CN110930182A - Improved particle swarm optimization algorithm-based client classification method and device - Google Patents
Improved particle swarm optimization algorithm-based client classification method and device Download PDFInfo
- Publication number
- CN110930182A CN110930182A CN201911089171.5A CN201911089171A CN110930182A CN 110930182 A CN110930182 A CN 110930182A CN 201911089171 A CN201911089171 A CN 201911089171A CN 110930182 A CN110930182 A CN 110930182A
- Authority
- CN
- China
- Prior art keywords
- particle swarm
- particle
- preset
- population
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Game Theory and Decision Science (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention provides a client classification method and a client classification device based on an improved particle swarm optimization algorithm, wherein the method comprises the following steps: initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm; performing iterative updating operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until a preset iteration number is reached; after the iteration times are preset, respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update, carrying out cross operation and mutation operation for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition; and obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients. The method can avoid trapping in local optimal solution, improve later convergence speed and improve search precision through organic fusion of genetic algorithm.
Description
Technical Field
The invention relates to the field of customer classification, in particular to a customer classification method and device based on an improved particle swarm optimization algorithm.
Background
Particle Swarm Optimization (PSO) is a new evolutionary computing technique from simulation of bird Swarm predation behavior proposed by Kennedy and Eberhart in 1995, and optimizes the search through Swarm intelligence guidance generated by cooperation and competition among Swarm particles. The algorithm is a new global intelligent optimization algorithm and has the advantages of simple realization, less parameters needing to be set, high convergence speed and the like. The main essential characteristics of the standard particle swarm optimization algorithm are high in convergence speed, simple in operation and high in universality.
The particle swarm optimization algorithm is combined with a cluster analysis method, so that the classification of customers can be realized, and the targeted management of users can be realized. However, a large amount of practice and research show that the standard particle swarm optimization algorithm has the defects of poor local search capability, "premature convergence", low later iteration efficiency and the like, cannot ensure the search of a global optimal solution, is easy to fall into a local minimum solution, and causes the clustering algorithm to easily fall into a local extreme value, so that the accuracy of the obtained classification result is low.
Disclosure of Invention
In order to solve the above problems, embodiments of the present invention provide a client classification method and device based on an improved particle swarm optimization algorithm.
In a first aspect, an embodiment of the present invention provides a client classification method based on an improved particle swarm optimization algorithm, including: initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm; performing iterative updating operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until a preset iteration number is reached; after the iteration times are preset, respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update, carrying out cross operation and mutation operation for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition; and obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients based on a clustering algorithm.
Further, the selecting the particle swarm comprises: calculating the selection probability of each individual inherited to the next generation according to a preset fitness function, and calculating the cumulative selection probability of the individual to the current iteration times; and determining each individual as a selection result of the next iteration individual according to the accumulated selection probability of each individual.
Further, the performing a crossover operation includes: according to a preset cross probability function, the individuals obtained by the selection operation are selected again; performing cross operation on the individuals obtained by re-selection and the individual optimal result and the group optimal result respectively; wherein the cross probability function is a decreasing function with respect to the number of iterations.
Further, after the performing the crossover operation and before the performing the mutation operation, the method further includes: and calculating the group fitness variance of the current iterative population, and if the group fitness variance is smaller than a preset threshold, performing variation operation.
Further, the performing mutation operations include: determining individuals subjected to mutation operation in the individuals subjected to the cross operation according to a preset mutation probability function, and performing mutation operation; wherein the variation probability function is an increasing function with respect to the number of iterations.
Further, after the performing the mutation operation, the method further includes: adjusting the inertia weight of the particle swarm algorithm according to a preset inertia weight function; wherein the inertial weight function is a non-linearly decreasing function with respect to the number of iterations.
Further, after the performing the mutation operation, the method further includes: if the position of the population individual exceeds the set maximum position value or minimum position value, fine adjustment is carried out on the position of the population individual through a preset random function; and if the speed of the population individuals exceeds the set maximum speed value or minimum speed value, fine adjustment is carried out on the speed of the population individuals through a preset random function.
In a second aspect, an embodiment of the present invention provides a client classification device based on an improved particle swarm optimization algorithm, including: the initialization module is used for initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm; the first search module is used for carrying out iterative update operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until the preset iteration times are reached; the second search module is used for respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update after the iteration times are preset, carrying out cross operation and mutation operation, and is used for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition; and the classification module is used for obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients based on a clustering algorithm.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements, when executing the computer program, the steps of the client classification method based on the improved particle swarm optimization algorithm in the first aspect of the present invention.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the client classification method based on the improved particle swarm optimization algorithm of the first aspect of the present invention.
The client classification method and device based on the improved particle swarm optimization algorithm provided by the embodiment of the invention can overcome the defects that the standard particle swarm optimization algorithm is easy to fall into the local optimal solution, the later convergence speed is low, the search precision is not high, and the like. By setting the genetic algorithm, the optimization algorithm of the particle swarm is improved to classify clients based on the clustering algorithm, and after iteration of preset times, the genetic algorithm is organically integrated, so that the situation that the clients fall into a local optimal solution can be avoided, the later convergence speed is improved, the search precision is improved, and an accurate clustering center is obtained, thereby avoiding the defect that the traditional clustering algorithm depending on the clustering center easily falls into a local extreme value, and improving the accuracy of classification.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a flowchart of a customer classification method based on an improved particle swarm optimization algorithm according to an embodiment of the present invention;
FIG. 2 is a graph illustrating genetic cross-probability adjustment for each particle according to an embodiment of the present invention;
FIG. 3 is a graph illustrating the adjustment of the genetic crossover probability of a single particle according to an embodiment of the present invention;
FIG. 4 is a graph illustrating the adjustment of genetic variation probability of each particle according to an embodiment of the present invention;
FIG. 5 is a graph illustrating the adjustment of the probability of genetic variation of a single particle according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a Logistic mapping value according to an embodiment of the present invention;
fig. 7 is a schematic diagram of distribution of Logistic mapping particles according to an embodiment of the present invention;
FIG. 8 is a schematic diagram of a Logistic probability density function according to an embodiment of the present invention;
FIG. 9 is a diagram illustrating improved Logistic mapping values provided by an embodiment of the present invention;
FIG. 10 is a schematic diagram of improved Logistic mapping particle distribution provided by an embodiment of the present invention;
fig. 11 is a structural diagram of a client classification device based on an improved particle swarm optimization algorithm according to an embodiment of the present invention;
fig. 12 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The genetic algorithm is a group-based parallel search method proposed by Holland in 1975, and the group forms excellent genes through continuous selection, intersection and variation and transmits the excellent genes to the next generation, so that knowledge about a search space can be automatically acquired and accumulated in the search process, and the search process is adaptively controlled, thereby obtaining an optimal solution or a quasi-optimal solution. The genetic algorithm and the particle swarm optimization algorithm have strong complementarity: the particle swarm optimization has the advantages that the function of the optimized object does not require resolvability, the searching speed is high, but the problems of easy falling into local extremum, low optimizing precision and the like exist; the genetic algorithm has extremely strong search precision and evolution function, is simple and universal, has strong robustness, is suitable for parallel processing, and has advantages in global optimization.
