Enhancing K-Means Clustering Model to Improve Rice Harvest Productivity Areas in Aceh Utara Using Purity

Abstract

To optimize the performance of the clustering process using K-Means, an optimalization approach employing the Purity algorithm is needed. This research was tested on a dataset of rice harvest productivity areas in Aceh Utara Regency by comprehensively analyzing the number of iterations and the DBI values produced by K-Means and Purity K-Means in clustering priority and non-priority rice production areas. This is in line with the efforts of the Regional Government to implement rice production intensification programs in Aceh Utara Regency. From the testing of Purity K-Means, an average of 5, 2, 2, 5, and 3 iterations were obtained from all tested datasets sequentially from 2019 to 2023. Meanwhile, from the testing of conventional K-Means, the average number of iterations obtained was 5.4, 4.8, 4.2, 5.6, and 3.8 iterations, sequentially. This indicates that the clustering performance conducted by Purity K-Means is better than conventional K-Means. The DBI values obtained from Purity K-Means for the entire dataset sequentially are 0.6781, 0.4175, 0.4419, 0.6182, and 0.4973. This value is lower compared to the DBI values obtained from conventional K-Means, which are 0.7178, 0.6025, 0.4971, 0.7222, and 0.5519, respectively. This also indicates that the validity level of the clustering results performed by Purity K-Means is higher than conventional K-Means.