TWM633222U - Electronic apparatus for providing multi-objective optimal solution based on genetic algorithm - Google Patents

Abstract

An electronic apparatus for providing multi-objective optimal solution based on genetic algorithm is provided, which includes a storage device with multiple modules and a processor for accessing and executing the multiple modules. The modules includes: an initialization module generating an initialize population; a fitness value calculation module calculating fitness values of all chromosomes; a fitness sharing module obtaining a filtered sample set from the initialize population based on the fitness values; a selection module obtaining a parent collection from the filtered sample set; a crossover module performing crossover on the parent collection to generate an offspring set; and a mutation module performing mutation based on the offspring set to update the initialize population. The fitness value calculation module, the fitness sharing module, the selection module, the crossover module and the mutation module are repeatedly executed until a termination conditions is met, and an optimal solution is obtained.

Description

Electronic device for providing multi-objective optimal solution based on genetic algorithm

The present invention relates to an electronic device, and in particular to an electronic device providing multi-objective optimal solutions based on a genetic algorithm.

In order to expand finance, each branch of the bank often needs to discover new cooperative merchants individually. In practice, which branch a merchant assigns to serve is a trade-off in the case of multi-objective conflicts, and it is a matter of multi-objective optimization. For example, only the branch closest to the merchant is considered, but the business attributes, manpower allocation, branch service capacity, and business target volume of each branch are not considered. Since different branches have different business attributes, manpower allocation, branch service capacity, and branch business target volume, how to assign the most suitable service bank branch to a cooperative merchant is one of the problems to be solved in this field.

This novel creation provides an electronic device that provides multi-objective optimal solutions based on genetic algorithms, and can obtain global optimal solutions.

The electronic device for providing multi-objective optimal solutions based on a genetic algorithm of the present invention includes: a storage device including a plurality of modules; and a processor coupled to the storage device and configured to access and execute the multiple modules, wherein these modules include: an initialization module, which generates an initialization group, and the initialization group includes a plurality of chromosomes; a fitness value calculation module, which respectively calculates a plurality of fitness values of the plurality of chromosomes; a fitness sharing module, Find the best fitness value among the fitness values, take out the chromosomes that are within a specified range from the chromosome with the best fitness value, and classify the chromosomes with the best fitness value and the chromosomes within the specified range into the screening sample set ; a selection module to obtain a set of mothers from the screening sample set; a mating module to perform mating on the set of mothers to generate a set of offspring; and a mutation module to perform mutation based on the set of offspring to update the initialization population. If the termination condition is not satisfied, the processor repeatedly executes the fitness value calculation module, fitness sharing module, selection module, mating module and mutation module until the termination condition is met and an optimal solution is obtained.

In an embodiment of the present invention, the selection module adopts grade roulette selection to obtain a parent set from the screening sample set.

In an embodiment of the present invention, the module further includes a termination judging module, which is used to judge whether the termination condition is satisfied.

In an embodiment of the new creation, the module further includes: a setting module, which is used to set multiple parameters, wherein the parameters include the number of chromosomes included in the initialization population, the length of chromosomes, the probability of mating, and the probability of mutation and termination conditions.

In an embodiment of the present invention, each chromosome includes a plurality of genes, and these genes respectively represent feasible solutions assigned by different merchants to different bank branches.

In an embodiment of the new creation, the mating module adopts one of single-point mating, double-point mating, multi-point mating and uniform mating.

In an embodiment of the present invention, the mutation module adopts one of single-point mutation, double-point mutation, mobile mutation and uniform mutation.

Based on the above, this disclosure uses the fitness sharing mechanism before the calculation selection operation, which can reduce the convergence speed of the algorithm, solve the problem of falling into the regional optimal solution due to excessive convergence in the traditional genetic algorithm, and then obtain the global optimal solution ,.

In order to make the content of the present disclosure easier to understand, the following specific embodiments are taken as examples in which the present disclosure can indeed be implemented. In addition, wherever possible, elements/components/steps using the same reference numerals in the drawings and embodiments represent the same or similar parts.

FIG. 1 is a block diagram of an electronic device for providing multi-objective optimal solutions based on a genetic algorithm according to an embodiment of the present invention. The electronic device 100 includes a processor 110 and a storage device 120 . The electronic device 100 of this embodiment provides multi-objective optimal solutions based on the genetic algorithm.

