Understanding the SNN Input Parameters and How They Affect the Clustering Results

2. Related Work

Clustering is the task of identifying sets of segments or clusters that group similar objects. A cluster is a collection of data objects that have more similarities between them and are dissimilar to objects that belong to other clusters (Han & Kamber, 2001).

Density-based clustering approaches were developed based on the notion of density (Han & Kamber, 2001). These algorithms perceive clusters as dense regions of objects in a space separated by regions of relatively low density. This kind of algorithms is useful to filter out noise and for discovering clusters of arbitrary shapes (Ye & others, 2003). DBSCAN (Ester et al., 1996) and OPTICS (Ankerst, Breunig, Kriegel, & Sander, 1999) are major representatives of this class of clustering algorithms, being DBSCAN the most representative density-based clustering algorithm. Many of the available density-based algorithms were derived from DBSCAN, which was introduced by Ester (Ester et al., 1996) and was specially designed to treat spatial databases.