CN106227798A - A kind of Ontology Matching method based on compact Cooperative Evolutionary Algorithm - Google Patents

Abstract

A kind of Ontology Matching method based on compact Cooperative Evolutionary Algorithm, sets up the Optimized model of Ontology Matching problem, builds similarity matrix；Use compact Cooperative Evolutionary Algorithm solving-optimizing model, obtain optimum Ontology Matching result；Initialize the most individual probability vector PV_better and probability vector PV_worse of poor individuality, generate elite solution ind_BElite of the most individual initial probability vector and elite solution ind_Bworse of the probability vector of poor individuality by PV_better and PV_worse；Each individual UVR exposure information had both included that the weight of the mapping result for integrated different measuring similarities also included the threshold value for filtering body mapping result；The matching result that the measuring similarity using average weighted method integration different produces.The present invention inherently reduces time and the amount of ram that body matching system based on evolution algorithm consumes in running, thus improves the efficiency of Ontology Matching process.

Description

A kind of Ontology Matching method based on compact Cooperative Evolutionary Algorithm

[technical field]

The invention belongs to computer information technology field, specifically refer to the Ontology Matching side of a kind of compact Cooperative Evolutionary Algorithm Method.

[background technology]

Along with the development of semantic net, occur in that substantial amounts of body.Owing to application purpose is similar, many bodies coexist in same In individual field.Yet with the subjectivity of people, the different bodies of same application may define same by different modes Individual entity object, creates bulk heteroj problem.In order to realize different application systems at semantic level by body Cooperation, it is thus necessary to determine that semantic corresponding relation between element in different bodies.For having the big rule of up to a million conceptual entities For the body of mould, it is unpractical for completing Ontology Matching by the way of artificial.Accordingly, it would be desirable to develop efficient body Match system is automatically performed Ontology Matching process.

Owing to user cannot wait too long of system response process, therefore for dynamic application scenarios, body Match system is extremely emphasized to complete Ontology Matching process within the limited operation time.In this sense, except body Joining outside the quality of result, the efficiency (the operation time needed for matching process and the amount of ram of consumption) of Ontology Matching process is to closing Important.Although the most existing body matching system based on evolution algorithm all have employed different strategies improves Ontology Matching The efficiency of process, but the basic algorithm framework owing to using does not has the lifting of essence, so efficient and intelligent Ontology Matching Process remains a challenge.

In body matching system based on evolution algorithm, foremost is GOAL (Genetics for Ontology ALignments).GOAL cannot directly calculate the coupling between two bodies, but determined the weight of optimum by evolution algorithm Configure with integrated different measuring similarity technology.Additionally, Vitiello etc. proposed to be solved by HYBRID EVOLUTIONARY ALGORITHMS in 2012 Ontology Matching problem.Owing to HYBRID EVOLUTIONARY ALGORITHMS adds local searching operator in conventional evolutionary algorithm, improve evolution and seek The efficiency of excellent process.

Compaction algorithm is that a class estimates distributed algorithm, can be represented by probability-distribution function by complete population. Chinese invention patent 201510865803.8 discloses a kind of Ontology Matching method based on compact evolution algorithm, therein compact Evolution algorithm only utilizes a PV vector, it is difficult to take into account global search and local search procedure so that the difficult quality of result obtains To ensureing.

The internal memory that existing body matching system based on evolution algorithm consumes during coupling body is excessive, operation Time is oversize.

[summary of the invention]

The technical problem to be solved is to provide a kind of Ontology Matching method of compact Cooperative Evolutionary Algorithm, from this Reduce time and amount of ram that body matching system based on evolution algorithm consumes in running in matter, thus improve body The efficiency of matching process.

The present invention is achieved in that

A kind of Ontology Matching method based on compact Cooperative Evolutionary Algorithm, comprises the steps:

Step 1): set up the Optimized model of Ontology Matching problem, build similarity matrix；Specifically include:

Step 1.1): set up the Optimized model of this volume elements matching problem:

$\left\{\begin{array}{cc}\max & f\left(X\right)=f-measure\left(X\right)\\ s.t.& X={(x_1,x_2,...,x_{n+1})}^T\\ & x_i\∈\[0,1\],i=\mathrm{1...}n+1\end{array}\right.---\left(1\right)$

Wherein, f-measure (X) is for the quality of metrics match result, and n represents the individual of the measuring similarity technology of employing Count or treat integrated similarity matrix number, x_n+1Represent the threshold value for filtering final matching results；

