Document clustering using GIS visualizing and EM clustering method

Abstract

This paper uses expectation-maximization clustering algorithm and a simple multidimensional projection method for visualization and data reduction. The multidimensional data is projected into a 2D Cartesian coordinate system. We run EM and K-Means algorithms on the transformed data. The system uses Microsoft Spatial Data Base Engine as a GIS tool for visualization. We used Expectation-Maximization (EM) and K-Means clustering algorithms of the Microsoft Analysis Services. The simple multidimensional projection method used in this paper tries to preserve the similarity relationships in original datasets.