A New Dynamic Niching Genetic Algorithm And The Application For Data Clustering

Genetic algorithms because of its simple to use and its superior performance on optimization, making it received wide attention and continuous research. Niching genetic algorithms are a product of genetic algorithm for continuous development, the introduction of niching technology can better maintain the diversity of the species. But the algorithm still has some drawbacks and requires further improvement.In this paper, we propose a dynamic niching genetic algorithm with diverse niche radii (DNDR) for multimodal optimization with application for data clustering. In our method, we first randomly generate a value to initialize the niche radius of all niches. After that, each niche radius is updated individually during the evolutionary process by employing two newly introduced operators. The first operator exploits the relationship between the relative positions of the nearest niches to determine whether the two niches should be combined. The second operator, however, devotes to eliminate individual that does not belong to a niche. These operators are implemented on all the niches in the population. As a result, diverse niche radii can be maintained during the evolutionary process. After implementation of these two operators, the FS is then applied to individuals of the same niche. In order to ensure the newly generated niche candidates being sufficient diverse, the value used to initialize the niche radius is updated at the end of the algorithm.The proposed method has been evaluated on a set of standard multimodal functions, the results show that our method can achieve superior performance and significantly outperform normal niching methods. At the same time, in order to better verify validity of this algorithm, the improved algorithm is applied to the clustering analysis. Experimental results show that the improved genetic clustering algorithm can achieve superior performance.