The genetic algorithm is to change/generate a new solution by adding random numbers, pairwise permutations, etc. in a solution set of a specific size, and select a better solution to keep it in the next generation to ensure that the solution is mutated. The direction can gradually converge. In the genetic algorithm, the following nouns and concepts will be used. A population is composed of a bunch of individuals. An individual is composed of a bunch of genes. Genes are the basic units of chromosomes. . A generation is the result of an evolutionary process. Chromosomes are represented by a series of integers, which are suitable for representing solutions to combinatorial optimization problems.

The processor 110 is, for example, a central processing unit (Central Processing Unit, CPU), a physical processing unit (Physics Processing Unit, PPU), a programmable microprocessor (Microprocessor), an embedded control chip, a digital signal processor (Digital Signal Processor) , DSP), Application Specific Integrated Circuits (Application Specific Integrated Circuits, ASIC) or other similar devices. The processor 110 can be coupled to the storage device 120 for accessing and executing multiple modules and various application programs stored in the storage device 120 .

The storage device 120 is, for example, any type of fixed or removable random access memory (Random Access Memory, RAM), read-only memory (Read-Only Memory, ROM), flash memory (Flash memory), hard Disc or other similar device or combination of these devices. The storage device 220 includes a plurality of modules, and the modules will be executed by the processor 110 after being installed. The modules include a setting module 121 , an initialization module 122 , a fitness calculation module 123 , a termination judgment module 124 , a fitness sharing module 125 , a selection module 126 , a mating module 127 and a mutation module 128 .

In general, the genetic algorithm includes the following four stages: the first stage, randomly generating n chromosomes (initialization population); the second stage, using the fitness function to calculate the fitness value of all chromosomes; the third stage, according to each chromosome The fitness value is used to select and copy chromosomes: the fourth stage: perform crossover and mutation actions on the remaining chromosomes. Completing the actions of the second to fourth stages above becomes an evolutionary process. The best solution can be obtained by repeating the actions of the second to fourth stages above until the termination condition (convergence) is met.

FIG. 2 is a flowchart of a method for providing multi-objective optimal solutions based on a genetic algorithm according to an embodiment of the present invention. In this embodiment, the above-mentioned modules are combined for description, please refer to FIG. 1 and FIG. 2 at the same time.

In step S205, the setting module 121 sets a plurality of parameters. Here, the parameters include the number of chromosomes included in the initialization population, chromosome length, mating probability, mutation probability and termination conditions. Chromosome number represents how many chromosomes each generation has. Chromosome length indicates how many genes are contained in a chromosome. The termination condition can be: the limit of the number of evolutions, that is, the maximum number of evolution generations; the resource limit of calculation (such as calculation time, memory occupied by calculation, etc.); the condition that a chromosome has met the optimal value, that is, the optimal value It has been found; the fitness value has reached saturation, and continued evolution will not produce chromosomes with better fitness values; or human intervention; and a combination of two or more of the above.

In step S210, the initialization module 122 is used to generate an initialization population (initialize population). The initialization population includes M chromosomes, and each chromosome includes N genes, where M and N are positive integers. For example, Table 1 is an example of assigning banks to multiple merchants. That is, N genes represent feasible solutions of N merchants and corresponding assigned bank branches. In terms of gene #1 of chromosome number #1, (A1, B1) represents a feasible solution in which merchant A1 is assigned to bank branch B1. In terms of gene #1 of chromosome number #2, (A1, B2) represents a feasible solution for merchant A1 to be assigned to bank branch B2. In Table 1, gene #1 represents the feasible solution corresponding to merchant A1, gene #2 represents the feasible solution corresponding to merchant A2, and gene #N represents the feasible solution corresponding to merchant AN. It can be seen from this that N merchants have M feasible solutions.

Table 1 chromosome number Gene #1 Gene #2 … Gene #N #1 (A1, B1) (A2, B2) … (AN, B6) #2 (A1, B2) (A2, B3) … (AN, B1) … … … … … #M (A1, B4) (A2, B5) … (AN, B3)

In step S215 , the fitness value calculation module 123 respectively calculates a plurality of fitness values (fitness values) of the plurality of chromosomes. The fitness value calculation module 123 calculates the fitness value using a fitness function. The fitness value represents the numerical performance of the degree of chromosome fitness, and different fitness functions will correspond to different calculation methods in different application scenarios.

