CN104615892B - A kind of multilayer elite role method excavated for traditional Chinese medical science case history diagnostic rule - Google Patents

Abstract

The invention discloses a multi-layer elite role method for excavating diagnostic rules of special TCM medical records. The method firstly designs elite role subpopulation concentration selection probabilities and assigns TCM medical record attributes to different types of "common-elite" role populations. Carry out preprocessing of correlation and interdependence attributes in special TCM medical records; then construct a dynamic equilibrium strategy based on multi-layer elite roles, and form multi-layer elite roles by dynamically moving in and out of different elite role populations in each pentagonal evolution area Global equilibrium point; finally, the elite subset vector with the strongest global search and local refinement capabilities is selected from the multi-layer elite roles, and the strongest elite optimization array is constructed to realize the rapid mining of special TCM medical record diagnosis rules. The invention can better overcome the problems of fuzzy attributes, attribute correlation and interdependence of special TCM electronic medical records, effectively improve the mining efficiency of its diagnostic rules, and has strong robustness and practicability.

Description

A Multi-layer Elite Role Method for Mining Diagnostic Rules of TCM Medical Records

Technical field:

The invention relates to the field of intelligent processing of TCM information, in particular to a multi-layer elite role method for excavating diagnostic rules of TCM medical records.

Background technique:

As a unique cultural and scientific heritage of the Chinese nation, traditional Chinese medicine has made great contributions to the world. TCM diagnosis is mainly based on the patient's symptoms through the process of "look, smell, ask, and feel". Due to the intricate intersection of TCM syndrome types and symptoms, it is difficult to rely on computer systems for accurate TCM syndrome differentiation and treatment. However, the TCM medical record system stores Collecting a large amount of useful information, extracting the important diagnostic rules of TCM, and carrying out the classification and rule mining of TCM diseases are of great significance and value to improve the diagnosis outcome and outcome analysis of TCM diseases.

Due to the ambiguity and uncertainty of TCM medical records and the existence of a large number of redundant medical record attributes in the TCM medical record database, it is difficult to establish an intelligent auxiliary diagnosis system for TCM medical records to mine disease rules. Especially for some TCM medical records, there are complex correlations and interdependencies among their attributes. It is difficult to dig out really useful diagnostic rules by using the general rule mining method to preprocess and classify TCM medical records. In addition, general rule mining methods are often unable to deal with the multi-dimensional and complex inline relationships among TCM medical record attributes, and cannot guarantee that the final extracted TCM medical record diagnosis rules are the optimal rule feature set for the target. Sometimes supplementary use will interfere with its normal disease judgment, and may lead to misdiagnosis in severe cases. Therefore, under the premise of not affecting the mining accuracy of diagnostic rules, we must seek a rule mining method for TCM medical records that can effectively deal with complex correlation and interdependence attributes.

Invention content:

The present invention overcomes the above deficiencies and provides a multi-layer elite role method for excavating diagnostic rules of medical records of traditional Chinese medicine.

The present invention is achieved through the following technical solutions:

A multi-layer elite role method for mining diagnostic rules of TCM medical records, the specific steps are as follows:

A. Assign the attributes of TCM medical records to different "ordinary-elite" role evolution populations, pre-process the relevant and interdependent attributes in TCM medical records, and design the elite concentration ECD(Elitist _i ) and selection probability P( x _i ), Elitist _i is the i-th elite role population, x _i is the i-th selection of TCM medical records, assigning attribute-related and interdependent TCM medical records to the elite role population, there will be no probability of being selected P( _xi ) The selected general TCM medical record attributes are assigned to the general role population;

B. Construct a multi-layer elite role dynamic equilibrium strategy for mining diagnostic rules of TCM medical records, through the j-th elite in the i-th elite role subpopulation dynamic immigration rate and eviction rate Carry out adaptive migration of elites to achieve the global equilibrium point Mbest of multi-level elite roles;

C. Realize the mining of TCM medical record diagnosis rules based on multi-layer elite role subpopulation vectors, and select n elite subset optimization vectors with the strongest optimization ability from the multi-layer elite role set Realize the global search and local refined mining of n TCM medical record attribute sets MPR ₁ , MPR ₂ ,...,MPR _n , and extract the corresponding TCM medical record attribute subsets MPR ₁ , MPR ₂ ,...,MPR _n respectively The diagnostic rule mining set DRS ₁ , DRS ₂ ,...,DRS _n ;

