Implementation of Feature Selection to Reduce the Number of Features in Determining the Initial Centroid of K-Means Algorithm

Clustering is a data mining method to group data based on its features or attributes. One reasonably popular clustering algorithm is K-Means. K-Means algorithm is often optimized with methods such as the genetic algorithm (GA) to overcome the problem of determining the initial random centroid. Many features in a dataset can reduce the accuracy and increase the computational time of model execution. Feature selection is an algorithm that can reduce data dimension by removing less relevant features for modeling. Therefore, this research will implement Feature selection on the K-Means algorithm optimized with the Dynamic Artificial Chromosome Genetic Algorithm (DAC GA). From the experimental results with ten datasets, it is found that reducing the number of features with feature selection can speed up the computation time of DAC GA to K-Means process by 17,5%. However, all experiments resulted in higher Sum of Square Distance (SSD) and Davies Bouldin Index (DBI) values in clustering results with selected features.