In step S220, the termination judging module 124 judges whether the termination condition is satisfied. If the termination condition is not satisfied, continue to execute steps S225-S240; if the termination condition is satisfied, then in step S245, it is determined that the best solution has been obtained. The termination condition can be: the limit of the number of evolutions, that is, the maximum number of evolution generations; the resource limit of calculation (such as calculation time, memory occupied by calculation, etc.); the condition that a chromosome has met the optimal value, that is, the optimal value It has been found; the fitness value has reached saturation, and continued evolution will not produce chromosomes with better fitness values; or human intervention; and a combination of two or more of the above.

In step S225, the fitness sharing module 125 finds the best fitness value among the plurality of fitness values, takes out the chromosome with the best fitness value within a specified range, and assigns the chromosome with the best fitness value Chromosomes and chromosomes within specified ranges are sorted into screening sample sets.

In order to avoid premature convergence of the algorithm and fall into the optimal solution of the region, the fitness value of the excellent chromosome is shared with other chromosomes, so that other chromosomes also have the opportunity to be retained for the next generation, so as to improve the heterogeneity of the chromosome population. This is fitness sharing . And this mechanism uses the distance between chromosomes as the basis of sharing to avoid excessive sharing of excellent chromosome fitness to too many chromosomes.

In one embodiment, the fitness sharing module 125 can use a sharing function to find other chromosomes that are similar to the best chromosome corresponding to the best fitness value. The sharing function is a function that expresses the degree of close relationship between two individuals in a group, and the sharing function between individual x _i and individual x _j is sh ( d _ij ). Among them, d _ij is the intuitionistic fuzzy distance between individual x _i and individual x _j , which is used to measure the similarity between two individuals. The larger sh ( d _ij ) indicates that the two are closely related, that is, the similarity is high.

In another embodiment, it may also be set to take the one with the largest fitness value as the best fitness value. Then, share the chromosome corresponding to the best fitness value (hereinafter referred to as the best chromosome) to other chromosomes. Can be added to the screening sample set. Assuming that the fitness values of chromosomes include 4, 7, 12, 13, 10, and 15, then choose 15 as the best fitness value, and set a shared value radius of 3, then set the specified range to be greater than or equal to 12 (=15-3 ) and less than or equal to 15. Next, the chromosomes whose fitness values fall within 12-15 are classified into the screening sample set.

In step S230, the selection module 126 obtains a parent set from the screening sample set. That is, the selection module 126 is used to keep good chromosomes and exclude bad chromosomes.

In one embodiment, the selection module 126 adopts a rank based wheel selection. The selection module 126 first sorts the chromosomes according to the large fitness value. That is, each chromosome will be sorted according to its fitness value. Next, a ratio is calculated using roulette wheel selection. The roulette selection is to divide the roulette into N parts, and the size of the disk is determined according to the fitness value of each chromosome. The larger the fitness value, the larger the area. The grade roulette selection will increase or decrease the proportion of the area occupied by the chromosome according to the ranking of its fitness value. The larger the proportion of the area occupied by each chromosome on the roulette, the greater the probability of being selected into the mother collection (mating pool). The selection module 126 randomly selects a point of the roulette wheel, and the corresponding chromosome is selected into the parent set.

In step S235, the mating module 127 performs mating on the parent set to generate the offspring set. Mating is a major genetic operation. The mating process is to randomly select two chromosomes in the parent set (mating pool) to exchange information with each other, and then form another two new chromosomes. Whether two chromosomes will mate is determined based on the probability of mating, and the location of mating is determined randomly. The crossover may be single point crossover, two point crossover, multi-point crossover or uniform crossover.

In step S240, the mutation module 128 performs mutation based on the offspring set to update the initialization population. Mutation is to change a certain gene value with a certain probability (mutation probability) for the genetic genes of the offspring set, so as to prevent chromosomes from missing important information or falling into a local optimal solution during the process of replication and mating. The mutation is single point mutation, two point mutation, shift mutation, inversion mutation or uniform mutation. For example, in step S205, a mutation probability is firstly set for each chromosome. In step S240, when the mutation rate of the chromosome is lower than the set mutation rate, mutation will be performed.

