CN104809500A - Efficient improved FSPSO (Particle Swarm Optimization based on prey behavior of Fish Schooling) - Google Patents

Abstract

The invention discloses an efficient improved particle swarm algorithm based on the behavior of fish shoals. The invention simulates this intelligent behavior. The currently optimal particle finds the current global optimal position information through its own optimal position information provided by other few random particles. better location. When the school is attacked by other predators, the weaker fish that cannot escape quickly will be eaten. The invention simulates these behaviors, and replaces weak particles near the current global worst particle with randomly generated particles, thus enhancing the diversity of groups, and FSPSO can effectively avoid local optimal values. The invention can be specifically applied in the solving process of complex optimization problems such as function optimization, knapsack problem, traveling salesman problem, assembly line operation problem and graphic image processing.

Description

An Efficient Improved Particle Swarm Algorithm Based on the Predating Behavior of Fish Schools

technical field

The invention belongs to the technical field of evolutionary algorithms, and relates to a particle swarm optimization algorithm, in particular to an efficient improved particle swarm algorithm based on the behavior of fish rushing for food.

Background technique

Particle Swarm Optimization (PSO) is a novel evolutionary algorithm proposed by Kennedy and Eberhart through the observation and research of certain social behaviors of birds. Since PSO has few setting parameters and is easy to implement, it can be used to solve a large number of complex optimization problems such as nonlinear, non-differentiable and multi-peak, and has been successfully applied to parameter optimization, feature extraction, traveling salesman problem, pattern recognition and many other fields. However, in complex optimization problems, the particle swarm optimization algorithm has premature convergence, and it is easy to fall into a local optimal solution. Angeline et al. proposed the concept of hybrid PSO algorithm by referring to the idea of genetic algorithm, which improved the convergence speed and accuracy of the algorithm. The fitness-distance-proportion algorithm based on particle swarm optimization (Fitness-Distance-Ratio based Particle Swarm Optimization, FDR-PSO) proposed by Peram et al. in 2003, each particle in the algorithm is based on a certain fitness value-distance- The principle of proportionality is to approach multiple particles with better fitness values nearby to different degrees, not just to the best particle currently found. This algorithm improves the premature convergence problem of PSO algorithm, and its performance has been greatly improved in optimizing complex functions. Bergh proposed a cooperative PSO (Cooperative Particle Swarm Optimizer, CPSO), which makes it easier for particles to jump out of local minimum points and achieve higher convergence accuracy. In order to prevent the PSO algorithm from falling into a local optimum, Liang et al. proposed the Comprehensive Learning Particle Swarm Optimizer (CLPSO) in 2006, so that the velocity update of each particle is based on the historical optimal position of all other particles, so that To achieve the purpose of comprehensive learning. Bergh proposed the cooperative PSO (Cooperative Particle Swarm Optimizer, CPSO), which makes it easier for particles to jump out of local minimum points and achieve higher convergence accuracy. In order to prevent the PSO algorithm from falling into a local optimum, Liang et al. proposed the Comprehensive Learning Particle Swarm Optimizer (CLPSO) in 2006, so that the velocity update of each particle is based on the historical optimal position of all other particles. In order to achieve the purpose of comprehensive learning. By analyzing the probability distribution of the search center, Wu Xiaojun proposed a uniform search PSO (Uniform Search Particle Swarm Optimization, UPSO) algorithm in which the search center is evenly distributed between two extreme values. This algorithm has high search efficiency and good convergence performance.

The above algorithms are still prone to fall into local optimum when optimizing complex high-dimensional multi-modular functions.

Contents of the invention

In order to solve the above-mentioned technical problems, the present invention simulates the behavior of fish shoals, and proposes an efficient improved particle swarm optimization algorithm based on the fish swarm behavior.

