}_i}\left(x_i\right)}\]$

In upper formulaReferring to actual average daily fuzzy value corresponding to intake in the nutrition curve of i-th kind of nutrient substance, n is the species number of the nutrient substance needing consideration in meals. Compensation between target value in usual multiobjectives decision problem, it is between best and the worst target value, obtain certain equilibrium, average operator often is used to realize this conversion, but compared with the average (arithmetic equal value, geometric mean and harmomic mean) that intake is recommended in the reference of Chinese residents nutrient substance, P is closer to minimum value, and it characterizes logic "AND"; IfThen P �� 0, when illustrating that use P evaluates dietary nutrition situation, the overall nutritional status of individuality can be had very strong impact by the nutrient substance of low degree of being subordinate to, and can objectively respond individual actual nutritional status; In proximate calculation, P is convex function, can obtain optimum value, can judge whether the nutritional status of a kind of diet formula is better than the nutritional status of another kind of diet formula. Therefore P as the total appraisal index of individual nutrition situation, after the nutrition curve of various nutrient substance is determined, can judge the quality of the current nutrition intake situation of this individuality relatively easily by P value.

25 objective functions that the multiple goal nutrition of the present invention is optimized in model are asymmetric fuzzy membership curvilinear function, it is possible to adopt cut set method to solve. In cut set method, each nutrient substance x in nutrition curve_iInterval (0, degree of being subordinate to �� (x c) is allowed for fuzzy_i), represent " satisfaction " to fuzzy constraint. �� (x_iDuring)=0, this constraint condition is not met, �� (x_iDuring)=1, this constraint condition is strictly met, as 0 < �� (x_i) < when 1, this constraint portions is met.

Use a_mRepresent x_mTo the satisfaction of fuzzy constraint, it is designated as: a_m=�� (x_m) fuzzy permission interval (0, in c), degree of being subordinate to �� (x_mThe interval formation of) >=�� is the �� horizontal cut set on number theory territory.

To the demand difference of Different Nutrition element under proportionlity between this model nutrition, various state, the such as pregnancy period is not only different from common woman to demands such as calcium, iron, zinc, protein, food fibre, iodine, energy, folic acid, also along with the change in pregnant week, the demand of some nutrient substance also can be changed by same person, and these factors all need to pay attention in optimization model and optimizing process. The 25 kinds of nutritions comprising energy are all the targets that needs are optimized in this model, and the blend proportion between them is also the target needing to optimize. In addition individual residing state comprise the disease (obesity, diabetes, hypertension etc.) relevant to nutrition intake, environmental factors, living habit (stay up late, take microbiotic, strenuous exercise, pressure big etc.), food habits (partially oily, partially salty, meat be lead, vegetarian diet be main etc.), shared by the individual required three big macroelements of the different steps (pregnancy, lactation, old man, teenager etc.) of physiological status, the ratio of total energy is not quite similar, and the demand of some special nutrition element is also along with individual state has bigger change. Therefore these needs need the content that emphasis is considered when designing model, constraint condition and overall evaluation index.

Although P can be used as the total appraisal index of individual nutrition situation, after the nutrition curve of various nutrient substance is determined, also can be judged the quality of the current nutrition intake situation of this individuality relatively easily by P value. But above-mentioned various factors cannot be taken into account, it is thus desirable to P is redesigned. The 2nd embodiment according to the present invention, it is contemplated that the plain different important degree in target model of Different Nutrition under different situations, for the difference of the situation demand of Different Individual, introduces weights omega_jImproving the total appraisal index of nutritional status, new total appraisal index P ' is as follows:

And $\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j=1$

WhereinIt is that i-th kind of nutrient substance is at the interval A of fuzzy permission_iOn membership function, ��_iBeing its weight, n is the species number of the nutrient substance needing consideration in meals.

For new total appraisal index P ', it is asked for optimum solution, so that it may separate to obtain the optimum absorption of multiple goal nutrient substance. Specific algorithm is as follows:

(1) initialization step: get initial value for n kind nutrient substanceIf iteration error is ��, make loop variable j=0, for each x_i, what have following L-R type is subordinate to curve

It is subordinate to curve according to above-mentioned, obtains at the interval A of fuzzy permission_iIn membership function

(2) by membership functionSubstitute into following global optimization index formula,

Wherein��_iIt is x_iWeighted value, set in advance according to different situations.

