Classification of Bandung Regency Well Water Based on Mineral Elements using Agglomerative Clustering
Katharina Oginawati, Indah Rachmatiah Siti Salami, Nurul Fahimah, Yuli Sri Afrianti, Grace Angelia, Fadhil Hanif Sulaiman

Bandung Institute of Technology

Abstract

Wells are sources of water used to meet daily needs, such as drinking water. However, the quality of well water cannot always be considered safe depending on the mineral elements in it. Therefore, a method is needed to classify well water quality safely. Cluster analysis is a statistical method for grouping data by maximizing the similarity of data in a cluster and minimizing the similarity of data with other clusters [1]. This research was conducted on 6 mineral elements (i.e., As, Cd, Fe, Mn, Pb, and Zn) in well water spread across 160 locations in Bandung Regency. The locations are clustered using Agglomerative Clustering with various linkage parameters (i.e., Single, Average, and Complete) as well as various distance metrics (i.e., Euclidean and Manhattan) [2]. The aim of this research is to identify the most optimal linkage cluster analysis method based on distance metrics and evaluate well water quality based on regional characteristics analysis. Each linkage is evaluated to determine the optimal number of clusters (k) using the Mojena method. Then, the optimal linkage is selected using the Silhouette Coefficient. The two best cluster methods were obtained- those are for metric Euclidean is Average Linkage and for metric Manhattan is Single Linkage. Next, each cluster member in both methods is mapped based on drinking water quality standards, and then the mean is calculated for each cluster. This is useful for determining the level of well water quality in each cluster. These clusters have a range of well water quality levels from 1 to 4, with the higher the level the better. Then, an analysis of regional characteristics was carried out to evaluate the map of the distribution of well water quality in each region. The results show that the Average linkage with Euclidean metrics is more representative of well water data than the Single linkage with Manhattan metrics. The results of this research can be an important basis for safe water resource management.

Keywords: Well, Agglomerative, Bandung Regency

Topic: Others