The technical scheme adopted in the present invention is: an efficient improved particle swarm algorithm based on the fish swarming behavior, which is characterized in that it includes the following steps:

Step 1: Initialize the parameters of an efficient improved particle swarm optimization algorithm based on the behavior of fish swarms, the parameters include the number of groups n, the maximum number of iterations maxk, inertia weight w, learning factors c ₁ and c ₂ , and the number of search particles m, search factor c ₃ and range factor c ₄ , where 0≤m≤n/10;

Step 2: Update the position and velocity of each particle in the improved particle swarm;

Step 3: The currently globally optimal particle finds a current globally optimal position through its own optimal position information provided by m random particles;

Step 4: Replace the weak particles near the current global worst particle with randomly generated particles;

Step 5: Judgment, whether the efficient improved particle swarm optimization algorithm based on fish predation behavior converges or reaches the maximum number of iterations?

If so, output the position of the global optimal solution, which is the solution of the optimization problem;

If not, go back to step 2.

As a preference, the specific implementation process of step 3 is to randomly select m particles, and their own optimal positions are: P ₁ , P ₂ ,..., P _m , the search position of the current global optimal particle in the jth direction for:

XX _j ＝P _g +r ₃ ·c ₃ ·(P _j -P _g ), where 1≤j≤n, r ₃ is a uniformly distributed random number between [0,1], c ₃ is the search factor; Judgment: If the optimal value in XX _j is better than P _g , then replace P _g with this optimal value; otherwise, the value of P _g remains unchanged. The current global optimal value improves the global search ability of the algorithm by learning from the own optimal value of a few particles.

Preferably, for the weak particles described in step 4, if the position of the global worst particle group is: P _b =(P _b1 ,P _b2 ,...,P _bD ), the position of the jth weak particle is Y _j , since the weak particle is near the global worst particle, the distance between Y _j and P _b satisfies the relation where c ₄ is the range factor, and Range is the search range of the particle. Weak particles are replaced by randomly generated particles, which enhances the diversity of the population and can effectively avoid local optimal values.

An efficient improved particle swarm optimization based on prey behavior of Fish Schooling (FSPSO for short) proposed by the present invention. The school of fish can transmit enlightenment information through the water wave, and find a better position by sensing the water wave information during the predation. The present invention simulates this kind of intelligent behavior, and the currently globally optimal particle searches for a current globally optimal position through its own optimal position information provided by other few random particles. When the school is attacked by other predators, the weaker fish that cannot escape quickly will be eaten. The invention simulates these behaviors, and replaces weak particles near the current global worst particle with randomly generated particles, thus enhancing the diversity of groups, and FSPSO can effectively avoid local optimal values. The invention can be specifically applied in the solving process of complex optimization problems such as function optimization, knapsack problem, traveling salesman problem, assembly line operation problem and graphic image processing.

Description of drawings

Fig. 1: is the global optimal particle search pattern of the embodiment of the present invention;

Figure 2: is a schematic diagram of weak and small particles near the global worst particle in the embodiment of the present invention;

Fig. 3: is the convergence diagram for the function f ₂ (x) of the embodiment of the present invention;

Fig. 4: is the convergence diagram for the function f ₇ (x) of the embodiment of the present invention;

Fig. 5: is the convergence diagram for the knapsack problem of the embodiment of the present invention.

Detailed ways

In order to facilitate those of ordinary skill in the art to understand and implement the present invention, the present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the implementation examples described here are only used to illustrate and explain the present invention, and are not intended to limit this invention.

Assuming that the number of particles in the PSO algorithm is n, the search space is D-dimensional, and the position and velocity of the i-th particle are x _i and v _i respectively, then:

x _i = (x _i1 ,...,x _id ,...,x _iD ) (1),

v _i = (v _i1 ,...,v _id ,...,v _iD ) (2);

The position of the particle with the best fitness in the population is recorded as:

P _g ＝(P _g1 ,P _g2 ,...,P _gD ) (3);

The best position of the solution space experienced by the i-th particle is:

P _i = (P _i1 ,P _i2 ,...,P _iD ) (4);

Then the equation of particle swarm algorithm position and velocity update is:

${\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; w</span>}{\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; gd</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

Among them, w is the inertia weight, k is the number of iterations, c ₁ and c ₂ are learning factors, r ₁ and r ₂ are random numbers uniformly distributed in the interval [0, 1]. The variation ranges of the position and velocity of the dth particle are [XMIN _d , XMAX _d ] and [VMIN _d , VMAX _d ] respectively. If the value calculated by formulas (5) and (6) exceeds this range, set it as the boundary value.