D. Output the global optimal diagnostic rule set of TCM medical records under the optimization of the strongest elite

The further improvement of the present invention lies in: Step A is specifically: assigning the attributes of the TCM medical records to different "common-elite" role evolution populations, performing preprocessing of the correlation and interdependence attributes in the TCM medical records, and designing the elite concentration of the elite role population ECD(Elitist _i ) and selection probability P( _xi ), Elitist _i is the i-th elite role population, x _i is the i-th selection of TCM medical records, assigning attribute-related and interdependent TCM medical records to the elite role population, The general TCM medical record attributes that are not selected by the selection probability P( _xi ) are assigned to the general role population, and the specific steps are as follows:

a. Calculate the elite role concentration ECD(Elitist _i ) of each elite population as

Where Elitist _i is the i-th elite character population, f(x _j ) is the fitness of the j-th elite in the elite character population, is the average fitness of the i-th elite character population;

b. The selection probability P(x _i ) for constructing the i-th elite role population to select the attributes of TCM medical records is

c. Assign the attribute-related and interdependent TCM medical record attributes in the TCM medical records to the elite role population Elitist _i with the selection probability P( _xi ), and assign the general TCM medical record attributes that are not selected by the selection probability P( _xi ) to In the common character population;

d. Each TCM medical record attribute set assigned to the elite role population Mapped to [0,1] evolutionary population space for normalization processing, the mapping formula is as follows:

where f(x _j ) is the fitness of the jth elite in the elite character population, is the average fitness of the i-th elite role population, f _max and f _min are the maximum fitness and minimum fitness of the i-th elite role population;

if Then MPR _i is 0;

If f(x _j ) and equal, then MPR _i is 1.

The further improvement of the present invention lies in: step B is specifically: constructing a multi-layer elite role dynamic equilibrium strategy for excavating diagnostic rules of medical records of traditional Chinese medicine, through different elite role subpopulation elite dynamic immigration rate and eviction rate Carry out adaptive migration of elites to achieve the global equilibrium point Mbest of multi-layer elite roles. The specific steps are as follows:

a. Divide the evolutionary space into n equal-area pentagons Pe ₁ , Pe ₂ ,...,Pe _n , all evolutionary elite character populations are evenly distributed to different pentagons, and then divide each pentagon Divided into several triangular small evolutionary regions of equal size;

b. Select the elite evolutionary individual with the best fitness in the i-th pentagonal evolutionary area, that is, the elite role And sort them according to the degree of fitness from low to low, so as to form the multi-layer elite role set Er _i in the i-th pentagon area as follows:

Where N _elt is the maximum number of multi-layer elite characters in the i-th pentagon area;

c. Calculate each elite role Dynamic immigration rate in Er _i and eviction rate Calculated as follows:

Among them, I _i is the maximum dynamic migration rate of elite characters in the ith pentagonal area, E _i is the maximum dynamic migration rate of elite characters in the ith pentagonal area, and j is the elite in the multi-layered elite role set Er _i The sequence number of the role arrangement;

d. Multi-layer elite roles focus on each elite role will be based on the rate of immigration and eviction rate Perform dynamic adaptive migration in each adjacent pentagonal area, calculate the average fitness of elite characters in each pentagonal evolution area, and then send it to the population of elite characters in adjacent pentagonal areas, thereby adjusting their dynamic migration Intake rate and eviction rate

e. Repeat the above steps until the elite roles divided in each pentagonal area are no longer migrated, and the overall equilibrium state of multi-layer elite roles is reached;

f. The global equilibrium point Mbest for constructing multi-layer elite roles is as follows:

Through the above-mentioned dynamic balance adjustment based on multi-level elite roles, in each pentagonal space Pe _i , multi-level elite roles will be able to find their respective best immigration rates and eviction rate Then, the dynamic adaptive migration of each elite role is carried out to obtain the global equilibrium point of the elite role in each pentagonal evolutionary area, so that a better population of elite roles can be selected for attribute-related and interdependent TCM diagnosis rules mining.