Fig. 1 is a flowchart of a client classification method based on an improved particle swarm optimization algorithm according to an embodiment of the present invention, and as shown in fig. 1, the embodiment of the present invention provides a client classification method based on an improved particle swarm optimization algorithm, including:
101. and initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm.
The mixed particle swarm of the embodiment is composed of chromosomes, each chromosome is an individual, and each individual corresponds to an initial solution of an optimization problem. The number of individuals is called the size of the population or the size of the population, and the number of individuals usually adopts a constant and can also be overlapped according to a certain strategy for a plurality of timesBut is changed. The ith individual position in the population is X ═ Xi1,xi2,…,xiD) Velocity is V ═ Vi1,vi2,…,viD) The population scale is N, and the value of i is [1, N]And D is the number to be classified.
Coding is the primary problem of hybrid particle swarm optimization algorithms, which is to convert the feasible solution of a problem from its solution space to a search space that the algorithm can handle. The Holland's encoding method is binary, but for many applications of hybrid particle swarm optimization algorithms, it is difficult for this simple encoding method to directly describe the nature of the problem.
As an alternative embodiment, setting initial values for particle velocity and particle position includes: and setting random initial values for the particle speed and the particle position in a real number encoding mode. In the embodiment, a real number coding mode is adopted to code the individual, each gene value of the individual is represented by a real number in a certain range, and the length of the individual code is equal to the number of decision variables. And real number coding is adopted, so that the decoding problem after coding does not exist.
In general, in the absence of a priori knowledge of the problem to be solved, the initial population of the evolutionary algorithm is generally generated in a random manner, so the present embodiment generates the population by using a method of randomly generating the initial population.
And (4) randomly initializing parameters in X and V through real numbers to obtain initial population individuals.
The initial population generating and encoding method adopted by the hybrid particle swarm optimization algorithm can ensure that the initial population covers the whole definition domain. The advantages of the real number encoding method are as follows: (1) large numbers are suitable for table demonstration in genetic algorithms; (2) facilitating genetic searches in larger spaces; (3) the accuracy requirement of the genetic algorithm is improved; (4) the computational complexity of the genetic algorithm is improved, and the computational efficiency is improved; (5) the mixing action of the algorithm and the classical optimization method is facilitated; (6) and the genetic operator of the special problem is convenient to design.
102. And performing iterative updating operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until a preset iteration number is reached.
In 102, an existing particle swarm algorithm is adopted, individuals in the population are updated through a preset fitness function, and the fitness function reflects characteristics of clients. Compared with the total iteration times, the preset times K 'do not need genetic operation when the particle swarm optimization algorithm is in a global random search stage (the iteration times are smaller than the preset times), and the genetic algorithm operation is started after the iteration times reach the preset times K', so that the particle swarm optimization can be utilized to the maximum extent to approach to the global optimal solution, and the convergence speed of the algorithm is accelerated.
K ═ a × K, in this embodiment, a ═ 0.2 is taken, and K is the maximum iteration number of the evolution process.
103. After the iteration times are preset, after each update, respectively carrying out selection operation on the particle swarm according to a genetic algorithm, carrying out cross operation and mutation operation for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition.
In 103, after the iterative update times reach the preset iterative times K', and after the existing population inertia weight, particle velocity, and particle position are updated each time, the individuals in the mixed particle swarm are optimized by using a genetic algorithm, wherein genetic operators used in the algorithm include selection, intersection, and mutation operations. And optimizing the individuals in the mixed particle swarm by adopting a genetic algorithm to obtain an optimized population, wherein the optimized population is used for updating according to the particle swarm optimization algorithm in the next iteration.
The preset iteration times are set to avoid fusing genetic algorithms too early, and the search advantages of the particle swarm algorithm are fully utilized in the early stage. It should be noted that the preset number of iterations may be set as needed, and may be set to 1, that is, the genetic algorithm is fused at the beginning.
104. And obtaining a clustering center according to the particle swarm which reaches the total iteration times or meets the convergence condition, and classifying the clients based on a clustering algorithm.
And after the iteration total times K are finished or the convergence condition of the particle swarm optimization is met, terminating the optimization of the particle swarm optimization. The convergence condition of the particle swarm algorithm can be that the individual optimal result is consistent with the group optimal result. The corresponding clustering center can be determined according to the optimizing result obtained in the whole iteration process. And according to the clustering center, a clustering method based on division, such as K-mean clustering, is adopted, so that the classification of the customers is carried out.
The client classification method based on the improved particle swarm optimization algorithm provided by the embodiment of the invention can overcome the defects that the standard particle swarm optimization algorithm is easy to fall into the local optimal solution, the later convergence speed is low, the search precision is not high, and the like. By setting the genetic algorithm, the optimization calculation of the particle swarm is improved, the method classifies clients based on the clustering algorithm, and after the iteration of preset times, the organic fusion of the genetic algorithm is carried out, so that the situation that the clients fall into a local optimal solution can be avoided, the later convergence speed is improved, the search precision is improved, and an accurate clustering center is obtained, thereby avoiding the defect that the traditional clustering algorithm depending on the clustering center easily falls into a local extreme value, and improving the classification accuracy.
Based on the content of the foregoing embodiment, as an optional embodiment, the selecting operation on the particle swarm includes: calculating the selection probability of each individual inherited to the next generation according to a preset fitness function, and calculating the cumulative selection probability of the individual to the current iteration times; and determining each individual as a selection result of the next iteration individual according to the accumulated selection probability of each individual.