The study found that fish schools can generate complex water waves, and the heuristic information of predation can be transmitted through the water waves during the predation behavior. To simulate this kind of intelligent behavior, the currently optimal particle finds a better position globally through the information of its own optimal position provided by other random particles.

Without loss of generality, m particles that provide heuristic information are randomly selected, and their own optimal positions are:

P ₁ ,P ₂ ,...,P _m (7);

The direction of the current global optimal particle search for better particles is shown in Figure 1, and the search position of the jth direction of the current global optimal particle search is:

XX _j =P _g +r ₃ ·c ₃ ·(P _j -P _g ) (8);

Where 1≤j≤n, _r3 is a uniformly distributed random number between [0,1], and _c3 is a search factor. If the optimal value in XX _j is better than P _g , then replace P _g with this optimal value; otherwise, the value of P _g remains unchanged. In order to reduce the amount of calculation, the value of m should not be too large. The current global optimal value improves the global search ability of the algorithm by learning from the own optimal value of a few particles.

Fish schools will be attacked by other predators during the feeding process. According to Darwin's theory of the jungle, weak fish that cannot escape quickly will be eaten. The invention simulates these behaviors, and replaces weak particles near the current global worst particle with randomly generated particles.

The position of the global worst particle swarm is:

P _b = (P _b1 ,P _b2 ,...,P _bD ) (9);

The position of the jth weak particle is Y _j , since the weak particle is near the global worst particle, the distance between Y _j and P _b is very small, and the relationship between them is shown in Figure 2. Y _j is in the circle centered on P _b , then the relationship between them satisfies:

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; D.</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; the\; ji</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; bi</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; Range</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

where c ₄ is the range factor, and Range is the search range of the particle. Weak particles are replaced by randomly generated particles, which enhances the diversity of the population, and FSPSO can effectively avoid local optimal values.

The present invention proposes the FSPSO algorithm by simulating the fish rushing behavior, and the specific steps are as follows:

Step 1: Initialize the parameters of the FSPSO algorithm. The parameters include the number of groups n, the maximum number of iterations maxk, the inertia weight w, the learning factors c ₁ and c ₂ , the number of search particles m, the search factor c ₃ and the range factor c ₄ , where 0 ≤m≤n/10;

Step 2: Update the position and velocity of each particle in the group according to equations (5) and (6);

Step 3: The current globally optimal particle finds the current globally optimal position through the information of its own optimal position provided by m random particles, and improves according to formula (8);

Step 4: Replace the weak particles near the current global worst particle with randomly generated particles, and improve according to formula (10);

Step 5: Determine whether the FSPSO algorithm converges or reaches the maximum number of iterations?

If so, output the position of the global optimal solution, which is the solution of the optimization problem;

If not, go back to step 2.

In the FSPSO algorithm, the time complexity of calculating the current optimal particle P _g , the worst particle P _b and each particle’s own optimal position _Pi is O(n*D), and the weak particles near the worst particle are replaced by random particles. The time complexity of the particle is O(n*D), and the time complexity of the current global optimal particle searching for a better position is O(m*n*D). Since the value of m is relatively small, the time complexity of FSPSO is O(n*D*maxk), which is the same as that of PSO. In FSPSO, the positions of P _i , P _g and P _b need to be stored. Therefore, the space complexity of FSPSO is O(n*D), which is the same as that of PSO. To sum up, the time and space complexity of FSPSO is the same as that of PSO, so the performance of FSPSO is very good.