Step 1.2): treat body O1=(C1, P1, I1), O2=(C2, P2, I2) for given two；Wherein O1 and O2 table respectively Show two bodies；C1 and C2 is respectively the set of concept in O1 and O2；P1 and P2 is respectively in O1 and O2 relation between concept Set；I1 and I2 is respectively the set of example in O1 and O2, and described example is the practical object that concept is corresponding；Concept, concept it Between relation and example be referred to as the entity in body；It is utilized respectively different measuring similarity technology and generates similarity matrix, often Plant the corresponding similarity matrix of measuring similarity technology；The row and column of similarity matrix is respectively the entity in O1 and O2, phase It is to the corresponding example similarity to carrying out similarity evaluation generation by measuring similarity technology like the element in degree matrix Value；

Step 2): use compact Cooperative Evolutionary Algorithm solution procedure 1) in Optimized model, obtain optimum Ontology Matching Result；Specifically include:

Step 2.1) initialize the most individual probability vector PV_better and probability vector PV_ of poor individuality Worse, by PV_better and PV_worse generate the most individual initial probability vector elite solution ind_BElite and Elite solution ind_Bworse of the probability vector of poor individuality；Each individual UVR exposure information had both included for integrated different similarities The weight of the mapping result of tolerance also includes the threshold value for filtering body mapping result；Use average weighted method integration not The matching result that same measuring similarity produces, is described in detail below:

$\mathrm{\φ}\left(\stackrel{\→}{s}\right(c),\stackrel{\→}{w})=\underset{i=1}{\overset{n}{\Σ}}{w}_i{s}_i\left(c\right)---\left(2\right)$

Wherein It is that the mapping result that different measuring similarities obtains is vectorial,It it is weight Vector, n is the number of measuring similarity technology；

Encode by being indirectly indicative different weights at cut-point defined in interval [0,1], it is assumed that p is required power Weight number, then cut-point set can be expressed as c '={ c '₁,c′₂,...,c′_p-1}；

Decoding process is divided into two steps: is first arranged according to ascending order by the element in cut-point set, obtains new collection Close c={c₁,c₂,...,c_p-1, then according to the weight that below equation calculating is different:

$w_k=\left\{\begin{array}{c}c1,k=1\\ c_k-c_{k-1},1<k<p\\ 1-c_{p-1},k=p\end{array}\right.---\left(3\right)$

It is [0,1] for the threshold value of filtering body mapping result by a coded representation, its span；

Step 2.2): update PV_better by index crossover operator, update PV_ by conventional PV update algorithm worse；

Step 2.3): tied by the respective Ontology Matching of coding acquisition of information in ind_BElite and ind_WElite Really alignment_BElite and alignment_WElite, by fitness function the two quality good or not of tolerance the most really The most up-to-date fixed PV_better and PV_worse, it is achieved mutual between two PV vectors；

Step 2.4): if end condition is unsatisfactory for, return step 2.2), otherwise export ind_BElite.

Further, described fitness function, for evaluating the body obtained by the weight in individual UVR exposure and threshold value The object function of mapping result quality, object function is the f-measure value of Ontology Matching result.

It is an advantage of the current invention that: the present invention is directed to existing body matching system based on Cooperative Evolutionary Algorithm at body Matching process builds complete colony participate in evolutionary process and cause the excessive problem of memory consumption and comment in actual applications Internal memory that valency individuality consumes and time the biggest problem, it is proposed that use compact Cooperative Evolutionary Algorithm interior to reduce that colony consumes Storage, comes Local Search and the global search process of tuning algorithm by the cooperation of two probability vectors.The present invention uses compact Cooperative Evolutionary Algorithm replaces traditional evolution algorithm to reduce the memory consumption during Ontology Matching, overcomes tradition simultaneously The defect of evolution algorithm Premature Convergence.The compact Cooperative Evolutionary Algorithm that the present invention proposes is by two PV vector (i.e. PV_better Being responsible for local search procedure, PV_worse is responsible for global search process) cooperation with mutual, taken into account the global search in algorithm With local search procedure, improve the quality of Ontology Matching result.

[accompanying drawing explanation]

The invention will be further described the most in conjunction with the embodiments.

Fig. 1 is the compact Cooperative Evolutionary Algorithm flow chart of the solving-optimizing model of the present invention.

Fig. 2 is the most individual probability vector PV_better flow chart in the present invention.

Fig. 3 is the probability vector PV_worse flow chart of the poor individuality in the present invention.