The further improvement of the present invention lies in: Step C is specifically: realizing the diagnosis rule mining of TCM medical records based on the multi-layer elite role population vector, and selecting n elite subset optimization vectors with the strongest optimization ability from the multi-layer elite role set Realize the global search and local refined mining of n TCM medical record attribute sets MPR ₁ , MPR ₂ ,...,MPR _n , and extract the corresponding diagnostic rule mining sets DRS ₁ , DRS ₂ ,...,DRS _n respectively , the specific steps are as follows:

a. The elite roles in each elite role set Er _i in the global equilibrium point Mbest of multi-layer elite roles are sorted in descending order according to their optimization capabilities, in order

b. Select n elite subset optimization vectors with the strongest optimization ability from the multi-layer elite role set Sequentially assigned to n TCM medical record attribute subsets MPR ₁ , MPR ₂ ,..., MPR _n for global search and mining of diagnostic rules, expressed as:

c. Optimize the elite subset vector with the strongest ability According to the dimensions of the attribute subsets of TCM medical records, they are expanded into n×n dimensional arrays to form the strongest elite optimization array;

d. The strongest elite optimization array containing n dimensions is sequentially subjected to partial refinement mining of diagnostic rules in TCM medical record attribute subsets MPR ₁ , MPR ₂ ,..., MPR _n , and the corresponding TCM diagnostic rule set is extracted as DRS ₁ ,DRS ₂ ,...,DRS _n ;

e. Judging whether the accuracy of the TCM medical record diagnosis rule set DRS _i excavated by the above elite subset vector meets the requirements, if so, then output the global optimal diagnosis rule set of the TCM medical record; if not, repeat the local refinement mining until The accuracy of rule mining meets the requirements.

Compared with the prior art, the present invention has the following advantages:

1. The balance of multi-level elite roles:

In this method, when constructing the multi-level elite roles used for mining diagnostic rules of TCM medical records, the dynamic equilibrium strategy of multi-level elite roles is adopted, and the dynamic adaptive migration of elite roles in the multi-level elite role set Er _i in the pentagonal evolution space is adopted. In and out, to achieve the global equilibrium point Mbest of multi-level elite roles, the multi-level elite roles constructed in this way are more conducive to dealing with TCM medical records with attribute correlation and interdependence, and achieve better diagnosis rule mining results.

2. Global search and local refinement mining optimization performance:

In this method, when realizing the optimization process of mining and optimizing the diagnostic rules of TCM medical records based on the multi-layer elite role subset vector, the three stages of global search mining, local refinement mining and repeated implementation of local refinement mining are respectively optimized, so as to better Dealing with the uncertainty brought about by related and interdependent TCM medical record attributes to the actual mining results is an actual mining effect that cannot be achieved by the current general TCM diagnostic rule mining methods.

3. Strong applicability of diagnostic rules of TCM medical records:

This method can better overcome the problems of fuzzy attributes of special TCM electronic medical records and multi-dimensional and complex inline relationships between medical record attributes, and ensure that the final extracted TCM medical record diagnosis rule set is the optimal feature rule set for the target, effectively improving the quality of TCM. Mining efficiency of diagnostic rules in medical records. The actual clinical application shows that the diagnostic rules of traditional Chinese medicine excavated by this method have good consistency with the diagnostic rules inferred by actual clinicians based on diagnostic experience. This method has strong objectivity and scientificity. Auxiliary diagnosis and TCM clinical decision support analysis will have strong application value.

Description of drawings

Fig. 1 is the general structural diagram of the present invention;

Figure 2 is a schematic diagram of a dynamic balance strategy for multi-level elite roles;

Figure 3 is a schematic diagram of mining diagnostic rules of TCM medical records based on multi-layer elite role subset vectors;

Figure 4 is the strongest elite optimized array.

detailed description

In order to deepen the understanding of the present invention, the present invention will be further described below in conjunction with the embodiments and accompanying drawings. The embodiments are only used to explain the present invention and do not constitute a limitation to the protection scope of the present invention.

