CN113887691B - Whale evolution system and method for service composition problem - Google Patents

Abstract

The invention discloses a whale evolution method and system oriented to the service composition problem. The steps of the method of the present invention are as follows: Step 1: transform the web service composition problem into a single-objective optimization problem, and establish a fitness function; step 2: carry out the web service composition problem. Integer coding; Step 3: Build the WOA algorithm model; Step 4: Perform integer coding on the WOA algorithm and convert it into the DWOA algorithm; Step 5: Use the DWOA algorithm to start iterative optimization of the problem model; Step 6: End the search and output the search The optimal service combination coding sequence is combined with the QoS value of the optimal web service at this time. When the order of magnitude of the service combination sequence reaches a certain scale, the use of the present invention to solve the optimal service combination will significantly improve the optimal value of the random search method, and it will have higher optimization efficiency compared with the traversal search method. According to the characteristic requirements of reliability and real-time performance in the optimization problem of web service composition, it provides better composition services.

Description

Whale evolution system and method for service composition problem

technical field

The invention relates to the application of a whale optimization algorithm in the Web service composition problem, in particular to a whale evolution system and method oriented to the service composition problem.

Background technique

Under the current development of information technology, the problem of Web service composition has become one of the hotspots in today's information technology applications. Web services are software modules with self-describing and self-contained functions implemented over the Internet that can be published, located, and consumed on the Internet using a set of standards (such as SOAP, WSDL, and UDDI). Such software modules perform tasks, solve problems, and process transactions on behalf of users or applications. In most cases, users usually need to use a series of service combinations to achieve their goals. When a user creates a Web service composition using different categories of Web sub-services, there are usually multiple candidate services with the same function under one category. Due to the large number of current web services, modules that can implement the same function often include Thousands of QoS (Quality-of-Service) different functional modules. Therefore, how to select the optimal sub-service combination from a large number of Web service candidates is a problem that needs to be solved. A high-quality Web service composition should include the characteristics of short response time, low cost, high reliability, and good availability, which are the conventional evaluation criteria of Web services, QoS.

QoS is an important method used to evaluate network function services, including various evaluation criteria. From the service attributes, it can be basically divided into two categories: positive attributes and negative attributes. The higher the positive attribute score, the better, and the negative attribute on the contrary. Among them, success rate, reliability, throughput rate, etc. are classified as positive attributes, while response time, service price, etc. are classified as negative attributes. The evaluation of web services often requires a combination of positive and negative attributes, rather than focusing on a single aspect.

Further explain the web service composition structure process. According to the actual scene division, the common web service composition process composition workflow can be divided into sequential, concurrent, selective and loop structure, and according to different service attributes, the data processing mode will be different in different workflow modes.

This is the Web Services Composition problem description. For a large-scale candidate service set, if each target combination has n types of abstract service categories, and each category consists of m specific candidate Web services, then the number of all possible combinations is m ⁿ . This is a typical NP-hard problem. The swarm intelligence optimization algorithm can better solve this kind of problem, and can find a service combination that is close to the optimal solution without traversing all possible solutions. Efficient solution. Aiming at these goals and requirements, this paper transforms the web service composition problem into a QoS-constrained multi-objective web service composition optimization problem, and uses the whale algorithm WOA (Whale optimization Algorithm) to optimize multiple target parameters at the same time, and finally dynamically selects a set of web service compositions. the optimal solution.

Whale optimization algorithm WOA is a new swarm intelligence algorithm proposed by Mirjalili and Lewis in 2016. The algorithm constructs a swarm intelligent optimization algorithm by simulating the spiral motion of whales and the hunting behavior of bubble nets in the process of chasing their prey. The optimization principles of WOA are divided into the following three categories: encircling behavior, hunting behavior, and searching behavior. As a new type of optimization algorithm, WOA has the advantages of simple structure, few adjustment parameters, easy implementation, and fast operation speed. It has been proved by experiments that it can be better used in this field.

For the WOA algorithm, there is currently no model applied to the service composition problem. In most models using swarm intelligence algorithms applied to service composition problems, it is difficult to balance the search speed and the optimal value: the search speed is fast but the search optimal value is poor, and vice versa.