(3) j increases by 1, if j=n, forwards step (7) to;

(4) willN value in array arranges from big to small according to weighted value, gets the value coming jth position and (is assumed to be x_i), by x_iValue increase a predetermined step-length value, thus obtain new array, according to described new array, recalculate global optimization index, obtain

(5) ifSet up, forward step (6) to, otherwise forward step (7) to

(6) makeReturn step (3);

(7) judgeWhether setting up, if set up, forwarding step (10) to;

(8) if j=n, then j=0 is made;

(9) step (3) is forwarded to;

(10) x is exported_�ˡ�=[x₁, x₂... x_n]^TAndTerminate.

See Fig. 1, inquiring about to facilitate user and implement nutrition design, the present invention is based on above-mentioned multiple goal nutrition optimization method, it is proposed that a kind of multiple goal nutrition optimum decision supporting system, this system comprises one or more client's subsystem, and a computing subsystem. Wherein client's subsystem receives various nutrients type and the initial value of user's input, send it to computing subsystem, computing subsystem is according to above-mentioned input, the multiple goal nutrition optimization method of the present invention is used to calculate, obtain personalized nutritional scheme, the Output rusults of calculating is returned to client's subsystem.

Below for one 32 years old, height 159 centimetres, body weight 52.45kg, pregnant early stage women. Coordinating its meals to investigate, motion monitoring, body tissue measure, and adopt Model for Multi-Objective Optimization herein and optimize algorithm, obtain following personalized nutritional scheme:

Above-described embodiment of the present invention, does not form limiting the scope of the present invention. Any amendment, equivalent replacement and improvement etc. done within the spirit and principles in the present invention, all should be included within the claims of the present invention.

Claims (3)

1. a multiple goal nutrition optimum decision supporting system, it is characterised in that, this decision support system (DSS) comprises one or more client's subsystem, and a computing subsystem, wherein

Described client's subsystem receives the nutrient categories of described user input, and the initial value of nutrient substanceWherein n is the species number of nutrient substance, x_iIt it is the daily intaking amount of i-th kind of nutrient substance. Then the information that user inputs is sent to described computing subsystem by described client's subsystem;

Described computing subsystem calculates according to the information received, and exports the personalized nutritional scheme of described user, the result of output returns to described client's subsystem, and the detailed process wherein calculated is as follows:

(1) make iteration error be ��, make loop variable j=0, for each x_i, that sets up following L-R type is subordinate to curve,

Wherein, R is real number field, is subordinate to curve at the interval A of fuzzy permission_i=(0, it is divided into three sections on c), a and b is predetermined segmentation value, function F_LAnd F_RIt is left piecewise function and right piecewise function; It is subordinate to curve according to above-mentioned, obtains at the interval A of fuzzy permission_iIn membership function

(2) by membership functionSubstitute into following global optimization index formula, obtain

Wherein��_iIt is x_iWeighted value;

(3) j increases by 1, if j=n, forwards step (7) to;

(4) willN value in array arranges from big to small according to weighted value, the value coming jth position increases a predetermined step-length value, thus obtains new array, according to described new array, recalculate global optimization index, obtain

(5) ifSet up, forward step (6) to, otherwise forward step (7) to;

(6) makeReturn step (3);

(7) judgeWhether setting up, if set up, forwarding step (10) to;

(8) if j=n, then j=0 is made;

(9) step (3) is forwarded to;

(10) export $X_{{\mathrm{\&lambda;}}^{,}}={[X_1,X_2,...X_n]}^T$ AndTerminate.

2. multiple goal nutrition optimum decision supporting system according to claim 1, the nutrient substance wherein inputted is that Chinese food becomes the 25 kinds of macronutrient listed in submeter 2004.

3. multiple goal nutrition optimum decision supporting system according to claim 1-2, being wherein subordinate to curve is a convex function.