Object Analyst provides unsupervised classification based on the k-means clustering algorithm. Clustering algorithms search for generic data patterns among the attribute [variable] spaces. No target variable is identified, as such. Clustering algorithms search for patterns among all the input attributes related to objects and group them into relatively homogeneous clusters.
The similarity among the members within a certain cluster is maximum and the similarity to the members of other clusters is minimum. The similarity or dissimilarity is defined in some distance function. The objective of a good clustering method is to produce high-quality clusters in which intra-cluster similarity (within cluster) is high and inter-cluster (among clusters) similarity is low.
The goal of clustering is normally descriptive and to discover new categories. The algorithm learns from observation rather than examples and, therefore, does not require a prior hypothesis or training set. The core objective is finding the natural boundaries in attribute space for the number of clusters you specify.
© PCI Geomatics Enterprises, Inc.®, 2026. All rights reserved.