Analisis Perbandingan Clustering Based dengan Density Based dalam Mendeteksi Outlier

Abstract

Data is information that is collected and recorded, which can then be processed or analyzed. Data can take the form of numbers, text, images, sound, or other forms that represent facts or characteristics of an object, event, or phenomenon. However, the larger and more complex the volume of data generated, the more difficult it is to find outliers in the data without the appropriate tools and techniques that can disrupt the statistical analysis of the data. This research was conducted to find a better method for detecting outliers by comparing clustering-based and density-based methods. In the clustering-based method, K-means is used to cluster data that are similar to each other compared to other clusters. It calculates the Euclidean distance for data furthest from the cluster and identifies that data as an outlier. On the other hand, the density-based method uses DBSCAN to cluster data with sufficiently high density and identifies data with relatively low density as outliers. Based on this research, it was found that the number of outliers detected by DBSCAN is lower compared to K-means. The results show that 23 data points are identified as outliers using DBSCAN, while 27 data points are identified as outliers using K-means. According to these results, it can be concluded that with the same data, the density-based method can identify outliers better than the clustering-based method.