SUMMARY OF THE INVENTION

In view of the above problems existing in the prior art, the present invention provides a whale evolution system oriented to the service composition problem. The present invention adopts the Whale Optimization Algorithm (WOA), which focuses on the optimization environment under the condition of more target interference, and the algorithm is more sensitive to the situation that it is easy to fall into the local optimum. Therefore, it has a good target optimization efficiency in a multi-peak environment, and is suitable for the optimization of Web service composition problems. Since the coordinate values of each dimension of the search agent in the WOA algorithm are continuous, and the number of each specific candidate service that constitutes the service combination is discrete, it is necessary to use a fuzzy function to convert the coordinate value of each dimension of the search agent into the corresponding Integer encoding to solve.

For realizing the above-mentioned technical purpose, the present invention adopts following technical scheme:

Whale evolution method for service composition problem, the specific steps are as follows:

Step 1: Convert the web service composition problem into a single-objective optimization problem and establish a fitness function;

Step 2: Integer encoding for the web service composition problem;

Step 3: Establish the WOA algorithm model;

Step 4: Perform integer coding on the WOA algorithm and convert it into the DWOA algorithm;

Step 5: Use the DWOA algorithm to start iterative optimization of the problem model;

Step 6: End the search, and output the searched optimal service combination coding sequence and the current optimal web service combination QoS value.

Preferably, step 1 is as follows: first normalize the QoS values of different web service attributes by formula (1), then set the weight value of each service attribute according to the required optimization constraints, and use formula (2) Allocate the weight of each service attribute QoS value to get the final fitness function;

f(x)=F(minT(s), minA(s), minS(s), minR(s),...)(2)

Among them, i is the service type, j is the jth seed service QoS _i,j (S) is the attribute value before normalization, QoS' _i,j (S) is the attribute value after normalization, QoS _{j_max} and QoS _{j_min} respectively is the maximum and minimum attribute values of this class; in formula (2), T is the response time, A is the effectiveness, S is the success rate, R is the stability, and belongs to the normalized QoS value of each service attribute of the web.

Preferably, step 2 is as follows: each web service combination is formed by selecting m _d specific candidate services from n types of abstract services, arranging various abstract services under the sequential structure workflow, and finally obtaining a series of codes Combination; wherein, the transformation range of a single code is determined by the number of candidate services for each type of abstract service in the actual situation, and the value is from 1 to m _d and is an integer, so the search range of each dimension is determined by the actual coding range: ub means search The upper bound is generally equivalent to m _d ; lb represents the search lower bound, generally a fixed value of 1.

Preferably, step 3 is as follows: set the maximum number of iterations T _max , and the initial iteration value t=1; the model consists of three behavior calculation principles:

Surrounding behavior: Formula (3) is the update formula for the individual position, formula (4) is the distance difference between the remaining individuals and the target position, and the expression is as follows:

X(j+1)=X ^* (j)-A·D (3)

D=|C·X ^* (j)-X(j)| (4)

Among them, X(j+1) is the next movement position of the current individual, j is the number of iterations, and X ^* (j) is the optimal individual position of each generation; the expressions of A and C are as follows:

A=2a·r _a -a (5)

C=2·r _c (6)

Among them, _ra and _rc are two random vectors with the value [0, 1], the convergence factor a decreases linearly from 2 to 0, and the expression is as follows:

Hunting behavior: Equation (8) is the distance between the i-th individual and the optimal individual, and Equation (9) is the bubble net hunting behavior of the search agent;

D=|C·X ^* (j)-X(j)| (8)

X(j+1)=D·e ^bl ·cos(2πl)+X ^* (j) (9)

Among them, b is a spiral linear parameter, l is a random parameter variable, and its value is between [-1, 1];

In order to ensure that the search agent shrinks and surrounds and the spiral approximation is carried out synchronously in this process, it is realized by the random number P, and the expression is as follows:

Search behavior: At this time |A|≥1, the algorithm selects a random individual X _rand as the target, and the rest of the search individuals move toward it; Equation (11) is the search behavior of the search agent, and Equation (12) is the relationship between the i-th individual and this generation. the distance between selected random individuals;

X(j+1)=X _rand (j)-A·D (11)

D=|C·X _rand (j)-X(j)| (12).