[detailed description of the invention]

A kind of Ontology Matching method based on compact Cooperative Evolutionary Algorithm, comprises the steps:

Step 1): set up the Optimized model of Ontology Matching problem, build similarity matrix；Specifically include:

Step 1.1): set up the Optimized model of this volume elements matching problem:

$\left\{\begin{array}{cc}\max & f\left(X\right)=f-measure\left(X\right)\\ s.t.& X={(x_1,x_2,...,x_{n+1})}^T\\ & x_i\&Element;\&lsqb;0,1\&rsqb;,i=\mathrm{1...}n+1\end{array}\right.---\left(1\right)$

Wherein, f-measure (X) is for the quality of metrics match result, and n represents the individual of the measuring similarity technology of employing Count or treat integrated similarity matrix number, x_n+1Represent the threshold value for filtering final matching results；

Step 1.2): treat body O1=(C1, P1, I1), O2=(C2, P2, I2) for given two；Wherein O1 and O2 table respectively Show two bodies；C1 and C2 is respectively the set of concept in O1 and O2；P1 and P2 is respectively in O1 and O2 relation between concept Set；I1 and I2 is respectively the set of example in O1 and O2, and described example is the practical object that concept is corresponding；Concept, concept it Between relation and example be referred to as the entity in body；It is utilized respectively different measuring similarity technology and generates similarity matrix, often Plant the corresponding similarity matrix of measuring similarity technology；The row and column of similarity matrix is respectively the entity in O1 and O2, phase It is to the corresponding example similarity to carrying out similarity evaluation generation by measuring similarity technology like the element in degree matrix Value；

Step 2): use compact Cooperative Evolutionary Algorithm solution procedure 1) in Optimized model, obtain optimum Ontology Matching Result；As shown in Figure 1 to Figure 3, specifically include:

Step 2.1) initialize the most individual probability vector PV_better and probability vector PV_ of poor individuality Worse, by PV_better and PV_worse generate the most individual initial probability vector elite solution ind_BElite and Elite solution ind_Bworse of the probability vector of poor individuality；Each individual UVR exposure information had both included for integrated different similarities The weight of the mapping result of tolerance also includes the threshold value for filtering body mapping result；Use average weighted method integration not The matching result that same measuring similarity produces, is described in detail below:

$\mathrm{\&phi;}\left(\stackrel{\&RightArrow;}{s}\right(c),\stackrel{\&RightArrow;}{w})=\underset{i=1}{\overset{n}{\&Sigma;}}{w}_i{s}_i\left(c\right)---\left(2\right)$

Wherein It is that the mapping result that different measuring similarities obtains is vectorial,It it is weight Vector, n is the number of measuring similarity technology；

Encode by being indirectly indicative different weights at cut-point defined in interval [0,1], it is assumed that p is required power Weight number, then cut-point set can be expressed as c '={ c '₁,c′₂,...,c′_p-1}；

Decoding process is divided into two steps: is first arranged according to ascending order by the element in cut-point set, obtains new collection Close c={c₁,c₂,...,c_p-1, then according to the weight that below equation calculating is different:

$w_k=\left\{\begin{array}{c}c1,k=1\\ c_k-c_{k-1},1<k<p\\ 1-c_{p-1},k=p\end{array}\right.---\left(3\right)$

It is [0,1] for the threshold value of filtering body mapping result by a coded representation, its span；

Step 2.2): update PV_better by index crossover operator, update PV_ by conventional PV update algorithm worse；

Step 2.3): tied by the respective Ontology Matching of coding acquisition of information in ind_BElite and ind_WElite Really alignment_BElite and alignment_WElite, by fitness function (f-measure) the two Functionality, quality and appealing design of tolerance Bad and further determine that up-to-date PV_better and PV_worse, it is achieved mutual between two PV vectors；

Step 2.4): if end condition is unsatisfactory for, return step 2.2), otherwise export ind_BElite.

Wherein fitness function, for evaluating the Ontology Mapping result matter obtained by the weight in individual UVR exposure and threshold value The object function of amount, object function is the f-measure value of Ontology Matching result.

Wherein selection opertor, first carries out descending sort according to the crowding of Different Individual in colony, and before selecting to come The individuality of half part, therefrom random reproduction body one by one is until forming new colony.

Wherein single-point crossover operator, first determines a cut-point in father's individuality at random, and this cut-point is by two fathers Body is divided into two parts: left-hand component and right-hand component, then by the coding of two father's individuality right-hand components of exchange to produce Two sons of tissue regeneration promoting are individual.

Wherein Mutation operator, first determines the individual bits of coded knowing from experience generation variation according to mutation probability, then will The value of these bits of coded is revised as 0 from 1, or is revised as 1 from 0.