In order to evaluate the performance of FSPSO, the algorithm is used to optimize nine standard test functions. The test functions are shown in Table 1. The experimental development environment is as follows: Matlab R2012b, CPU is i7-2600k3.4GHz.

Table 1 Test function

In Table 1, f ₁ (x)-f ₆ (x) are single-mode functions, and f ₇ (x)-f ₉ (x) are multi-mode functions. In these functions, N is the dimension of the function, the minimum value of these functions is 0, and the optimal solution is [0] ^N . FSPSO, PSO and GA are used to optimize the above functions. In these three algorithms, the dimension is 30 (N=D=30), the number of populations is 50, and the maximum number of iterations is 1000. In PSO, c ₁ =c ₂ =2, and the inertia factor w decreases linearly with the number of iterations from 0.9 to 0.7. The appropriate parameters of FSPSO are found through comparative experiments as follows: stepa=1.5, stepb=0.0001, m=4, and the settings of other parameters of FSPSO are the same as in PSO. The parameters of GA are set as follows: the crossover probability is 0.95, the random selection mechanism is adopted, and the probability of Gaussian mutation is 0.1. The optimization experiment of each function is run 30 times, and the optimal value, average value, median value and variance of the 30 solutions are shown in Table 2.

Table 2 Results of three algorithm optimization functions

It can be seen from Table 2 that the optimal solutions of FSPSO for all functions are better than those of GA and PSO, so the global search ability of FSPSO is better than the other two algorithms. The average value of FSPSO is better than PSO for all functions, except for f ₁ (x), f ₃ (x) and f ₈ (x), the average value of FSPSO is better than GA. Therefore, the overall search performance of FSPSO is better than the other two algorithms. Except for f ₇ (x), the median value of FSPSO is smaller than that of PSO, and except for f ₃ (x) and f ₈ (x), the median value of FSPSO is smaller than that of GA. The variance of FSPSO is smaller than that of PSO except for f ₇ (x), and the variance of FSPSO is smaller than that of GA except for f ₁ (x), f ₃ (x) and f ₈ (x). Therefore, the stability of FSPSO is better than the other two algorithms.

The convergence curves are used to evaluate the performance of the three algorithms. In order not to lose generality, the single-mode function f ₂ (x) and the multi-mode function f ₇ (x) are randomly selected for analysis. In order to better analyze the convergence curve, the value of the solution in the iteration is logarithmic, and the convergence curves of the three algorithms are shown in Figure 3 and Figure 4. It can be seen from the convergence curve that, compared with the other two algorithms, FSPSO has better global search ability and convergence. The number of iterations for FSPSO to find the global optimal value is less than other algorithms, so the convergence efficiency is higher. Compared with PSO, FSPSO is more effective. Therefore, FSPSO shows a significant improvement in the optimization function, and the overall performance is also very good.

The knapsack problem is a typical NP-hard problem. GA, PSO and FSPSO are used to optimize a knapsack problem. In this problem, the value matrix c and volume matrix w are as follows:

c＝[220,208,198,192,180,180,165,162,160,158,155,130,125,122,120,118,115,110,105,101,100,100,98,96,95,90,88,82,80,77,75,73,72,70,69,66,65,63,60,58,56,50,30,20,15 ,10,8,5,3,1]

w=[80,82,85,70,72,70,66,50,55,25,50,55,40,48,50,32,22,60,30,32,40,38,35,32 ,25,28,30,22,50,30,45,30,60,50,20,65,20,25,30,10,20,25,15,10,10,10,4,4,2 ,1]

The volume limit is 1000. Because the knapsack problem is a discrete problem, the discretization method of each particle position in the present invention is:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; if</span>}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; sig</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; else</span>}& \end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 11</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

$\mathrm{<span\; class="notranslate">\; sig</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; id</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

Where r ₄ is a random number uniformly distributed between [0,1]. The experimental environment and algorithm parameter settings are the same as in the function optimization problem. The iterative curves of the three algorithms for solving the knapsack problem are shown in Figure 5. It can be seen from Figure 5 that the global search ability of FSPSO is stronger than the other two algorithms. The experiment of optimizing the knapsack problem with three algorithms runs 30 times, and the optimal value, worst value, average value, median value and variance of the 30 solutions are shown in Table 3.

