Analiza dyskryminacyjna - stan aktualny i kierunki rozwoju

The aim of the discriminant analysis is to partition the multivariate feature space into subspaces in order to separate observations belonging to different classes. In other words, its task is to find a model that can give class descriptions on the basis of a set containing previously classified observations. Then the model is applied to classify new ones with a minimum error. Founded in 1936 by Fisher, the discriminant analysis had become an important part of multivariate statistical analysis. It has many applications and is an obligatory procedure in many available data mining systems. In Poland prof. Józef Kolonko has been one of the pioneering statisticians interested in discriminant analysis. He published his book on discriminant analysis in 1980, based on cybernetics. Therefore the analysis had a broader meaning, including both supervised and unsupervised classification.