CN105282229A - Web service composition method based on improved QPSO algorithm - Google Patents

Abstract

The invention discloses a web service composition method based on an improved QPSO algorithm. Based on the Web service task numbers, each task has ni (i=1, 2, 3.....m) candidate Web services having same service functions and different QoSs, the Web services are encoded at a fixed length based on the task numbers, the number of the services is shown in the description, services meeting user demands and constrained conditions can be found, and service instance numbers corresponding to particles are resolved based on the improved QPSO algorithm, namely, an optimum service sequence can be obtained, and a larger service function can be provided for the Web service composition. The iteration frequency for achieving an optimal solution is reduced and the practical significance for building complex composition services based on Internet is achieved.

Description

Based on the web service composition method improving QPSO algorithm

Technical field

The present invention relates to and solve Web service combination optimization problem, specifically a kind of web service composition method based on improving QPSO algorithm.

Background technology

Web service, as a kind of emerging Web application model, is a brand-new distributed computing platform.Single Web service function is limited, Web service combination (WebServiceComposition, WSC) by distribution loosely-coupled multiple Web service combination composite service of becoming to satisfy the demands on internet, while realizing software reuse, tissue is made to have the ability of " adjust to changed conditions, fast reaction ".WSC is subject to academia and business circles are extensively paid attention to, and emerges a large amount of WSC correlative studys.How dynamically existing Web service combination is got up to be formed new, that can meet different demand, value-added complex services and become new application demand and study hotspot.

Web service combination based on QoS (QualityofService) belongs to non-linear, Mixed Integer Multiple Goal Programming problem, needs the running time considering Web service, the factors such as service cost when Modling model.Service Combination Algorithm at present based on QoS mainly contains the method for exhaustion, linear programming technique, genetic algorithm, simulated annealing etc.Web service combination problem based on QoS belongs to np hard problem, although adopt the combined optimization method of limit can obtain optimal solution in theory, amount of calculation is very big, autgmentability is poor, Time and place expense is large; Nonlinear QoS attribute and constraints need be carried out linear transformation by linear programming technique, and the practicality of algorithm suffers restraints; Genetic algorithm local search ability is more weak, and convergence rate is slow, is easily absorbed in local optimum; Simulated annealing belongs to single-point optimizing on solution combinatorial optimization problem, does not have advantage to Web service combination problem.

Particle swarm optimization algorithm (ParticleSwarmOptimization, PSO) be a kind of global optimization approach, its basic thought be by individual in population between cooperation and information sharing find optimal solution, but the search volume of particle is a limited region, whole feasible space can not be covered, can not ensure to converge to globally optimal solution.2004, the people such as Sun proposed a kind of new PSO algorithm model from quantum-mechanical angle and improve standard P SO.This model, based on DELTA potential well, thinks that particle has quantum behavior, and proposes the particle swarm optimization algorithm (Quantum-behavedParticleSwarmOptimization, QPSO) based on quantum behavior according to this model.In vector subspace, particle is searched in whole solution space, and its global search performance is better than PSO algorithm, but the particle in population is easy to attracted by the best particle position of the current overall situation and quickly converge on a local optimum, causes precocity.

Summary of the invention

The object of the invention is by the improvement to QPSO algorithm, providing the web service composition method based on improving QPSO algorithm, to solve the Web service scheme more meeting user's request.

Based on the web service composition method improving QPSO algorithm, comprise the following steps,

Step 1, definition M particle, represent M bar service combination path;

Step 2, the dimension defining each particle are D, represent Web service combination number of tasks, the numbering of the corresponding a certain service of value of D;

Step 3, basis provide the QoS initialization particle position vector of service;

Step 4, QoS property parameters particle often being tieed up service substitute into the QoS computation model of Web service combination, calculate the polymerization QoS attribute that Web service combination is corresponding;

Step 5, employing improve the evolution that QPSO algorithm carries out particle;

Step 6, output form the optimal service of Services Composition.

Described Service Quality Metrics is that service execution time is less than 100s, and availability is greater than 0.8.

The detailed process of described step 5 is,

Step 51, initialization population, according to the random site of the fitness setting particle of particle, and the desired positions p of initialization particle experience _i, all particles experience in population desired positions p _gwith iterations t;

Step 52, according to current iteration number of times dynamic conditioning QPSO shrinkage expansion coefficient

β＝(1-0.5)*(Maxiter-t)/Maxiter+0.5，

Its value is with iterations from 1.0 to 0.5 linear decrease, and wherein Maxiter represents maximum iteration time;

Boolean distance pbd under step 53, calculating current iteration number of times between each particle and global optimum's position particle _i,g: pbd _i,g, particle weights w _i, with the weight average particle desired positions point mbest of boolean's distance weighting _w;

Step 54, evolution equation according to population, generate new particle, and determine whether to upgrade p _iand p _g;

Whether step 55, the number of particles judging whether to reach maximum iteration time or generation equal the service plan number of user's request, namely generate enough number of particles, if then terminate to calculate, otherwise get back to step 52.

