The Kernel Trick and Clustering - the Kernel K-Means method

Expanded-Poster Poster
Main results apllying Kernel Trick on K-Means method

Abstract

The main objective of clustering methods is to segment the observations in subgroups or groupings, from measures of similarity between them. One of the classic examples is K-means, that Euclidean distance is used between similarity measures. However, there are some minor limitations, especially with regard to data that are not linearly separable. Kernel K-Means are as alternative to solve this problem. The methods were implemented in the R, and, as a result, it was noticed that in addition to solving well the issue of lack of linearity Kernel K-Means, is also a result of clustering when compared to classic K-Means.

Date
Event
23º SINAPE
Location
São Pedro, São Paulo, Brazil
The complete article about the Kernel K-Means can be found in the Expanded-Poster button, while the poster that was present it’s in the Poster Button.
Avatar
Mateus Maia
Master’s Student in Statistics

My research interests include machine learning, statistical learning, ensemble methods and data-driven solutions to real world problems.