Table 3 The results of three algorithms for optimizing the knapsack problem

algorithm best the worst average middle variance GA 2909 2670 2788.5 2787 49.5 PSO 2955 2687 2820.3 2812 58.7 FSPSO 3030 2899 2957.4 2961 31.8

It can be seen from Table 3 that the optimal value, the worst value, the average value and the median value obtained by FSPSO are greater than the corresponding values obtained by GA and PSO, and the variance obtained by FSPSO is smaller than that of the other two algorithms. Therefore, the global search ability and convergence of FSPSO are better than the other two algorithms, indicating that FSPSO is more effective for the knapsack problem.

The present invention proposes an improved particle swarm algorithm FSPSO based on the behavior of fish swarms. The current globally optimal particle finds a better position through its own optimal position information provided by other few random particles, so that the global search ability of the algorithm is enhanced. up. The weak particles near the current global worst particle are replaced by randomly generated particles, which increases the diversity of the population and avoids local optimal solutions. By optimizing standard test functions and typical knapsack problems, it is proved that FSPSO is more efficient for solving both continuous and discrete optimization problems.

It should be understood that the parts not described in detail in this specification belong to the prior art.

It should be understood that the above-mentioned descriptions for the preferred embodiments are relatively detailed, and should not therefore be considered as limiting the scope of the patent protection of the present invention. Within the scope of protection, replacements or modifications can also be made, all of which fall within the protection scope of the present invention, and the scope of protection of the present invention should be based on the appended claims.

Claims (3)

1. an efficient improved particle swarm algorithm based on fish swarm behavior, is characterized in that, comprises the following steps: Step 1: Initialize the parameters of an efficient improved particle swarm optimization algorithm based on the behavior of fish swarms, the parameters include the number of groups n, the maximum number of iterations maxk, inertia weight w, learning factors c ₁ and c ₂ , and the number of search particles m, search factor c ₃ and range factor c ₄ , where 0≤m≤n/10; Step 2: Update the position and velocity of each particle in the improved particle swarm; Step 3: The currently globally optimal particle finds a current globally optimal position through its own optimal position information provided by m random particles; Step 4: Replace the weak particles near the current global worst particle with randomly generated particles; Step 5: Judgment, whether the efficient improved particle swarm optimization algorithm based on fish predation behavior converges or reaches the maximum number of iterations? If so, output the position of the global optimal solution, which is the solution of the optimization problem; If not, go back to step 2. 2. the efficient improved particle swarm algorithm based on the behavior of shoals of fish pouncing food according to claim 1, is characterized in that: the concrete realization process of step 3 is, randomly selects m particles, and their own optimum position is: P ₁ ,P ₂ ,...,P _m , the search position of the current global optimal particle in the jth direction is: XX _j ＝P _g +r ₃ ·c ₃ ·(P _j -P _g ), where 1≤j ≤n, r ₃ is a uniformly distributed random number between [0,1], c ₃ is the search factor; Judgment: If the optimal value in XX _j is better than P _g , use this optimal value to replace P _g ; otherwise , the value of P _g remains unchanged. 3. the efficient improved particle swarm algorithm based on fish swarming behavior according to claim 1, characterized in that: for the weak and small particles described in step 4, if the position of the global worst particle swarm is: P _b =( P _b1 ,P _b2 ,...,P _bD ), the position of the jth weak particle is Y _j , since the weak particle is near the global worst particle, the distance between Y _j and P _b satisfies the relation where c ₄ is the range factor, and Range is the search range of the particle.