Overview

GhostMiner is a data mining package from Fujitsu which featured data access from databases or spreadsheets; data visualisation; data preparation and selection; and a choice of data mining algorithms.

You can get an overview of Ghost Miner from this Online Tutorial %EXT%. If you have problems with the images, download the zipped version instead - it's a HTML file with better quality images.

Data visualisation methods include Multi Dimensional Scaling (MDS), and plots showing the distribution of values for each parameter, broken down by class.

Data pre-processing includes normalisation of numeric values, and dimension reduction using MDS.

Tree based methods, k-Nearest Neighbour (KNN), Misc.Neural Network and Misc.Support Vector Machine (SVM) algorithms are available for classification/prediction with testing using Cross Validation and Monte Carlo methods for model validation. Classification models found by different algorithms may be combined for greater robustness and accuracy.

Description of key features %EXT%