The selection operation (or called replication operation) is to realize an optimal preservation strategy, ensure that good particle groups are not eliminated, ensure that good excellent modes are not damaged, and embody the mechanism of survival and elimination of unsuitable people. The probability that the individual with high fitness is inherited to the next generation group is high, and the probability that the individual with low fitness is inherited to the next generation group is low, so that the main purpose is to avoid gene deletion and improve the overall convergence and the calculation efficiency. The operators performing the selection operation are called selection operators (copy operators), and the commonly used selection operators are: roulette operators, elite operators, sequential operators, clone operators, and the like, and the roulette operator is taken as an example for the present embodiment.
In an improved particle swarm optimization algorithm fusing genetic ideas, firstly, new particle swarms generated by each iteration are compared according to a preset fitness function, a roulette operator (the probability of each individual being selected is in direct proportion to the fitness) is adopted for selection, M particles with high fitness are selected (M is b multiplied by N, in the embodiment, b is 0.6, and N is the total number of original particle swarms), and the method specifically comprises the following operations:
(1) calculating the fitness f of each individual in the populationiAnd the sum of the fitness values f of all particles in the current populations，i＝1,2,…,N；
(2) Calculating the selection probability (relative fitness) p of each individual inherited into the next generation groupi，
The selection probability is determined according to the proportion of the fitness value of the individual to the sum of the fitness values of all the individuals;
(3) calculating the cumulative probability of each individual to the current iteration number
(4) Based on roulette algorithm, in the interval [0,1]]Generating a uniformly distributed pseudo-random number r e [0,1 ∈]If, if
Selecting an individual i to enter the next generation;
(5) repeating the step (4) for M times, namely generating M individuals to form a next generation population.
According to the customer classification method based on the improved particle swarm optimization algorithm, the particles of the next iteration are selected according to the cumulative selection probability of the individuals, the cumulative selection probability is in direct proportion to the size of the adaptive value, the individuals with high fitness are selected with higher probability, and the individuals with low fitness are selected with lower probability. And obtaining the cumulative selection probability according to the individual fitness of the particles, determining the survival of the individual, continuously improving the average fitness of the particle swarm, and enhancing the local searching capability of the particle swarm. In order to improve the convergence rate, a roulette selection operator can be adopted to perform 'winning or rejecting', appropriate particles are selected to be crossed, and the situation that the particles fall into a local optimal solution is prevented.
Based on the content of the foregoing embodiment, as an alternative embodiment, the interleaving operation is performed, and includes: according to a preset cross probability function, the individuals obtained by the selection operation are selected again; performing cross operation on the individuals obtained by re-selection and the individual optimal result and the group optimal result respectively; wherein the cross probability function is a decreasing function with respect to the number of iterations.
Considering that the number of particles with high fitness is large, although the convergence speed of the algorithm is high, the algorithm is caused to fall into a local extreme value, and the phenomenon of premature convergence occurs. If the number of particles with low fitness is large, the convergence rate of the algorithm is slow although diversity of the population can be maintained.
Because the late stage of the recent generation is easy to trap in the local convergence, in order to prevent the local optimal solution from trapping and improve the convergence speed, the embodiment designs that the cross operation is started according to the cross probability after a certain number of iterations K', so that the local search capability is improved, and the convergence speed is accelerated. And (N-M)/2 particles are selected from the selected particle swarm according to the cross probability to respectively carry out cross operation with the individual optimum and the group optimum, and the particles with good fitness are selected to form a new group again, so that the total number N of the original particle swarm is maintained unchanged, and the rate of premature convergence is reduced by the cross operation. The crossover enables the exchange of genetic material between individuals to produce better individuals, and this operation enables the number of individuals in the population with better fitness values to be increased rapidly, thereby improving the efficient practicality of the algorithm.
The crossover operator continuously updates the population and has crossover probability PcThe updating rate of the population individuals is determined by the size of the gene, and the excellent genetic pattern can be damaged if the value is too large; the over-small value of the value can cause slow searching speed of the algorithm, and the probability P of the population being difficult to evolve and cross-occurcGenerally between 0.25 and 1.00. Therefore, in the early stage of population evolution, in order to expand the overall search range and accelerate the population updating speed, P should be increasedcA value of (d); in the later evolution stage, the overall solution set of the population tends to be stable, in order to make good genesThe structure is preserved, and P should be reduced appropriatelycThe value of (c). In summary, the cross probability function of the present invention is a decreasing function with respect to the number of iterations.
Preferably, the cross probability function is an increasing function with respect to the fitness value, and is specifically determined according to the fitness value, the fitness maximum value of the current iteration, the fitness minimum value, and the number of iterations.
The method for determining the cross probability function comprises the following steps: determined according to the Logistic function. The Logistic function is as follows:
accordingly, the determined cross probability function comprises:
wherein p isciThe probability of generating cross operators for the individual i has higher cross probability at the initial stage of evolution and lower cross probability at the later stage of evolution; k is the current iteration number; k' is the iteration number when the intersection occurs; k is the maximum iteration number; f. ofiThe fitness function value of the individual i.
(fi-fmin)/(fmax-fmin) The value range is [0,1]]I.e. corresponding to a [0,1]]The random variable of (2) and the particles with high individual self-fitness have higher cross probability and the particles with low individual self-fitness have lower cross probability, thus effectively preventing the particles from falling into the local optimal solution. When (f)i-fmin)/(fmax-fmin) When the value changes, fig. 2 is a graph of adjusting the genetic crossover probability of each particle provided by the embodiment of the present invention, the curve of the crossover probability along with the evolution times is shown in fig. 2, when (f)i-fmin)/(fmax-fmin) When the value is close to 1, fig. 3 is a graph for adjusting the genetic crossover probability of a single particle provided by the embodiment of the present invention, and the curve of the crossover probability along with the evolution times is shown in fig. 3, which shows that the overall probability is close to 1The trend decreases with the number of iterations, consistent with the analysis described above.