Preferably, step 4 is as follows: the number of search agents is set to n, and the coordinate value range of a certain dimension is an integer from 1 to m _d ; the fuzzy function fd of formula (13) is used to convert the coordinate values of each dimension of the search agent For the corresponding integer value, update the coordinates of the previous generation to integer coordinates, so as to convert the coordinates of each dimension of each search agent into the corresponding integer encoded coordinates according to the coding sequence of the web service combination; according to the service combination coding string of the search agent, Substitute into the objective function to calculate the current fitness value of each search agent, compare and update the position coordinates of the search agent with the best current fitness as X ^* ;

Among them, x ^t _i,d is the coordinate value of the t-th generation of search agent i on the d-th dimension, z is an integer on [1,m _d ]; the random variable Y is the result of a Bernoulli experiment with a probability of 0.5 , the value of the iff(P,u,v) function depends on whether the P proposition is true, if so, it is u, otherwise it is v; the iff(P,u,v) function is introduced.

Preferably, step 5 is as follows: update the random parameter variables l and P, where l is used to control the amplitude of the spiral motion, and its value is between [-1, 1], and P is used to control the position update, which is [0, 1] ]; compare the values of the judgment parameters P and A respectively, determine the position update expression of the next generation search agent, and update the position of the search agent by formula (14).

Preferably, step 6 is specifically as follows: update the number of iterations t; compare with the maximum number of iterations T _max , if t≤T _max , go back to step 3; if t>T _max , end the search, and output the optimal search service The combined coding sequence and the optimal web service at this time combine the QoS value for the user to call the actual service combination.

The invention also discloses a whale evolution system oriented to the service composition problem, including the following modules:

The fitness function building module converts the web service composition problem into a single-objective optimization problem and establishes the fitness function;

Integer encoding module, which performs integer encoding for web service composition problems;

WOA algorithm model establishment module, establish WOA algorithm model;

The conversion module performs integer coding on the WOA algorithm and converts it into the DWOA algorithm;

The optimization module uses the DWOA algorithm to start iterative optimization of the problem model;

The output module ends the search, and outputs the searched optimal service combination coding sequence and the current optimal web service combination QoS value.

The corresponding relationship between DWOA algorithm and web service composition problem is further explained. The optimization objective function of the algorithm is the fitness function of the web composition service; the coordinates of each search agent of the algorithm correspond to the integer service composition code of the problem model; the coordinates of each dimension of the search agent correspond to various abstract services in the problem model; the coordinates of each dimension of the search agent The value corresponds to the number of sub-services in the abstract service; the optimal search agent corresponds to the current optimal web service composition; the optimal fitness value corresponds to the optimal comprehensive attribute value of the web service composition.

Since most of the models that use swarm intelligence algorithms are applied to service composition problems, it is difficult to balance the search speed and the optimal value: the search speed is fast but the optimal value searched is poor, and vice versa. The WOA algorithm adopted in the present invention can better balance these two objectives.

When the order of magnitude of the service combination sequence reaches a certain scale, the use of the present invention to solve the optimal service combination will significantly improve the optimal value of the random search method, and it will have higher optimization efficiency compared with the traversal search method. According to the characteristic requirements of reliability and real-time performance in the optimization problem of web service composition, it provides better composition services.

Description of drawings

Figure 1 shows the method of mapping from concrete candidate services of various abstract services to service combination sequences, which corresponds to the coding sequence in the algorithm. Among them, n represents different abstract service classes, and m represents the number of specific candidate services in the abstract service class. It should be noted that, for the convenience of tabulation, m _d described in step 1 is uniformly represented by m.

FIG. 2 is an example of web composition service, which is a workflow diagram of sequential web service composition. Each circle represents a single service, starting work from Start and entering the service workflow T ₁ , accessing various types of services according to the workflow setting sequence, and ending the service workflow End after T _n . In practical applications, users can choose the required control flow structure by themselves.

