Classification of Go Public Companies in Indonesia by using Fisher's Discriminant Function, Euclidean and Mahalanobis Distance

The classification of objects into a group is expected to be solved with minimum errors. Euclidean distance, Mahalanobis distance and Fisher's quadratic discriminant are used for classification of companies listed on Indonesia Stock Exchange (IDX). Eleven variables are obtained from indicators of Kompas100 Index stocks selection. Then, the dimensionality of the data are also reduced by principal component analysis (PCA), which combine linearly the original correlated variables into new uncorrelated variables (principal component). PCA is used to obtain the visualization of data which have dimension more than three. Based on the results of testing data with eleven variables classification, Mahalanobis distance gives the most minimum classification error. However in visualization, Fisher's quadratic discriminant gives the best result.