Clustering Research Based On The Multivariant Optimization Algorithm (MOA)

With the rapid development of the science technology and the era of progress, the computer technology is growing popularity, and advantages of the storage media is more and more obvious. Under the context, a kind of multi-group variant optimization algorithm was proposed, that is Multivariant Optimization Algorithm (MOA).Which algorithm’s core idea is that to search the solution space comprehensively through global and local search alternately. What need to be clear is that in this process, the global search is responsible for randomly exploring the whole solution space while the local search is responsible for exploiting a few potential local areas. The algorithm is able to achieve as expected function, mainly due to its construction of special structure of the MOA, which can fully use the informations during looking for the optimization process, and reach the effects of share and memorizing. The structure also help to make process memory happen when global and local search to alternate search. Ultimately, to achieve the purpose of efficient sharing of information and memory storage.In this paper, the purpose to analysis and study is to further validate the MOA clustering algorithm which is based MOA algorithm, has higher optimization ability and better stability. Firstly, designing and realising the test software platform to compare clustering algorithms performance. Then, through the five standards UCI data sets to compare optimization capability of the clustering algorithm and other five population intelligent optimization clustering algorithms in this platform, including ant colony clustering algorithm, particle clustering class algorithm, genetic clustering algorithm, fish clustering algorithm and fireflies clustering algorithm. Finally, doing these six kinds of clustering algorithms experiments under the same tests times and iteration times, then output each fitness value in the a figure to compare the performance optimization. The experimental results show that even though the MOA clustering algorithm convergence has a slower speed, the result is much better than other five algorithms. Therefore, it comes to the conclusion that MOA clustering algorithm is amore scientific and reasonable way, in the same time it has better search ability, better stability and better convergence.