Fig. 3 is a block diagram of the whale evolution system oriented to the service composition problem of the present invention.

Fig. 4 is a flow chart of the whale evolution method oriented to the service composition problem of the present invention.

Fig. 5 is a DWOA flow chart of service-oriented composition of a preferred embodiment.

Detailed ways

The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Example 1

As shown in Figure 4-5, this embodiment is a whale evolution method for the service composition problem. Since each workflow structure of the service composition can be transformed into a sequence structure, the model is implemented according to the sequence structure shown in Figure 2, and the steps are as follows:

Step 1: Convert the web service composition problem into a single-objective optimization problem and establish a fitness function. First, the QoS values of different web service attributes are normalized by formula (1), the sum of each attribute value is calculated according to the sequence structure in Table 1, and finally the weight value of each service attribute is set according to the required optimization constraints. Equation (2) assigns the weight of each service attribute QoS value to obtain the final fitness function.

Table 1 Aggregation calculation formula of QoS value

f(x)=F(minT(s), minA(s), minS(s), minR(s),...) (2)

where i is the service type, j is the jth seed service QoS _i,j (S) is the attribute value before normalization, QoS' _i,j (S) is the attribute value after normalization, QoS _{j_max} and QoS _{j_min} are respectively The maximum and minimum property values for this class. In formula (2), T is the response time, A is the effectiveness, S is the success rate, and R is the stability, which belongs to the normalized QoS value of each service attribute of the web.

Step 2: Integer encoding of the web service composition problem. Each web service combination is composed of m _d specific candidate services selected from n types of abstract services, and the various abstract services are arranged under the sequential structure workflow, and finally a series of coding combinations are obtained. The transformation range of a single code is determined by the number of candidate services for each type of abstract service in the actual situation, and the value ranges from 1 to m _d and is an integer. Therefore, the search range of each dimension of the algorithm is determined by the actual coding range: ub represents the upper bound of the search, which is generally equivalent to m _d ; lb represents the lower bound of the search, which is generally a fixed value of 1.

Step 3: Establish the WOA algorithm model. The maximum number of iterations T _max of the algorithm is set, and the initial iteration value t=1. The model consists of three behavioral calculation principles.

Surrounding behavior: Formula (3) is the update formula for the individual position, formula (4) is the distance difference between the remaining individuals and the target position, and the expression is as follows:

X(j+1)=X ^* (j)-A·D (3)

D=|C·X ^* (j)-X(j)| (4)

where X(j+1) is the next movement position of the current individual, j is the number of iterations, and X ^* (j) is the optimal individual position for each generation. The expressions for A and C are as follows:

A=2a·r _a -a (5)

C=2·r _c (6)

where _{r a} and rc are two random vectors with values [0,1], the convergence factor a decreases linearly from 2 to 0, and its expression is as follows:

Hunting behavior: Equation (8) is the distance between the ith individual and the optimal individual, and Equation (9) is the bubble net hunting behavior of the search agent.

D=|C·X ^* (j)-X(j)| (8)

X(j+1)=D·e ^bl ·cos(2πl)+X ^* (j) (9)

where b is a spiral linear parameter, and l is a random parameter variable with a value between [-1, 1].

In order to ensure that the search agent shrinks and surrounds and the spiral approximation is carried out synchronously in this process, the algorithm is realized by the random number P, and the expression is as follows:

Search behavior: At this time |A|≥1, the algorithm selects a random individual X _rand as the target, and the rest of the search individuals move toward it. Equation (11) is the search behavior of the search agent, and Equation (12) is the distance between the i-th individual and the random individual selected by the generation.