A multi-layer elite role method collaborative balance method for mining diagnostic rules of traditional Chinese medicine medical records, the specific steps are as follows:

A. Assign the attributes of TCM medical records to different "ordinary-elite" role evolution populations, pre-process the relevant and interdependent attributes in TCM medical records, and design the elite concentration ECD(Elitist _i ) and selection probability P( x _i ), Elitist _i is the i-th elite role population, x _i is the i-th selection of TCM medical records, assigning attribute-related and interdependent TCM medical records to the elite role population, there will be no probability of being selected P( _xi ) The selected general TCM medical record attributes are assigned to the general role population;

B. Construct a multi-layer elite role dynamic equilibrium strategy for mining diagnostic rules of TCM medical records, through the j-th elite in the i-th elite role subpopulation dynamic immigration rate and eviction rate Carry out adaptive migration of elites to achieve the global equilibrium point Mbest of multi-level elite roles;

C. Realize the mining of TCM medical record diagnosis rules based on multi-layer elite role subpopulation vectors, and select n elite subset optimization vectors with the strongest optimization ability from the multi-layer elite role set Realize the global search and local refined mining of n TCM medical record attribute sets MPR ₁ , MPR ₂ ,...,MPR _n , and extract the corresponding TCM medical record attribute subsets MPR ₁ , MPR ₂ ,...,MPR _n respectively The diagnostic rule mining set DRS ₁ , DRS ₂ ,...,DRS _n ;

D. Output the global optimal diagnostic rule set of TCM medical records under the optimization of the strongest elite

Step A is specifically as follows: assign the attributes of TCM medical records to different "common-elite" role evolution populations, perform preprocessing of related and interdependent attributes in TCM medical records, and design the elite concentration ECD (Elitist _i ) of the elite role population and select Probability P( _xi ), Elitist _i is the i-th elite role population, x _i is the i-th selection of TCM medical records, assigning attribute-related and interdependent TCM medical records to the elite role population, there will be no probability of being selected P( x _i ) The selected general TCM medical record attributes are assigned to the general role population, and the specific steps are as follows:

a. Calculate the elite role concentration ECD(Elitist _i ) of each elite population as

Where Elitist _i is the i-th elite character population, f(x _j ) is the fitness of the j-th elite in the elite character population, is the average fitness of the i-th elite character population;

b. The selection probability P(x _i ) for constructing the i-th elite role population to select the attributes of TCM medical records is

c. Assign the attribute-related and interdependent TCM medical record attributes in the TCM medical records to the elite role population Elitist _i with the selection probability P( _xi ), and assign the general TCM medical record attributes that are not selected by the selection probability P( _xi ) to In the common character population;

d. Each TCM medical record attribute set assigned to the elite role population Mapped to [0,1] evolutionary population space for normalization processing, the mapping formula is as follows:

where f(x _j ) is the fitness of the jth elite in the elite character population, is the average fitness of the i-th elite role population, f _max and f _min are the maximum fitness and minimum fitness of the i-th elite role population;

if Then MPR _i is 0;

If f(x _j ) and equal, then MPR _i is 1.

Step B is specifically: Construct a multi-layer elite role dynamic equilibrium strategy for mining diagnostic rules of TCM medical records, through different elite role subpopulations of elites dynamic immigration rate and eviction rate Carry out adaptive migration of elites to achieve the global equilibrium point Mbest of multi-layer elite roles. The specific steps are as follows:

a. Divide the evolutionary space into n equal-area pentagons Pe ₁ , Pe ₂ ,...,Pe _n , all evolutionary elite character populations are evenly distributed to different pentagons, and then divide each pentagon Divided into several triangular small evolutionary regions of equal size;

b. Select the elite evolutionary individual with the best fitness in the i-th pentagonal evolutionary area, that is, the elite role And sort them according to the degree of fitness from low to low, so as to form the multi-layer elite role set Er _i in the i-th pentagon area as follows:

Where N _elt is the maximum number of multi-layer elite characters in the i-th pentagon area;

c. Calculate each elite role Dynamic immigration rate in Er _i and eviction rate Calculated as follows:

Among them, I _i is the maximum dynamic migration rate of elite characters in the ith pentagonal area, E _i is the maximum dynamic migration rate of elite characters in the ith pentagonal area, and j is the elite in the multi-layered elite role set Er _i The sequence number of the role arrangement;

d. Multi-layer elite roles focus on each elite role will be based on the rate of immigration and eviction rate Perform dynamic adaptive migration in each adjacent pentagonal area, calculate the average fitness of elite characters in each pentagonal evolution area, and then send it to the population of elite characters in adjacent pentagonal areas, thereby adjusting their dynamic migration Intake rate and eviction rate

e. Repeat the above steps until the elite roles divided in each pentagonal area are no longer migrated, and the overall equilibrium state of multi-layer elite roles is reached;

f. The global equilibrium point Mbest for constructing multi-layer elite roles is as follows:

Through the above-mentioned dynamic balance adjustment based on multi-level elite roles, in each pentagonal space Pe _i , multi-level elite roles will be able to find their respective best immigration rates and eviction rate Then, the dynamic adaptive migration of each elite role is carried out to obtain the global equilibrium point of the elite role in each pentagonal evolutionary area, so that a better population of elite roles can be selected for attribute-related and interdependent TCM diagnosis rules mining.

Step C is specifically: realizing the diagnosis rule mining of TCM medical records based on the multi-layer elite role population vector, and selecting n elite subset optimization vectors with the strongest optimization ability from the multi-layer elite role set Realize the global search and local refined mining of n TCM medical record attribute sets MPR ₁ , MPR ₂ ,...,MPR _n , and extract the corresponding diagnostic rule mining sets DRS ₁ , DRS ₂ ,...,DRS _n respectively , the specific steps are as follows:

a. The elite roles in each elite role set Er _i in the global equilibrium point Mbest of multi-layer elite roles are sorted in descending order according to their optimization capabilities, in order

b. Select n elite subset optimization vectors with the strongest optimization ability from the multi-layer elite role set Sequentially assigned to n TCM medical record attribute subsets MPR ₁ , MPR ₂ ,..., MPR _n for global search and mining of diagnostic rules, expressed as:

c. Optimize the elite subset vector with the strongest ability According to the dimensions of the attribute subsets of TCM medical records, they are expanded into n×n dimensional arrays to form the strongest elite optimization array;

d. The strongest elite optimization array containing n dimensions is sequentially subjected to partial refinement mining of diagnostic rules in TCM medical record attribute subsets MPR ₁ , MPR ₂ ,..., MPR _n , and the corresponding TCM diagnostic rule set is extracted as DRS ₁ ,DRS ₂ ,...,DRS _n ;

e. Judging whether the accuracy of the TCM medical record diagnosis rule set DRS _i excavated by the above elite subset vector meets the requirements, if so, then output the global optimal diagnosis rule set of the TCM medical record; if not, repeat the local refinement mining until The accuracy of rule mining meets the requirements.

The invention discloses a multi-layer elite role method for excavating diagnostic rules of medical records of traditional Chinese medicine. This method first designs the concentration selection probability of elite role subpopulations, assigns TCM medical record attributes to different types of "common-elite" role populations, and preprocesses the correlation and interdependence attributes in TCM medical records; then constructs a multi-level elite role based The dynamic equilibrium strategy of the multi-layer elite role is formed through the dynamic migration and migration of different elite role populations in each pentagonal evolution area; finally, the most powerful global search and local refinement is selected from the multi-layer elite roles. The elite subset vector of optimization ability is used to construct the strongest elite optimization array to realize the rapid mining of diagnostic rules of TCM medical records.

The present invention can better deal with the characteristics of complex correlation and interdependence among the attributes of TCM medical records, overcome the problems of difficult mining of diagnostic rules, effectively improve the efficiency of mining diagnostic rules of TCM medical records, and has strong robustness and practicability . The diagnostic rules of TCM excavated by this method have a good consistent effect with the diagnostic rules inferred by actual clinicians based on diagnostic experience. This method has strong objectivity and scientificity. The field of decision support analysis will have strong application value.