And continuously applying a crossover operator in the iterative process to enable the genes of the excellent individuals to frequently appear in the population, and finally enabling the whole population to converge to an optimal solution. Crossover is the recombination of genes on two different chromosomes, the position of the crossover is randomly determined, and the operator performing the crossover operation is called crossover operator. The crossover operator is an important operator in genetic operation and a main method for generating new individuals, and parent individual information can be directly exchanged through the crossover operation, so that more excellent individuals in a solution space can be found, and the local optimal solution can be effectively prevented from being trapped. The commonly used crossover operators include a single-point crossover operator, a double-point crossover operator, a uniform crossover operator, a fusion crossover operator, an arithmetic crossover operator, and the like.
The generation of a new particle is influenced by the self, the historical optimal particle and the group optimal particle, so that the self particle is respectively crossed with the historical optimal particle and the group optimal particle, then the fitness of the particle generated by the individual extreme value is compared with the fitness generated by the group extreme value, and the particle with higher fitness is selected as the new particle. Therefore, in order to improve the crossing efficiency, the embodiment selects (N-M)/2 particles from the selected M population particles to perform double-point crossing with the individual history optimization and the population optimization respectively to generate a new population (the crossing operation is performed under the crossing probability) in a way of crossing from the interval [ l, D ]](D is the number of individual genes, namely the dimension of the particle) and randomly setting two cross points; then, the partial genes between the two intersections are exchanged, and two new individuals are finally generated. Let two parents be X respectively1＝a a a a a a a，X2B bb b b b, the filial generation individuals after crossing are respectively X1＝a a a b b a a，X2B b a a b, the calculation formula is as follows:
wherein p isciProbability of occurrence of crossover operators for individual i; pbestiIs a particle iA volume optimal solution; and the gbest is the optimal solution of the population under the current iteration. The hybrid particle swarm optimization algorithm of cross operation is introduced, when the fitness of the optimal particles of the particle swarm is continuously improved along with the optimization iteration, namely the particle swarm continuously moves towards the global optimal direction, the searching capability of the particle swarm to the position near the current optimal particle position is enhanced, the local optimal state can be effectively prevented, and the calculation efficiency of the algorithm is improved.
Based on the content of the foregoing embodiment, as an alternative embodiment, after the performing the crossover operation and before the performing the mutation operation, the method further includes: and calculating the group fitness variance of the current iteration population, and if the group fitness variance is smaller than a preset threshold, performing mutation operation.
Once a single particle in the particle swarm is constrained by a certain local mechanism, the local mechanism is difficult to draw out, and although the particle swarm is still likely to have premature convergence after the selection and the cross operation. Aiming at the defects of population diversity in the later evolution stage of the particle swarm optimization algorithm, the convergence precision of the algorithm is low and the algorithm is easy to fall into a local extreme value, when the algorithm falls into a local optimal solution (namely, the premature convergence phenomenon occurs), parts such as 1/4 population particles are selected in time according to the variation probability to perform variation operation, the variation can recover genetic materials lost or not developed by individuals, new search is started in a multi-dimensional solution space, and the diversity of the particles is enhanced to enable the particles to jump out of the constraint of the local extreme value.
However, since frequent mutation causes lack of stability in the algorithm, it is necessary to select a mutation timing according to the evolution state of the particle group so that the algorithm has the ability to jump out of the locally optimal solution. The premature convergence judgment is a basis of variation operation, and in order to make a population jump out of local optimality quickly, the embodiment adopts a population fitness variance (tracking population state) of a particle swarm to judge whether premature convergence occurs, where the population fitness variance includes:
wherein N is the total number of population particles; f. ofiRepresenting the ith particleA fitness value; f. ofavgRepresenting the average fitness value of the current population; f denotes a normalization factor, f { | f { [ max { ] { [ f ]i-favg|}。
σ2Reflects the convergence degree, sigma, of the current particle swarm2Smaller indicates that the particles gradually tend to a convergent state; sigma2When the particle swarm is close to 0, namely the fitness of the particles is almost the same, the particle swarm is trapped in premature convergence or is in global convergence; otherwise, σ2The larger the size, the more the population is in the random search phase. Thus, the population fitness variance σ is exploited2To determine the operation time of mutation operator when sigma is2<When the preset threshold value delta (different problems, different values of the preset threshold value delta, and set according to specific requirements) is not satisfied with the termination condition, it can be determined that the current particle is trapped in premature convergence, and a mutation operation is performed according to the mutation probability.
Based on the content of the foregoing embodiments, as an alternative embodiment, a mutation operation is performed, including: determining individuals subjected to mutation operation in the individuals subjected to the cross operation according to a preset mutation probability function, and performing mutation operation; wherein the variation probability function is an increasing function with respect to the number of iterations.
Probability of variation PmThe variation condition of the population is influenced, a large number of excellent individuals are damaged due to overlarge values, the optimal solution cannot be obtained possibly, the algorithm is similar to random search, and the characteristic of genetic evolution is lost; if the value is too small, the individual cannot be mutated to a more optimal solution, the convergence rate of the algorithm is slowed down, and the algorithm is difficult to jump out of the local optimum and the probability P of mutation occurrencemGenerally, the amount of the surfactant is 0.001 to 0.1. Therefore, at the initial stage of evolution, the possibility of individual variation is low; at the end of evolution, the probability of variation of population individuals is properly increased due to similar gene structures in population individual regions, so that generation of new individuals is encouraged, local optimality is facilitated to jump out, and population diversity is kept. In summary, the variation probability function in this embodiment is an increasing function with respect to the number of iterations.
Preferably, the variation probability function is an increasing function with respect to the fitness value, and is specifically determined according to the fitness value, the fitness maximum value of the current iteration, the fitness minimum value, and the iteration number.
The method for determining the variation probability function comprises the step of fitting according to a Logistic function to obtain the variation probability function. Wherein, the cross probability function obtained by fitting according to the Logistic function comprises:
wherein p ismiThe probability of generating mutation operators for the individual i is lower in the early evolution stage and higher in the later evolution stage; k is the current iteration number; k' is the iteration number when the intersection occurs; k is the maximum number of iterations.
(fi-fmin)/(fmax-fmin) The value range is [0,1]]I.e. corresponding to a [0,1]]The random variable of (2) is larger in variation probability of the particles with high individual self-fitness and smaller in variation probability of the particles with low individual self-fitness, so that the local optimal solution can be skipped. When (f)i-fmin)/(fmax-fmin) When the value changes, FIG. 4 is a graph of the adjustment of the genetic variation probability of each particle according to the embodiment of the present invention, as shown in FIG. 4, when (f)i-fmin)/(fmax-fmin) When the value is close to 1, fig. 5 is a graph of the adjustment of the genetic variation probability of a single particle provided by the embodiment of the present invention, and the change curve of the cross probability along with the number of evolutions is shown in fig. 5.