X(j+1)=X _rand (j)-A·D (11)

D=|C·X _rand (j)-X(j)| (12)

Step 4: Perform integer coding on the WOA algorithm and convert it into a DWOA (Discrete WhaleOptimization Algorithm) algorithm. The number of search agents is set to n, and the coordinates of one dimension range from an integer from 1 to m _d . Use the fuzzy function fd of formula (13) to convert the coordinate values of each dimension of the search agent into corresponding integer values, and update the coordinates of the previous generation to integer coordinates, so as to convert the coordinates of each dimension of each search agent according to the coding sequence of the web service combination. Code the coordinates for the corresponding integer type. According to the service combination code string of the search agent, it is brought into the objective function to calculate the current fitness value of each search agent, and the position coordinates of the search agent with the best current fitness are compared and updated as X ^* .

where x ^t _i,d is the coordinate value of the t-th generation of search agent i on the d-th dimension, and z is an integer in [1,m _d ]. The random variable Y is the result of a Bernoulli experiment with a probability of 0.5. The value of the iff(P, u, v) function depends on whether the P proposition is true, if it is u, otherwise it is v. The iff(P,u,v) function is introduced. The main reason is that the coordinates of the search agent will exceed the boundary when the WOA algorithm is actually applied. The introduction of this function prevents the coordinates of the search agent from being encoded as illegal codes after the boundary is exceeded. .

Step 5: Update the random parameter variables l and P, where l is used to control the amplitude of the spiral motion, and its value is between [-1, 1], and P is used to control the position update, which is a random value between [0, 1]. number. Comparing the values of the judgment parameters P and A respectively, the position update expression of the next generation search agent is determined, and the position of the search agent is updated by formula (14).

Step 6: Update the number of iterations t. Compare with the maximum number of iterations T _max , if t≤T _max , go back to step 3; if t>T _max , end the search, and output the searched optimal service combination code sequence and the optimal web service combination QoS at this time The value for the user to actually call the service composition.

Example 2

As shown in FIG. 3 , the whale evolution system for the service composition problem in this embodiment includes the following modules:

Fitness function establishment module: Convert the web service composition problem into a single-objective optimization problem, and establish a fitness function. First, the QoS values of different web service attributes are normalized by formula (1), the sum of each attribute value is calculated according to the sequence structure in Table 1, and finally the weight value of each service attribute is set according to the required optimization constraints. Equation (2) assigns the weight of each service attribute QoS value to obtain the final fitness function.

Table 1 Aggregation calculation formula of QoS value

f(x)=F(minT(s), minA(s), minS(s), minR(s),...) (2)

where i is the service type, j is the jth seed service QoS _i,j (S) is the attribute value before normalization, QoS' _i,j (S) is the attribute value after normalization, QoS _{j_max} and QoS _{j_min} are respectively The maximum and minimum property values for this class. In formula (2), T is the response time, A is the effectiveness, S is the success rate, and R is the stability, which belongs to the normalized QoS value of each service attribute of the web.

Integer encoding module: Integer encoding for web service composition problems. Each web service combination is composed of m _d specific candidate services selected from n types of abstract services, and the various abstract services are arranged under the sequential structure workflow, and finally a series of coding combinations are obtained. The transformation range of a single code is determined by the number of candidate services for each type of abstract service in the actual situation, and the value ranges from 1 to m _d and is an integer. Therefore, the search range of each dimension of the algorithm is determined by the actual coding range: ub represents the upper bound of the search, which is generally equivalent to m _d ; lb represents the lower bound of the search, which is generally a fixed value of 1.

WOA algorithm model building module: establish the WOA algorithm model. The maximum number of iterations T _max of the algorithm is set, and the initial iteration value t=1. The model consists of three behavioral calculation principles.

Surrounding behavior: Formula (3) is the update formula for the individual position, formula (4) is the distance difference between the remaining individuals and the target position, and the expression is as follows:

X(j+1)=X ^* (j)-A·D (3)

D=|C·X ^* (j)-X(j)| (4)

where X(j+1) is the next movement position of the current individual, j is the number of iterations, and X ^* (j) is the optimal individual position for each generation. The expressions for A and C are as follows:

A=2a·r _a -a (5)

C=2vr _c (6)

where _{r a} and rc are two random vectors with values [0,1], the convergence factor a decreases linearly from 2 to 0, and its expression is as follows:

Hunting behavior: Equation (8) is the distance between the ith individual and the optimal individual, and Equation (9) is the bubble net hunting behavior of the search agent.