The concrete computational process of described step 53 is as follows, boolean's distance between particle

${\mathrm{pbd}}_{i,j}:{\mathrm{pbd}}_{i,j}=\frac{1}{D}\underset{d=1}{\overset{D}{\Σ}}iff(s_{i,d}\≠s_{j,d},1,0)$

Wherein, $iff(s_{i,d}\≠s_{j,d},1,0)=\left\{\begin{array}{c}1,s_{i,d}\≠s_{j,d}\\ 0,s_{i,d}=s_{j,d}\end{array}\right.,$ S _i,drepresent the value that i-th particle d ties up, S _j,drepresent the

The value that j particle d ties up,

Then boolean's distance of particle and optimal location particle is expressed as:

${\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}=\frac{1}{D}\underset{d=1}{\overset{D}{\Σ}}iff(s_{i,d}\≠s_{g,d},1,0)$

Wherein S _{g, d}represent the value of optimal location particle d dimension,

Weight

$w_i=\frac{{\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}}{\underset{i=1}{\overset{M}{\Σ}}{\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}}$

Weight average particle desired positions point

${\mathrm{mbest}}_w=\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_i=(\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{i1},\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{i2},...\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{iD}).$

Described step 54 detailed process is,

According to Evolution of Population equation wherein in [0,1] upper equally distributed random number, 1

≤ d≤D, generates new particle x _id,

Ifrand(0,1)>0.5x _id＝p _id-β*abs(mbest _wd-x _id)*ln(1/u)

Elsex _id＝p _id+β*abs(mbest _wd-x _id)*ln(1/u)

If f is (x _i) > f (p _i) set up, then make p _i=x _i; If f is (p _i) > f (p _g) set up, then make p _g=p _i.

The optimal solution that the web service composition method that the present invention is based on the quanta particle algorithm of improvement is tried to achieve needs iterations less, and solving result is better, has realistic meaning to building the application of serving based on the complex combination of the Internet.

Accompanying drawing explanation

Fig. 1 is the flow chart of the QPSO algorithm improved;

Fig. 2 is the QoS computation model of the Web service combination of Web service combination order model and correspondence thereof;

Fig. 3 solves flow chart based on the Web service combination improving QPSO algorithm.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

Suppose have m task jointly to provide certain service function, each task has n _icandidate's Web service that (i=1,2,3......m) identical QoS of individual service function is different is available, fixes, by its block code, have according to Web service combination number of tasks plant service, find out the service of meeting consumers' demand with constraints, according to the particle that the QPSO algorithm improved is obtained, corresponding Service Instance is numbered, and can obtain optimal service sequence.

Adopt the web service composition method based on improving QPSO algorithm, as shown in Figure 3, step is as follows,

Step 1): define M particle, represent the service combination path that M kind is possible, population scale is M.

Step 2): the dimension defining each particle is D, represents Web service combination number of tasks, adopts integer block code mode, and the numbering of the corresponding a certain Service Instance of this dimension values, then a solution of Web service combination uses particle x _i=(x _1i, x _2i... x _di) represent, a Web service combination solution is x=(w herein ₁₂, w ₂₄, w ₃₆... w _d7).

Step 3): in the scheme of kind, Stochastic choice meets the M bar combinatorial path initialization particle position vector of Service Quality Metrics, and Service Quality Metrics can be that service execution time is less than 100s, and availability is greater than 0.8.

Step 4): use Web service combination QoS computation model as shown in Figure 2, calculate the polymerization QoS property value that Web service combination is corresponding, the time of implementation D () of such as Web service combination, Executing Cost P (), success rate R _l(), availability A (), reputation degree R _e() etc., the fitness f (x that each particle is corresponding _i), this method both can be the Services Composition improving QPSO algorithm based on single QoS attribute, also can be the Services Composition that many QoS attribute improves QPSO algorithm, if many QoS attribute solves Web service combination problem tries to achieve every bar combinatorial path Integrated Services Quality function by weighting method, standardized transformation is carried out to attribute vector, be worth larger, represent that QoS attribute corresponding to Web service is better.

Step 5): the individual local extremum p of particle _iequal its initial position value, p _gequal the top-quality particle vector value of QoS in M primary vector, iterations t=1.

Step 6): calculate QPSO shrinkage expansion factor beta=(1-0.5) * (Maxiter-t)/Maxiter+0.5.