Mutation is the alteration of the structure of a chromosome by changing its original structure (i.e., changing its value). The purpose of the mutation is to enhance the diversity of the population, making it jump out of the current local subspace immediately once it falls into local optimality. Mutation operators generally serve as auxiliary operators, and replace the gene values of certain gene positions in the chromosome with other alleles to form a new particle individual. Although mutation is an aid in genetic algorithms, it is an indispensable computational step, and generally begins to mutate when the population begins to fall into a locally optimal solution. The commonly used mutation operators include bitwise mutation, uniform mutation, Gaussian mutation, directed mutation, etc.
The motion state with randomness obtained by a deterministic equation is generally called chaotic motion, the chaotic motion has the characteristics of ergodicity, randomness, regularity and the like, and all the states can be traversed in a certain range according to the law of the chaotic motion without repetition. The chaotic system dynamic models are many, Logistic mapping is a common one-dimensional nonlinear iterative chaotic model, and compared with other power systems generating chaotic variables, the method is simpler and has small calculation amount. Therefore, the embodiment introduces the variation improvement based on the logistic mapping of the chaos idea, that is, the variation operation in the embodiment of the present invention includes performing variation according to the logistic mapping. Repeated searching can be avoided to keep the diversity of the population, chaotic variation is carried out according to variation probability when the premature convergence phenomenon occurs every time, and the positions of the particles are updated, so that the particle swarm is out of the constraint of the local optimal solution, and the searching precision is improved. The mutation algorithm is calculated as follows:
position vector x of current particle ii＝(xi1,xi2,…,xid) Mapping to Domain [0,1] of Logistic equations by the first formula]Get yi＝(yi1,yi2,…,yid) Calculating to obtain a chaotic sequence yi’＝(yi1’,yi2’,…,yid') (see second formula below this paragraph) and then mapped back to the original solution space of the problem, thereby producing a chaotic variable feasible solution sequence xi’＝(xi1’,xi2’,…,xid') (see the third formula below this paragraph), a mutation operation is performed. The method for establishing the mapping comprises the following steps: if the interval [ m, n]Mapping to the interval [ p, q ]]In this case, a one-to-one mapping is only required, and if g (x) ax + b, g (m) p and g (n) q are provided, the values of a and b can be obtained, that is, the real number of the corresponding interval is mapped to [0,1]The above. If the interval [ m, n]Mapping to an interval [0,1]]Upper (see first formula below this paragraph); if the interval is [0,1]]Mapping to an interval [ m, n ]]Upper (see the third formula below this paragraph).
y′i＝μyi(1-yi)yi∈[0,1]
x′i＝(n-m)y′i+m
Wherein, mu is a control parameter, and when mu is 4, the system is in a complete chaotic state. The Logistic mapping particle values, particle distribution states, and probability density function plots of Logistic over the mapping interval [0,1] are shown in fig. 6-10 (e.g., 500 particles total).
Fig. 6 is a schematic diagram of a Logistic mapping value provided in an embodiment of the present invention, fig. 7 is a schematic diagram of distribution of Logistic mapping particles provided in an embodiment of the present invention, and fig. 8 is a schematic diagram of a Logistic probability density function provided in an embodiment of the present invention. As shown in fig. 6-8, it can be seen from fig. 7 and 8 that the logistic mapping has the characteristics that the edge is suddenly changed, the middle is gentle, and the distribution probability at the two ends of the interval is greater than that at the middle part of the interval, and the uniform distribution characteristic of the mapping is poor, so that the position distribution of the particles is uneven, and the effect of enhancing the diversity of the particles is not good. Therefore, the embodiment of the present invention changes the chaotic sequence (the second formula above) into the following formula:
yi＝r×cos(μyi(1-yi)-0.4)yi∈[0,1]
where r is a number randomly generated in the interval [0,1], and μ ═ 4.
Fig. 9 is a schematic diagram of improved Logistic mapping values provided by an embodiment of the present invention, fig. 10 is a schematic diagram of particle distribution of improved Logistic mapping provided by an embodiment of the present invention, the improved Logistic mapping values and particle distribution states of the present invention are shown in fig. 9-10, and fig. 10 shows that the search blind area of the given model is significantly reduced, and it is easier for individual particles to vary to each corner, thereby enhancing population diversity and multi-scale. After the mutation operation, the new solution may be worse than the original solution, and this embodiment allows the objective function to be worse within a limited range.
Based on the content of the foregoing embodiment, as an alternative embodiment, after performing the mutation operation, the method further includes: adjusting the inertia weight of the particle swarm algorithm according to a preset inertia weight function; wherein the inertial weight function is a decreasing function with respect to the number of iterations.
Inertial weights have a large impact on the search capability of the particle, which is a balance of global and local search performance. The commonly used fixed or linear decreasing inertial weight inevitably generates a 'cross oscillation' phenomenon near the global optimal solution, so that the global search capability is not strong, and the convergence speed is influenced. The larger inertia weight can better perform global search, and the smaller inertia weight has better local search performance. For each particle, when the particle is far away from the global optimal solution, the flight speed needs to be increased to perform global search (i.e. increase the inertia weight); when the distance to the global optimal solution is closer, the flight speed is reduced (namely, the inertia weight is reduced), and local refinement search is carried out, namely, the inertia weight function is a function of nonlinear decrement of the iteration number. In order to make the algorithm more dynamically adaptive, the present embodiment designs a piecewise inertial weight that decreases nonlinearly with the number of iterations through rough fitting, where the preset inertial weight function includes:
wherein, the inertia weight w generally takes a value of [0.8,1.2], and k is the current iteration number. The strategy can maintain larger inertia weight and reduce slowly in the initial stage of the algorithm, so that the algorithm can maintain stronger global searching capability even if a better position is not searched in the initial stage, and the probability of searching the better position by the algorithm is improved; the inertia weight is slowly reduced in the later stage of the algorithm, the local searching capability of the algorithm is rapidly strengthened, and the convergence performance of the algorithm can be obviously improved compared with the fixed inertia weight.