Claims (1)

1. it is a kind of for traditional Chinese medical science case history diagnostic rule excavate multilayer elite role method, it is characterised in that：Comprise the following steps that：

A, traditional Chinese medical science case history attribute is assigned in different " common-elite " role's Advanced group species, carried out related in traditional Chinese medical science case history Attribute is pre-processed with interdepending, by the elite concentration ECD (Elitist for designing elite role population_i) and select probability P (x_g), Elitist_iIt is i-th elite role population, x_gFor g-th elite carries out traditional Chinese medical science case history selection, by attribute correlation and phase The traditional Chinese medical science case history for mutually relying on is distributed into elite role population, will be without selected probability P (x_g) traditional Chinese medical science case history attribute chosen In distribution to conventional character population；

A kind of multilayer elite role's dynamic equalization strategy excavated for traditional Chinese medical science case history diagnostic rule of B, construction, by i-th essence J-th elite in the sub- population of English roleWith dynamic rate of moving intoAnd emigrationElite adaptive-migration is carried out, Reach the global equilibrium point Mbest of multilayer elite role；

C, the traditional Chinese medical science case history diagnostic rule excavation for realizing being based on the sub- population vector of multilayer elite role, from multilayer elite role set Select the most strong elite subset optimization vector of p optimization abilityRealize w traditional Chinese medical science case history property set MPR₁, MPR₂,...,MPR_wGlobal search and local excavation of refining, traditional Chinese medical science case history attribute set MPR is extracted respectively₁,MPR₂,..., MPR_wEach self-corresponding diagnostic rule excavates collection DRS₁,DRS₂,...,DRS_w；

The global optimum's diagnostic rule collection of D, output traditional Chinese medical science case history under most strong elite optimization

The step A's comprises the following steps that：

A1, the elite role concentration ECD (Elitist for calculating each elite population_i) be

$ECD\left({\mathrm{Elitist}}_i\right)=\underset{g=1}{\overset{n}{\Σ}}|f_{{\mathrm{Elitist}}_i}-f\left(x_g\right)|,i=1,2,...,n,$

Wherein Elitist_iIt is i-th elite role population, f (x_g) it is g-th fitness of elite in elite role population,It is i-th average fitness of elite role population, n is elite role's sum；

B1, g-th select probability P (x of elite role selecting traditional Chinese medical science case history attribute of structure_g) be

$P\left(x_g\right)=\frac{ECD\left({\mathrm{Elitist}}_i\right)}{\underset{i=1}{\overset{n}{\Σ}}ECD\left({\mathrm{Elitist}}_i\right)}=\frac{\underset{g=1}{\overset{n}{\Σ}}|f_{{\mathrm{Elitist}}_i}-f\left(x_g\right)|}{\underset{i=1}{\overset{n}{\Σ}}\underset{g=1}{\overset{n}{\Σ}}|f_{{\mathrm{Elitist}}_i}-f\left(x_g\right)|};$

Wherein n is elite role's sum；

C1, with select probability P (x_g) the related and complementary traditional Chinese medical science case history attribute of attribute in traditional Chinese medical science case history is assigned to elite angle Color population Elitist_iIn, will be without selected probability P (x_g) traditional Chinese medical science case history attribute chosen is assigned in conventional character population；

D1, u-th traditional Chinese medical science case history property set MPR that will be assigned in elite role population_u[0,1] Advanced group species space is mapped to, It is normalized, mapping equation is as follows：

${\mathrm{MPR}}_u=\{\begin{array}{cc}1-4\×\frac{|f_{{\mathrm{Elitist}}_i}-f\left(x_g\right)|}{|f_{\max }-f_{\min }|},& \frac{|f_{{\mathrm{Elitist}}_i}-f\left(x_g\right)|}{|f_{\max }-f_{\min }|}\≤0.25\\ 0& other\end{array},$

Wherein f (x_g) it is g-th elite fitness in elite role population,It is the average adaptation of the i-th elite role population Degree, f_maxAnd f_minIt is the maximum adaptation degree and minimum fitness of the i-th elite role population；

IfThen MPR_uIt is 0；

If f (x_g) andIt is equal, then MPR_uIt is 1；

The step B is comprised the following steps that：

A2, evolution space is divided into the t pentagon Pe of area equation₁,Pe₂,...,Pe_t, all evolution elite role populations It is evenly distributed in different pentagons, each pentagon is then divided into several equal-sized triangle microevolution areas Domain；

B2, select the elite with adaptive optimal control degree in i-th pentagon evolution region and evolve individual, i.e. elite roleAnd They are ranked up according to fitness size to low, so as to form multilayer elite role set Er in i-th pentagonal regions_i It is as follows：