D=|C·X ^* (j)-X(j)| (8)

X(j+1)=D·e ^bl ·cos(2πl)+X ^* (j) (9)

where b is a spiral linear parameter, and l is a random parameter variable with a value between [-1, 1].

In order to ensure that the search agent shrinks and surrounds and the spiral approximation is carried out synchronously in this process, the algorithm is realized by the random number P, and the expression is as follows:

Search behavior: At this time |A|≥1, the algorithm selects a random individual X _rand as the target, and the rest of the search individuals move toward it. Equation (11) is the search behavior of the search agent, and Equation (12) is the distance between the i-th individual and the random individual selected by the generation.

X(j+1)=X _rand (j)-A·D (11)

D=|C·X _rand (j)-X(j)| (12)

Conversion module: Encode the WOA algorithm in integer type and convert it into DWOA (Discrete WhaleOptimization Algorithm) algorithm. The number of search agents is set to n, and the coordinates of one dimension range from an integer from 1 to m _d . Use the fuzzy function fd of formula (13) to convert the coordinate values of each dimension of the search agent into corresponding integer values, and update the coordinates of the previous generation to integer coordinates, so as to convert the coordinates of each dimension of each search agent according to the coding sequence of the web service combination. Code the coordinates for the corresponding integer type. According to the service combination code string of the search agent, it is brought into the objective function to calculate the current fitness value of each search agent, and the position coordinates of the search agent with the best current fitness are compared and updated as X ^* .

where x ^t _i,d is the coordinate value of the t-th generation of search agent i on the d-th dimension, and z is an integer in [1,m _d ]. The random variable Y is the result of a Bernoulli experiment with a probability of 0.5. The value of the iff(P, u, v) function depends on whether the P proposition is true, if it is u, otherwise it is v. The iff(P,u,v) function is introduced. The main reason is that the coordinates of the search agent will exceed the boundary when the WOA algorithm is actually applied. The introduction of this function prevents the coordinates of the search agent from being encoded as illegal codes after the boundary is exceeded. .

Optimization module: update random parameter variables l, P, where l is used to control the amplitude of the spiral motion, the value is between [-1, 1], and P is used to control the position update, which is between [0, 1] random number. Comparing the values of the judgment parameters P and A respectively, the position update expression of the next generation search agent is determined, and the position of the search agent is updated by formula (14).

Output module: update the number of iterations t. Compare with the maximum number of iterations T _max , if t ≤ T _max , return to the WOA algorithm model building module for processing; if t > T _max , end the search, and output the search optimal service combination code sequence and the optimal web The service composition QoS value is used for the actual invocation of the service composition by the user.

It should be pointed out that the present invention is not limited to the above-mentioned specific implementations, such as the selection of service evaluation criteria, the setting of the weight ratio of the fitness function, etc., and those skilled in the art can make various changes within the scope of the claims. or modification, which does not affect the essential content of the present invention.

Claims (1)

1. A whale evolution method for service combination problems is characterized by comprising the following specific steps:

step 1: converting a web service combination problem into a single-target optimization problem, and establishing a fitness function;

step 2: carrying out integer coding on the web service combination problem;

and step 3: establishing a WOA algorithm model;

and 4, step 4: carrying out integer coding on the WOA algorithm, and converting the WOA algorithm into a DWOA algorithm;

and 5: starting iterative optimization on the problem model by using a DWOA algorithm;

and 6: finishing the search, and outputting the optimal service combination coding sequence and the current optimal web service combination QoS value;

the step 1 is as follows: firstly, carrying out normalization processing on QoS values of different web service attributes through an equation (1), then setting a weight value of each service attribute according to a required optimization constraint condition, and distributing the weight of each service attribute QoS value through an equation (2) to obtain a final fitness function;

f(x)＝F(minT(s),minA(s),minS(s),minR(s),...) (2)

