K-mean clustering

Problems involving many features for large datasets become quickly intractable, and it is quite difficult to evaluate the independence between features. Any computation that requires some level of optimization and, at a minimum, the computation of first order derivatives, demands a significant amount of computing power to manipulate high-dimension matrices. As with many engineering fields, a divide and conquer approach to classifying very large datasets is quite appropriate. The objective is to reduce very large sets of observations into a small group of observations that share some common attributes:

Visualization of data clustering

This approach is known as vector quantization. Vector quantization is a method that divides a set of observations into groups of similar sizes. The main benefit of vector quantization is that analysis using a representative of each group is far simpler than an analysis of the entire dataset [4:2].

Clustering, also known as cluster analysis, is...