Influence of Data Pre-Processing on Genetic Algorithm Based Clustering

Clustering technique is used to partition or group a given dataset or pattern into disjoint clusters. In this paper, for data clustering, k-medians based clustering algorithm along with a Genetic Algorithm is used to cluster the data points. In distance-based partitioning, a major problem is that features with higher ranges have more influence on the distance measure resulting in poor clustering. To overcome this problem, in this paper, we have pre-processed the dataset using min-max scaling to bring all features into the same range. This has helped to achieve the expected clustering accuracy as the dataset has been smoothed. We have worked with Iris and Wine datasets. We have observed that clustering accuracy of our proposed method is improved by 1.124% for Iris dataset and 38.043% for Wine dataset compared to the clustering accuracy of the modified k-means method published earlier.