Step 7): under calculating this iterations, p _iwith p _gboolean's distance particle weights

$w_i=\frac{{\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}}{\underset{i=1}{\overset{M}{\Σ}}{\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}},$

Weight average particle desired positions point

${\mathrm{mbest}}_w=\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_i=(\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{i1},\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{i2},...\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{iD}).$

Step 8): according to wherein be in [0,1] upper equally distributed random number, generate new particle representation formula as follows:

Ifrand(0,1)>0.5x _id＝p _id-β*abs(mbest _wd-x _id)*ln(1/u)

Elsex _id＝p _id+β*abs(mbest _wd-x _id)*ln(1/u)

If f is (x _i) > f (p _i) set up, then p _i=x _icarry out p _ithe renewal of value, if relational expression f is (p _i) > f (p _g) set up, then p _g=p _icarry out p _gthe renewal of value;

Step 9): make t=t+1, judge whether to reach maximum iteration time Maxiter or generate enough number of particles, if meet, forward step 10 to), do not meet and then forward step 6 to);

Step 10): according to the particle of trying to achieve, export the optimal service scheme forming Services Composition according to coding.

Claims (5)

1., based on the web service composition method improving QPSO algorithm, it is characterized in that: comprise the following steps,

Step 1, definition M particle, represent M bar service combination path;

Step 2, the dimension defining each particle are D, represent Web service combination number of tasks, the numbering of the corresponding a certain service of value of D;

Step 3, basis provide the QoS initialization particle position vector of service;

Step 4, QoS property parameters particle often being tieed up service substitute into the QoS computation model of Web service combination, calculate the polymerization QoS attribute that Web service combination is corresponding;

Step 5, employing improve the evolution that QPSO algorithm carries out particle;

Step 6, output form the optimal service of Services Composition.

2. the web service composition method based on improving QPSO algorithm according to claim 1, is characterized in that: Service Quality Metrics described in step 3 is that service execution time is less than 100s, and availability is greater than 0.8.

3. the web service composition method based on improving QPSO algorithm according to claim 1, is characterized in that: the detailed process of described step 5 is,

Step 51, initialization population, according to the random site of the fitness setting particle of particle, and the desired positions p of initialization particle experience _i, all particles experience in population desired positions p _gwith iterations t;

Step 52, according to current iteration number of times dynamic conditioning QPSO shrinkage expansion coefficient

β＝(1-0.5)*(Maxiter-t)/Maxiter+0.5，

Its value is with iterations from 1.0 to 0.5 linear decrease, and wherein Maxiter represents maximum iteration time;

Boolean distance pbd under step 53, calculating current iteration number of times between each particle and global optimum's position particle _i,g: pbd _i,g, particle weights w _i, with the weight average particle desired positions point mbest of boolean's distance weighting _w;

Step 54, generate new particle according to Evolution of Population equation, and determine whether to upgrade p _iand p _g;

Whether step 55, the number of particles judging whether to reach maximum iteration time or generation equal the service plan number of user's request, namely generate enough number of particles, if then terminate to calculate, otherwise get back to step 52.

4. the web service composition method based on improving QPSO algorithm according to claim 3, is characterized in that: the concrete computational process of described step 53 is as follows, boolean's distance between particle

${\mathrm{pbd}}_{i,j}:{\mathrm{pbd}}_{i,j}=\frac{1}{D}\underset{d=1}{\overset{D}{\Σ}}iff(s_{i,d}\≠s_{j,d},1,0)$

Wherein, $iff(s_{i,d}\≠s_{j,d},1,0)=\left\{\begin{array}{c}1,s_{i,d}\≠s_{j,d}\\ 0,s_{i,d}=s_{j,d}\end{array}\right.,$ S _i,drepresent the value that i-th particle d ties up, S _j,drepresent the value that a jth particle d ties up,

Then boolean's distance of particle and optimal location particle is expressed as:

${\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}=\frac{1}{D}\underset{d=1}{\overset{D}{\Σ}}iff(s_{i,d}\≠s_{g,d},1,0)$

Wherein S _{g, d}represent the value of optimal location particle d dimension,

Weight

$w_i=\frac{{\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}}{\underset{i=1}{\overset{M}{\Σ}}{\mathrm{pbd}}_{i,g}:{\mathrm{pbd}}_{i,g}}$

Weight average particle desired positions point

${\mathrm{mbest}}_w=\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_i=(\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{i1},\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{i2},...\frac{1}{M}\underset{i=1}{\overset{M}{\Σ}}{w}_i{p}_{iD}).$

5. the web service composition method based on improving QPSO algorithm according to claim 3, it is characterized in that, described step 54 detailed process is,

According to Evolution of Population equation wherein that 1≤d≤D, generates new particle x in [0,1] upper equally distributed random number _id,

Ifrand(0,1)>0.5x _id＝p _id-β*abs(mbest _wd-x _id)*ln(1/u)

Elsex _id＝p _id+β*abs(mbest _wd-x _id)*ln(1/u)

If f is (x _i) > f (p _i) set up, then make p _i=x _i; If f is (p _i) > f (p _g) set up, then make p _g=p _i.