Based on the content of the above embodiment, as an optional embodiment, if the position of the population individual exceeds the set maximum position value or minimum position value, the position of the population individual is finely adjusted through a preset random function; and if the speed of the population individuals exceeds the set maximum speed value or minimum speed value, fine adjustment is carried out on the speed of the population individuals through a preset random function.
The hybrid particle swarm optimization algorithm carries out the evolution of the particle swarm optimization algorithm in the early stage, the genetic thought starts to be fused after certain iteration times (such as iteration to 200 times), and through operations of intersection, variation and the like, certain bits of the population may exceed the range of the problem solution, which requires corresponding processing. When the position and the speed of the particle have escape phenomena, the particle speed and position processing method adopted by the invention comprises the following steps:
wherein, XmaxAnd XminRespectively representing set maximum and minimum position values; vmaxAnd VminRespectively, the set maximum and minimum speed values are represented, and rand is a random function of (0, 1). The escape strategy is equivalent to introducing a new population, namely, small random disturbance is carried out on particles beyond the limit, so that the method is beneficial to an individual to find a better solution on the basis of keeping the current good performance.
Fig. 11 is a structural diagram of a client classification device based on an improved particle swarm optimization algorithm according to an embodiment of the present invention, and as shown in fig. 11, the client classification device based on the improved particle swarm optimization algorithm includes: an initialization module 111, a first search module 112, a second search module 113, and a classification module 114. The initialization module 111 is configured to initialize the particle speed and the particle position according to the classification number and the feature dimension, and set an initial value to establish an initial population of the particle swarm; the first search module 112 is configured to perform iterative update operations on the inertia weight, the particle velocity, and the particle position of the population according to a preset fitness function including the client feature data until a preset iteration number is reached; the second search module 113 is configured to, after the iteration number is preset, perform selection operation on the particle swarm according to a genetic algorithm after each update, perform a crossover operation and a mutation operation, and perform next iteration update until the iteration update reaches the total iteration number or meets a convergence condition; the classification module 114 obtains a clustering center according to the total number of iterations or the particle group satisfying the convergence condition, and classifies the clients based on a clustering algorithm.
The device embodiment provided in the embodiments of the present invention is for implementing the above method embodiments, and for details of the process and the details, reference is made to the above method embodiments, which are not described herein again.
The client classification device based on the improved particle swarm optimization algorithm provided by the embodiment of the invention can overcome the defects that the standard particle swarm optimization algorithm is easy to fall into the local optimal solution, the later convergence speed is low, the search precision is not high, and the like. By setting the genetic algorithm, the optimization calculation of the particle swarm is improved, the method classifies clients based on the clustering algorithm, and after the iteration of preset times, the organic fusion of the genetic algorithm is carried out, so that the situation that the clients fall into a local optimal solution can be avoided, the later convergence speed is improved, the search precision is improved, and an accurate clustering center is obtained, thereby avoiding the defect that the traditional clustering algorithm depending on the clustering center easily falls into a local extreme value, and improving the classification accuracy.
Fig. 12 is a schematic entity structure diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 12, the electronic device may include: a processor (processor)121, a communication Interface (Communications Interface)122, a memory (memory)123 and a bus 124, wherein the processor 121, the communication Interface 122 and the memory 123 complete communication with each other through the bus 124. The communication interface 122 may be used for information transfer of the electronic device. Processor 121 may call logic instructions in memory 123 to perform a method comprising: initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm; performing iterative updating operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until a preset iteration number is reached; after the iteration times are preset, respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update, carrying out cross operation and mutation operation for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition; and obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients based on a clustering algorithm.
In addition, the logic instructions in the memory 123 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the above-described method embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented to perform the transmission method provided in the foregoing embodiments when executed by a processor, and for example, the method includes: initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm; performing iterative updating operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until a preset iteration number is reached; after the iteration times are preset, respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update, carrying out cross operation and mutation operation for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition; and obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients based on a clustering algorithm.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods of the various embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A customer classification method based on an improved particle swarm optimization algorithm is characterized by comprising the following steps:
initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm;
performing iterative updating operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until a preset iteration number is reached;
after the iteration times are preset, respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update, carrying out cross operation and mutation operation for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition;
and obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients based on a clustering algorithm.
2. The improved particle swarm optimization algorithm-based client classification method according to claim 1, wherein the selecting operation on the particle swarm comprises:
calculating the selection probability of each individual inherited to the next generation according to a preset fitness function, and calculating the cumulative selection probability of the individual to the current iteration times;
and determining each individual as a selection result of the next iteration individual according to the accumulated selection probability of each individual.
3. The improved particle swarm optimization algorithm-based customer classification method according to claim 1, wherein the performing of the crossover operation comprises:
according to a preset cross probability function, the individuals obtained by the selection operation are selected again;
performing cross operation on the individuals obtained by re-selection and the individual optimal result and the group optimal result respectively;
wherein the cross probability function is a decreasing function with respect to the number of iterations.
4. The customer classification method based on the improved particle swarm optimization algorithm according to claim 1, wherein after the performing the crossover operation and before the performing the mutation operation, the method further comprises:
and calculating the group fitness variance of the current iterative population, and if the group fitness variance is smaller than a preset threshold, performing variation operation.
5. The customer classification method based on the improved particle swarm optimization algorithm according to any one of claims 1 to 4, wherein the performing mutation operation comprises:
determining individuals subjected to mutation operation in the individuals subjected to the cross operation according to a preset mutation probability function, and performing mutation operation;
wherein the variation probability function is an increasing function with respect to the number of iterations.
6. The customer classification method based on the improved particle swarm optimization algorithm according to claim 5, further comprising, after the performing the mutation operation:
adjusting the inertia weight of the particle swarm algorithm according to a preset inertia weight function;
wherein the inertial weight function is a non-linearly decreasing function with respect to the number of iterations.
7. The customer classification method based on the improved particle swarm optimization algorithm according to claim 5, further comprising, after the performing the mutation operation:
if the position of the population individual exceeds the set maximum position value or minimum position value, fine adjustment is carried out on the position of the population individual through a preset random function;
and if the speed of the population individuals exceeds the set maximum speed value or minimum speed value, fine adjustment is carried out on the speed of the population individuals through a preset random function.
