In the field of research, the growth of data mining using the k-means technique is well accepted, which involves the extraction of data from datasets with some limitations. To overcome the drawback of this technique we employed the kernel concepts and resolved the cluster inadequacy of separability. We have proposed an optimization technique to include a fisher’s discriminant analysis into the kernel of particle swarm optimization concept with GA (Genetic Algorithm) to evaluate fitness function. The fitness function value is required to select offspring for the next generation. The consequence was to reduce the noise and enhance the performance of clustering. The GA (Genetic Algorithm) was employed to optimize the objective of the fitness function by providing the input parameter. The kernel technique performs more fault identification features than principal component analysis. Results found are more beneficial by this method like fitness value, stopping criteria, and the average distance between individuals. In this research paper, we discuss the comparative analysis with the objective function of k-means and kernel fisher’s discriminant analysis in the domain of the large dataset. The fitness value of proposed KFDA is smaller than k-means fitness.