Abstract:
Clustering is one of the critical approaches in data mining, which aims to divide and group data into groups that have similar characteristics. Some of the main problems in clustering are grouping with high-dimensional datasets that have many attributes, both numerical and categorical data types, high time consumption, calculation complexity, and overhead, which makes some algorithms in the clustering process less efficient. The clustering algorithm often used is K-Means, but the algorithm needs to improve in the computational method that is quite long. The results of grouping data on K-Means must be defined first, as well as noise or outliers, due to outliers in grouping results and difficulties in finding global solutions that can reduce the quality of clustering results on the K-Means algorithm. Therefore, this research is focused on developing the K-Means Algorithm to improve model performance as well as the quality of the resulting clusters by combining the K-Means (KM) method with Invasive Weed Optimization (IWO), and Genetic Algorithm (GA) called the Hybrid IWOKM-GA method to produce data clustering with close genetic diversity. The results showed that the Hybrid IWOKM-GA method managed to find the best clustering results with a Cost Function Value value of 2400.51, almost three times when compared to the K-Means model combined with GA, which has a computational time of 328.08 seconds.