Chameleon Statistics is ideal for working with numerical data or images. Besides being an excellent general-purpose tool for clustering and classification, it will be especially valuable to anyone requiring outstanding quality graphical presentation of their data - whether that be for marketing to potential clients or for disseminating research results to peers. Chameleon Statistics Scientific Edition currently consists of the following major components:
Input Format(s): Datasets are lists of samples consisting of (numerical) feature vectors. Please see the FAQ for details about limitations on dataset size and related information.
Feature Selection: Principal components analysis, Variable weighting/masking.
methods. Single-link method (nearest neighbour), Ward's method
(sum of squares), Centroid method, Complete link (furthest neighbour), Group
average method, Median method, Lance Williams (parametric).
Partitional methods. K-Means.
Density Estimation: Multivariate Gaussian, Kernel methods, k-Nearest neighbours, Enhanced k-nearest neighbours.
Classification: Each of the density methods above can be used for classification.
Visualisation: Fully configurable 2D/3D scatter plots, density plots and contour plots, histograms, binary trees/dendograms, rotations/animations, bitmap displays.
Click on the links above for more detailed descriptions of each component.
*PLANNED FOR FUTURE RELEASE: Chameleon Statistics Professional Edition is currently under development. This will include all essential features of the Scientific Edition, as well as the following additional components: Automated Feature Extraction, Missing Value Handling, Multivariate Imputation, Mixed Data Type Handling & Anomaly and Outlier Detection.