wherein i is the service type, and j is the jth seed service QoS _i,j (S) is normalized Pre-Attribute value, QoS' _i,j (S) is normalized Attribute value, QoS _{j_max} And QoS _{j_min} The maximum and minimum attribute values of the class are respectively; in the formula (2), T is response time, A is effectiveness, S is success rate, R is stability, and the value belongs to the normalized QoS value of each service attribute of the web;

the step 2 is as follows: each web service combination is m selected from n types of abstract services _d Each specific candidate service is formed, various abstract services are arranged under the sequential structure working flow, and a string of coding combinations is finally obtained; wherein, the transformation range of the single code is determined by the number of candidate services of each kind of abstract service in actual conditions, and the value is from 1 to m _d And is an integer, so the search range of each dimension is determined by the actual coding range: ub denotes the upper search bound, which is equal to m _d (ii) a lb represents the lower search bound and is constant 1;

the step 3 is as follows: setting a maximum number of iterations T _max The iteration initial value t is 1; the model consists of three behavioral calculation principles:

bounding behavior: formula (3) is used to update the formula for the location of the individual, and formula (4) is used to update the distance difference between the remaining individual and the target location, and the expression is as follows:

X(j+1)＝X ^* (j)-A·D (3)

D＝|C·X ^* (j)-X(j)| (4)

wherein X (j +1) is the next motion position of the current individual, j is the number of iterations, X ^* (j) An optimal individual position for each generation; the expressions of A and C are as follows:

A＝2a·r _a -a (5)

C＝2·r _c (6)

wherein r is _a And r _c Are two values of [0,1]]The convergence factor a decreases linearly from 2 to 0, and the expression is as follows:

hunting behaviors: formula (8) is the distance between the ith individual and the optimal individual, and formula (9) is the bubble trap hunting behavior of the search agent;

D＝|C·X ^* (j)-X(j)| (8)

X(j+1)＝D·e ^bl ·cos(2πl)+X ^* (j) (9)

wherein, b is a spiral linear parameter, l is a random parameter variable, and the value is between [ -1,1 ];

in order to ensure that the search agent contracts and surrounds and is synchronously close to the spiral type in the process, the process is realized by a random number P, and the expression is as follows:

and (3) searching action: at the moment, the absolute value of A is more than or equal to 1, and the algorithm selects a random individual X _rand As a target, the rest of the search individuals move towards it; equation (11) is the search behavior of the search agent, and equation (12) is the distance between the ith individual and the random individual selected by the agent;

X(j+1)＝X _rand (j)-A·D (11)

D＝|C·X _rand (j)-X(j)| (12)；

the step 4 is as follows: setting the number of search agents as n, wherein the value range of the coordinate of a certain dimension is 1-m _d An integer of (d); converting each dimension coordinate value of the search agent into a corresponding integer value by using a fuzzy function fd shown in a formula (13), updating the previous generation coordinate into an integer type coordinate, and converting each dimension coordinate of each search agent into a corresponding integer type coding coordinate according to a web service combination coding sequence; respectively substituting the service combination code strings of the search agents into a target function to calculate the current fitness value of each search agent, and comparing and updating the position coordinate of the search agent with the best current fitness as X ^* ；

Wherein x is ^t _i,d Is the coordinate value of the search agent i in the d-dimension of the t-th generation, z is [1, m _d ]The above integer; the random variable Y is a Bernoulli test result with the probability of 0.5 once, and the value of the iff (P, u, v) function depends on whether the P proposition is true, if so, u, otherwise, v; introducing an iff (P, u, v) function;

the step 5 is as follows: updating random parameter variables l and P, wherein l is used for controlling the amplitude of the spiral motion and takes a value between [ -1 and 1], and P is used for controlling the position update and is a random number between [0 and 1 ]; respectively comparing the values of the judgment parameters P, A, determining a position updating expression of a next generation search agent, and updating the position of the search agent by the formula (14);

the step 6 is specifically as follows: updating the iteration times t; with the maximum number of iterations T _max Comparing if T is less than or equal to T _max Returning to the step 3; if T > T _max If so, ending the search, and outputting the optimal service combination coding sequence and the optimal web service combination QoS value at the moment for the user to call the actual service combination.

Images