8. A customer classification device based on an improved particle swarm optimization algorithm is characterized by comprising:
the initialization module is used for initializing the particle speed and the particle position according to the classification number and the characteristic dimension, and setting initial values to establish an initial population of the particle swarm;
the first search module is used for carrying out iterative update operation on the inertia weight, the particle speed and the particle position of the population according to a preset fitness function comprising the client characteristic data until the preset iteration times are reached;
the second search module is used for respectively carrying out selection operation on the particle swarm according to a genetic algorithm after each update after the iteration times are preset, carrying out cross operation and mutation operation, and is used for next iteration update until the iteration update reaches the total iteration times or meets a convergence condition;
and the classification module is used for obtaining a clustering center according to the total iteration times or the particle swarm meeting the convergence condition, and classifying the clients based on a clustering algorithm.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps of the improved particle swarm optimization algorithm based customer classification method according to any one of claims 1 to 7.
10. A non-transitory computer readable storage medium, having stored thereon a computer program, wherein the computer program, when being executed by a processor, implements the steps of the improved particle swarm optimization algorithm based client classification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911089171.5A CN110930182A (en) | 2019-11-08 | 2019-11-08 | Improved particle swarm optimization algorithm-based client classification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911089171.5A CN110930182A (en) | 2019-11-08 | 2019-11-08 | Improved particle swarm optimization algorithm-based client classification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110930182A true CN110930182A (en) | 2020-03-27 |
Family
ID=69852628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911089171.5A Pending CN110930182A (en) | 2019-11-08 | 2019-11-08 | Improved particle swarm optimization algorithm-based client classification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110930182A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111426323A (en) * | 2020-04-16 | 2020-07-17 | 南方电网科学研究院有限责任公司 | Routing planning method and device for inspection robot |
| CN111444569A (en) * | 2020-04-01 | 2020-07-24 | 南通大学 | Beam structure measuring point optimization method based on improved particle swarm optimization |
| CN111460735A (en) * | 2020-04-08 | 2020-07-28 | 深圳市瑞立视多媒体科技有限公司 | Genetic-inheritance-based camera layout function optimization method and related equipment |
| CN111489021A (en) * | 2020-03-31 | 2020-08-04 | 黑龙江大学 | Beet yield prediction method based on particle swarm optimization BP neural network |
| CN111506693A (en) * | 2020-04-21 | 2020-08-07 | 北京邮电大学 | Cooperative positioning method and device, electronic equipment and storage medium |
| CN111612528A (en) * | 2020-04-30 | 2020-09-01 | 中国移动通信集团江苏有限公司 | Method, device and equipment for determining user classification model and storage medium |
| CN111652267A (en) * | 2020-04-21 | 2020-09-11 | 清华大学 | Method and device for generating countermeasure sample, electronic equipment and storage medium |
| CN111883208A (en) * | 2020-06-24 | 2020-11-03 | 浪潮电子信息产业股份有限公司 | Gene sequence optimization method, device, equipment and medium |
| CN112070275A (en) * | 2020-08-14 | 2020-12-11 | 青岛海信网络科技股份有限公司 | Method, device and equipment for cooperatively limiting current of rail transit road network |
| CN112200224A (en) * | 2020-09-23 | 2021-01-08 | 温州大学 | Medical image feature processing method and device |
| CN112926245A (en) * | 2021-02-23 | 2021-06-08 | 广西电网有限责任公司电力科学研究院 | Arrester grading ring design method based on hybrid genetic-Taguchi algorithm |
| CN113205173A (en) * | 2021-05-28 | 2021-08-03 | 中国工商银行股份有限公司 | Electronic ticket selection method and device |
| CN113240065A (en) * | 2021-04-30 | 2021-08-10 | 西安电子科技大学 | Passive radar station distribution method based on improved particle swarm optimization algorithm |
| CN113554241A (en) * | 2021-09-02 | 2021-10-26 | 国网山东省电力公司泰安供电公司 | User layering method and prediction method based on user electricity complaint behaviors |
| CN114172963A (en) * | 2021-10-15 | 2022-03-11 | 合肥工业大学 | Multi-granularity task and service matching method and system based on resource utilization rate |
| CN114547954A (en) * | 2020-11-24 | 2022-05-27 | 中国移动通信集团浙江有限公司 | Logistics distribution center site selection method and device and computer equipment |
| CN114757724A (en) * | 2022-06-14 | 2022-07-15 | 湖南三湘银行股份有限公司 | Precise information pushing system and method based on genetic algorithm |
| CN113159459B (en) * | 2021-05-21 | 2023-04-18 | 南京林业大学 | Multi-forest-area air route scheduling planning method based on fusion algorithm |
| CN116205536A (en) * | 2023-03-06 | 2023-06-02 | 阿里巴巴（中国）有限公司 | Object evaluation method, computing device, and readable storage medium |
| CN116757333A (en) * | 2023-08-12 | 2023-09-15 | 中国人民解放军96901部队 | Classification dustbin optimal configuration method based on resident satisfaction |
-
2019
- 2019-11-08 CN CN201911089171.5A patent/CN110930182A/en active Pending
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111489021A (en) * | 2020-03-31 | 2020-08-04 | 黑龙江大学 | Beet yield prediction method based on particle swarm optimization BP neural network |
| CN111444569A (en) * | 2020-04-01 | 2020-07-24 | 南通大学 | Beam structure measuring point optimization method based on improved particle swarm optimization |
| CN111460735A (en) * | 2020-04-08 | 2020-07-28 | 深圳市瑞立视多媒体科技有限公司 | Genetic-inheritance-based camera layout function optimization method and related equipment |
| CN111460735B (en) * | 2020-04-08 | 2023-09-01 | 深圳市瑞立视多媒体科技有限公司 | Camera layout function optimization method based on genetic inheritance and related equipment |
| CN111426323B (en) * | 2020-04-16 | 2022-04-12 | 南方电网科学研究院有限责任公司 | Routing planning method and device for inspection robot |
