Getting Started

This project implements the KMeans Clustering Algorithm in Java using n-dimensional input data. Data can be read from a csv file or be randomly generated. The visual panel displays the clustering in higher dimensional data by performing a principal component analysis to showcase the 2 dimensions of maximum variance. The standard implementation performs complete clustering with an l-2 norm but has support for l1 norm implementation.

Folder Structure

The workspace the following classes

• CentroidSetter: Handles manual setting of centorids via Mouse input, when centroid setting is set to manual mode.
• KMeans: Contains the KMeans clustering algorithm.
• PCA: Finds principal compenents by looking at dimensions of maximal variance.
• Panel: Contains the JPanel to showcase visualisation
• Point: Blueprint for point object. Point object used to depict n dimensional data and calculate norms.
• PointGenerator: Generates points randomly if input data is not given and/or if centroids are set to automatic mode.
• ReadPoints: Provides an implementation to read data from a .csv file. Follows the format 'num points', /n 'num dimesnions' /n followed by the coordinates of points on each new line.
• Runner:Runs program, holds important field variables shared by all classes.

Version 2.0 updates (in process)

• Finding a Loss function that can exclude points so as to perform incomplete clusters instead of absolute clustering.
• Provide implementation for K fold cross validation to determine optimal clustering
• Provide implementation for dropping fringe or outlier points.
• Addition of Kernel functions (Radial Basis, Polynomial)
• Fix normalization method for resizing csv data to fit within bounds of the visual panel.