${\mathrm{Er}}_i=\underset{j=1}{\overset{N_{elt}}{\∪}}{\mathrm{Er}}_i^j,$

Wherein N_eltIt is the maximum quantity of multilayer elite role in i-th pentagonal regions, j is multilayer elite role set Er_iMiddle essence The serial number of English role arrangement；

C2, calculate each elite roleIn Er_iIn dynamic move into rateAnd emigrationComputing formula is such as Under：

${\mathrm{IMR}}_i^j={\mathrm{IMR}}_i^{j-1}+I_i\×(1-\frac{j}{N_{elt}}),$

${\mathrm{EMR}}_i^j={\mathrm{EMR}}_i^{j-1}+E_i\×\left(\frac{j}{N_{elt}}\right),$

Wherein I_iIt is elite role's maximum dynamic move into rate, E in i-th pentagonal regions_iIt is elite in i-th pentagonal regions The maximum dynamic emigration of role, j is multilayer elite role set Er_iThe serial number of middle elite role's arrangement；

Each elite role in d2, multilayer elite role setWill be according to the rate of moving intoAnd emigrationIn respective phase Dynamic self-adapting migration is carried out in adjacent pentagonal regions, the average adaptation of each pentagon evolution region elite role is calculated Degree, is then transferred to the elite role population of adjacent pentagonal regions, so as to adjust each dynamic rate of moving intoAnd emigration

E2, repeat the above steps, until the elite role being divided in each pentagonal regions no longer migrates, reach multilayer essence The overall equilibrium state of English role；

F2, build multilayer elite role global equilibrium point Mbest it is as follows：

$\begin{array}{c}Mbest=\left(\frac{1}{N_{elt}}\underset{j=1}{\overset{N_{elt}}{\Σ}}{\mathrm{Er}}_1^j,\frac{1}{N_{elt}}\underset{j=1}{\overset{N_{elt}}{\Σ}}{\mathrm{Er}}_2^j,\mathrm{...},\frac{1}{N_{elt}}\underset{j=1}{\overset{N_{elt}}{\Σ}}{\mathrm{Er}}_t^j\right)\\ =({\mathrm{Mbest}}_1,{\mathrm{Mbest}}_2,\mathrm{...},{\mathrm{Mbest}}_t)\end{array},$

Adjusted by the above-mentioned dynamic equalization based on multilayer elite role, the multilayer elite role in each pentagonal regions space To can find each best rate of moving intoAnd emigrationThen the dynamic self-adapting for carrying out each elite role is moved Move, the global equilibrium point of elite role in each pentagon evolution region is obtained, so as to pick out more excellent elite role's kind Group carries out the related and complementary traditional Chinese medical science case history diagnostic rule of attribute and excavates；

The step C is comprised the following steps that：

A3, will be in multilayer elite role overall situation equilibrium point Mbest each elite role set Er_iMiddle elite role is excellent according to its Change ability carries out descending sort from high to low, is followed successively byN is elite role's sum；

B3, select from multilayer elite role set the most strong elite subset optimization vector of p optimization ability It is sequentially allocated w traditional Chinese medical science case history attribute set MPR₁,MPR₂,...,MPR_wIn, excavated for diagnostic rule global search, its It is expressed as：

${\mathrm{MPR}}_1\left({\stackrel{\→}{Er}}_1^1\right),{\mathrm{MPR}}_2\left({\stackrel{\→}{Er}}_2^1\right),...,{\mathrm{MPR}}_w\left({\stackrel{\→}{Er}}_p^1\right);$

C3, by the most strong elite subset vector of optimization abilityRespectively according to the dimension of traditional Chinese medical science case history attribute set Degree is launched into s × s dimension arrays, forms most strong elite optimization array；

D3, will contain s dimension most strong elite optimization array carry out traditional Chinese medical science case history attribute set MPR successively₁,MPR₂,...,MPR_wIn examine The part of rule of breaking is refined excavation, extracts its corresponding tcm diagnosis rule set respectively DRS₁,DRS₂,...,DRS_w；

E3, judge that whether traditional Chinese medical science case history diagnostic rule that above-mentioned elite subset vector is excavated concentrates each rule set precision full Foot is required, as met, then exports traditional Chinese medical science case history global optimum diagnostic rule collection；Such as it is unsatisfactory for, then repeats local digging of refining Pick, until rule digging precision meets requiring.