| CN111426323A (en) * | 2020-04-16 | 2020-07-17 | 南方电网科学研究院有限责任公司 | Routing planning method and device for inspection robot |
| CN111506693A (en) * | 2020-04-21 | 2020-08-07 | 北京邮电大学 | Cooperative positioning method and device, electronic equipment and storage medium |
| CN111652267A (en) * | 2020-04-21 | 2020-09-11 | 清华大学 | Method and device for generating countermeasure sample, electronic equipment and storage medium |
| CN111506693B (en) * | 2020-04-21 | 2023-11-28 | 北京邮电大学 | Co-location method, device, electronic equipment and storage medium |
| CN111652267B (en) * | 2020-04-21 | 2023-01-31 | 清华大学 | Method and device for generating countermeasure sample, electronic equipment and storage medium |
| CN111612528A (en) * | 2020-04-30 | 2020-09-01 | 中国移动通信集团江苏有限公司 | Method, device and equipment for determining user classification model and storage medium |
| CN111883208A (en) * | 2020-06-24 | 2020-11-03 | 浪潮电子信息产业股份有限公司 | Gene sequence optimization method, device, equipment and medium |
| CN111883208B (en) * | 2020-06-24 | 2022-07-05 | 浪潮电子信息产业股份有限公司 | Gene sequence optimization method, device, equipment and medium |
| CN112070275A (en) * | 2020-08-14 | 2020-12-11 | 青岛海信网络科技股份有限公司 | Method, device and equipment for cooperatively limiting current of rail transit road network |
| CN112200224B (en) * | 2020-09-23 | 2023-12-01 | 温州大学 | Medical image feature processing method and device |
| CN112200224A (en) * | 2020-09-23 | 2021-01-08 | 温州大学 | Medical image feature processing method and device |
| CN114547954A (en) * | 2020-11-24 | 2022-05-27 | 中国移动通信集团浙江有限公司 | Logistics distribution center site selection method and device and computer equipment |
| CN112926245B (en) * | 2021-02-23 | 2022-12-16 | 广西电网有限责任公司电力科学研究院 | Arrester grading ring design method based on hybrid genetic-Taguchi algorithm |
| CN112926245A (en) * | 2021-02-23 | 2021-06-08 | 广西电网有限责任公司电力科学研究院 | Arrester grading ring design method based on hybrid genetic-Taguchi algorithm |
| CN113240065A (en) * | 2021-04-30 | 2021-08-10 | 西安电子科技大学 | Passive radar station distribution method based on improved particle swarm optimization algorithm |
| CN113159459B (en) * | 2021-05-21 | 2023-04-18 | 南京林业大学 | Multi-forest-area air route scheduling planning method based on fusion algorithm |
| CN113205173A (en) * | 2021-05-28 | 2021-08-03 | 中国工商银行股份有限公司 | Electronic ticket selection method and device |
| CN113554241B (en) * | 2021-09-02 | 2024-04-26 | 国网山东省电力公司泰安供电公司 | User layering method and prediction method based on user electricity complaint behaviors |
| CN113554241A (en) * | 2021-09-02 | 2021-10-26 | 国网山东省电力公司泰安供电公司 | User layering method and prediction method based on user electricity complaint behaviors |
| CN114172963A (en) * | 2021-10-15 | 2022-03-11 | 合肥工业大学 | Multi-granularity task and service matching method and system based on resource utilization rate |
| CN114757724A (en) * | 2022-06-14 | 2022-07-15 | 湖南三湘银行股份有限公司 | Precise information pushing system and method based on genetic algorithm |
| CN114757724B (en) * | 2022-06-14 | 2022-09-20 | 湖南三湘银行股份有限公司 | Precise information pushing system and method based on genetic algorithm |
| CN116205536A (en) * | 2023-03-06 | 2023-06-02 | 阿里巴巴（中国）有限公司 | Object evaluation method, computing device, and readable storage medium |
| CN116205536B (en) * | 2023-03-06 | 2023-10-13 | 阿里巴巴（中国）有限公司 | Object evaluation method, computing device, and readable storage medium |
| CN116757333A (en) * | 2023-08-12 | 2023-09-15 | 中国人民解放军96901部队 | Classification dustbin optimal configuration method based on resident satisfaction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110930182A (en) | Improved particle swarm optimization algorithm-based client classification method and device | |
| Strumberger et al. | Enhanced firefly algorithm for constrained numerical optimization | |
| Valdez et al. | Modular neural networks architecture optimization with a new nature inspired method using a fuzzy combination of particle swarm optimization and genetic algorithms | |
| Gharghory et al. | Optimal tuning of PID controller using adaptive hybrid particle swarm optimization algorithm | |
| CN107330902B (en) | Chaotic genetic BP neural network image segmentation method based on Arnold transformation | |
| CN109298930B (en) | Cloud workflow scheduling method and device based on multi-objective optimization | |
| Pilat et al. | Incorporating user preferences in MOEA/D through the coevolution of weights | |
| CN114742231A (en) | Multi-objective reinforcement learning method and device based on pareto optimization | |
| Arasomwan et al. | On adaptive chaotic inertia weights in particle swarm optimization | |
| Moriguchi et al. | Sustaining behavioral diversity in neat | |
| CN113141272B (en) | Network security situation analysis method based on iteration optimization RBF neural network | |
| CN106709572B (en) | A kind of data processing method and equipment | |
| CN111930435A (en) | Task unloading decision method based on PD-BPSO technology | |
| Datta | Efficient genetic algorithm on linear programming problem for fittest chromosomes | |
| CN107995027B (en) | Improved quantum particle swarm optimization algorithm and method applied to predicting network flow | |
| Singhal et al. | Greedy Shapley Client Selection for Communication-Efficient Federated Learning | |
| Jiang et al. | An improved quantum-behaved particle swarm optimization algorithm based on linear interpolation | |
| CN110322558A (en) | A kind of simplified method of the grid model folded based on fireworks algorithm tri patch | |
| CN107728469B (en) | A kind of laser particle control method | |
| Callaghan et al. | Evolutionary strategy guided reinforcement learning via multibuffer communication | |
| Hongwu | An adaptive chaotic particle swarm optimization | |
| Zhang | Derivative-free global optimization algorithms: population based methods and random search approaches | |
| Tansui et al. | An Enhanced Bat Algorithm with Random Walk for Solving Continuous Optimization Problems | |
| Toghraee | Evaluation of Meta-Heuristic Algorithms for Stable Feature Selection | |
| Poolsamran et al. | A modified marriage in honey-bee optimization for function optimization problems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200327 |
|
| RJ01 | Rejection of invention patent application after publication |