Afterwards, the initialization population is updated, that is, the better-performing offspring chromosomes (mutated chromosomes) are replaced by poorer parent chromosomes in the initialization population. Return to step S215, recalculate the fitness value for the updated initialization population, and judge whether the termination condition is met in step S220.

That is, after the processor 110 sequentially executes the initialization module 122, the fitness value calculation module 123, the fitness sharing module 125, the selection module 126, the mating module 127 and the mutation module 128, the judgment module 124 is terminated. Determine whether the termination condition is met. If the termination condition is not satisfied, the fitness value calculation module 123 , the fitness sharing module 125 , the selection module 126 , the mating module 127 and the mutation module 128 are executed repeatedly until the termination condition is satisfied, and an optimal solution is obtained.

Accordingly, when the above-mentioned electronic device 100 is used to designate bank branches for multiple merchants, the electronic device 100 can assign service bank branches to merchants according to the best solution.

To sum up, this new creation adds a fitness sharing mechanism before the selection operation, which can reduce the convergence speed of the algorithm and share the fitness value of good chromosomes with other chromosomes. On the one hand, good chromosomes can be suppressed, and on the other hand On the one hand, other chromosomes can also have the opportunity to be retained in the next generation, and in this way, the heterogeneity among chromosome groups can be improved. In addition, this disclosure uses the distance between chromosomes as the basis for whether to share, which can avoid excessive sharing of excellent chromosome fitness values to too many chromosomes. Accordingly, the present disclosure can obtain the global optimal solution, which solves the problem of falling into the regional optimal solution due to excessive convergence in the traditional genetic algorithm.

100: Electronic device 110: Processor 120: storage equipment 121: Setting Module 122: Initialize the module 123:Fitness value calculation module 124: Termination Judgment Module 125:Fitness sharing module 126:Select module 127: Mating Module 128: Mutation Module S205～S245: steps of the method for providing multi-objective optimal solution based on genetic algorithm

FIG. 1 is a block diagram of an electronic device for providing multi-objective optimal solutions based on a genetic algorithm according to an embodiment of the present invention. FIG. 2 is a flowchart of a method for providing multi-objective optimal solutions based on a genetic algorithm according to an embodiment of the present invention.

100: Electronic device

110: Processor

120: storage equipment

121: Setting Module

122: Initialize the module

123:Fitness value calculation module

124: Termination Judgment Module

125:Fitness sharing module

126:Select module

127: Mating Module

128: Mutation Module

Claims (7)

An electronic device for providing multi-objective optimal solutions based on a genetic algorithm, comprising: a storage device including a plurality of modules; and A processor, coupled to the storage device, and configured to access and execute the modules, wherein the modules include: An initialization module that generates an initialization population, the initialization population includes a plurality of chromosomes; An fitness value calculation module, which calculates multiple fitness values of the chromosomes; A fitness sharing module, which finds an optimal fitness value among the fitness values, takes out the chromosomes that are within a specified range from the chromosome with the best fitness value, and divides the chromosomes with the best fitness value and Chromosomes within the specified range are classified into a screening sample set; a selection module to obtain a parent generation set from the screening sample set; a mating module that performs a mating on the set of parents to produce a set of offspring; and a mutation module that performs a mutation based on the set of children to update the initialization population, Wherein, if a termination condition is not satisfied, the processor repeatedly executes the fitness value calculation module, the fitness sharing module, the selection module, the mating module and the mutation module until the termination condition is met. conditions to obtain an optimal solution. The electronic device as claimed in claim 1, wherein the selection module adopts a level roulette selection to obtain the parent set from the screening sample set. The electronic device as described in claim 1, wherein the modules further include: A termination judging module, judging whether the termination condition is satisfied. The electronic device as described in claim 1, wherein the modules further include: A setting module sets a plurality of parameters, wherein the parameters include a number of chromosomes included in the initialization population, a chromosome length, a mating probability, a mutation probability and the termination condition. The electronic device as claimed in claim 1, wherein each of the chromosomes includes a plurality of genes, and the genes respectively represent feasible solutions assigned by different merchants to different bank branches. The electronic device according to claim 1, wherein the mating module adopts one of single-point mating, double-point mating, multi-point mating and uniform mating. The electronic device according to claim 1, wherein the mutation module adopts one of single-point mutation, double-point mutation, mobile mutation and uniform mutation.

Images