The present invention is directed to existing body matching system based on Cooperative Evolutionary Algorithm built during Ontology Matching Whole colony participates in evolutionary process and causes the excessive problem of memory consumption and evaluate the internal memory that individuality consumes in actual applications The problem the biggest with the time, it is proposed that use compact Cooperative Evolutionary Algorithm to reduce the amount of ram that colony consumes, general by two The cooperation of rate vector carrys out Local Search and the global search process of tuning algorithm.The present invention uses compact Cooperative Evolutionary Algorithm to take For traditional evolution algorithm with the memory consumption during minimizing Ontology Matching, overcome conventional evolutionary algorithm Premature Convergence simultaneously Defect.By two PV vectors, (i.e. PV_better is responsible for Local Search mistake to the compact Cooperative Evolutionary Algorithm that the present invention proposes Journey, PV_worse is responsible for global search process) cooperation with mutual, taken into account the global search in algorithm and Local Search mistake Journey, improves the quality of Ontology Matching result.

The foregoing is only the present invention preferably implements use-case, is not intended to limit protection scope of the present invention.All Within the spirit and principles in the present invention, any amendment, equivalent and the improvement etc. made, should be included in the present invention's Within protection domain.

Claims (2)

1. an Ontology Matching method based on compact Cooperative Evolutionary Algorithm, it is characterised in that: comprise the steps:

Step 1): set up the Optimized model of Ontology Matching problem, build similarity matrix；Specifically include:

Step 1.1): set up the Optimized model of this volume elements matching problem:

Wherein, f-measure (X) for the quality of metrics match result, n represent the measuring similarity technology of employing number or Treat integrated similarity matrix number, x_n+1Represent the threshold value for filtering final matching results；

Step 1.2): treat body O1=(C1, P1, I1), O2=(C2, P2, I2) for given two；Wherein O1 and O2 represents two respectively Individual body；C1 and C2 is respectively the set of concept in O1 and O2；P1 and P2 is respectively the set of relation between concept in O1 and O2； I1 and I2 is respectively the set of example in O1 and O2, and described example is the practical object that concept is corresponding；Pass between concept, concept System and example are referred to as the entity in body；Being utilized respectively different measuring similarity technology and generate similarity matrix, every kind similar The corresponding similarity matrix of degree measurement technology；The row and column of similarity matrix is respectively the entity in O1 and O2, similarity moment Element in Zhen is to the corresponding example Similarity value to carrying out similarity evaluation generation by measuring similarity technology；

Step 2): use compact Cooperative Evolutionary Algorithm solution procedure 1) in Optimized model, obtain optimum Ontology Matching result； Specifically include:

Step 2.1) initialize the most individual probability vector PV_better and probability vector PV_worse of poor individuality, logical Cross PV_better and PV_worse and generate elite solution ind_BElite of the most individual initial probability vector and poor individuality Elite solution ind_Bworse of probability vector；Each individual UVR exposure information had both included reflecting for integrated different measuring similarity The weight penetrating result also includes the threshold value for filtering body mapping result；Use different similar of average weighted method integration The matching result that degree tolerance produces, is described in detail below:

Wherein It is that the mapping result that different measuring similarities obtains is vectorial,It is weight vectors, N is the number of measuring similarity technology；

Encode by being indirectly indicative different weights at cut-point defined in interval [0,1], it is assumed that p is required weight Number, then cut-point set can be expressed as c'={c'₁,c'₂,...,c'_p-1}；

Decoding process is divided into two steps: is first arranged according to ascending order by the element in cut-point set, obtains new set c= {c₁,c₂,...,c_p-1, then according to the weight that below equation calculating is different:

It is [0,1] for the threshold value of filtering body mapping result by a coded representation, its span；

Step 2.2): update PV_better by index crossover operator, update PV_worse by conventional PV update algorithm；

Step 2.3): by the coding acquisition of information respective Ontology Matching result in ind_BElite and ind_WElite Alignment_BElite and alignment_WElite, measures the two quality good or not by fitness function and further determines that Up-to-date PV_better and PV_worse, it is achieved mutual between two PV vectors；

Step 2.4): if end condition is unsatisfactory for, return step 2.2), otherwise export ind_Belite.

A kind of Ontology Matching method based on compact Cooperative Evolutionary Algorithm, it is characterised in that: described Fitness function, for evaluating the target letter of the Ontology Mapping outcome quality obtained by the weight in individual UVR exposure and threshold value Number, object function is the f-measure value of Ontology Matching result.