Font Size: a A A

A Cellular Genetic Algorithm Based On The Clustering Technology For Multi-objective Optimization

Posted on:2014-08-27Degree:MasterType:Thesis
Country:ChinaCandidate:X BianFull Text:PDF
GTID:2268330422960012Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The laws of nature is a beacon to guide the progress of human science, mostscientific studies come from the inspiration of nature. Human got inspiration whichfittest survives from the biological genetic evolution mechanisms, so some scholarsproposed the genetic algorithm. The algorithm is iterative calculation, through geneticoperation is gradually close to the optimal solution of the problem. Genetic algorithmis different from the other search and optimization algorithm for the advantage, sinceit has the adaptive intelligent, characteristics of implicit parallelism. The algorithm issimple and easy to realize, it’s commonly used to solve search and optimizationproblems with limited cost. With the development of the society, we face more andmore of multi-objective optimization problems, and more and more people focus theirattention on how can we solve multi-objective optimization problems effectively andsimply.In this paper, we described and summarized the cGA and MOP comprehensively.Also, in order to solve the problem of multi-objective optimization the author use thecGA as the model sample, combines the theory and the existing algorithm, adds theclustering classification and local search, to optimization the algorithm. Finally weproposed A Cellular Genetic Algorithm Based on the Clustering technology forMulti-objective Optimization, the algorithm using the K-means clustering accordingto the similarity of individuals in each generation, using local search strategy toimprove the individuals whom clustered, and then eventually optimizing thealgorithm’s results in the multi-objective problem.At the end of this paper, we analyze the K-means algorithm and the local searchstrategies according to the experiment, and also analyze the influence about clusteringnumber and operations to our algorithm. Then compared against three famousalgorithms through two metrics which are Generational Distance and Spread. Itclearly outperforms our algorithm is better. Additionally, we present the results of thefour functions experiments for the multi-objective optimization problem by the Pareto front, proved the new algorithm we proposed which using K-means algorithm andlocal improve strategy can obtain better results in solving multi-objective problems.
Keywords/Search Tags:Cellular Genetic Algorithm, Multi-objective optimization, K-means, Pareto optimal solution set
PDF Full Text Request
Related items