Authors:
Luminita State
1
;
Catalina Cocianu
2
;
Panayiotis Vlamos
3
and
Viorica Stefanescu
2
Affiliations:
1
University of Pitesti, Romania
;
2
Academy of Economic Studies, Romania
;
3
Ionian University, Greece
Keyword(s):
Principal component analysis, fuzzy clustering, supervised learning, cluster analysis, pattern recognition, data mining.
Related
Ontology
Subjects/Areas/Topics:
Business Analytics
;
Communication and Software Technologies and Architectures
;
Data Engineering
;
Data Warehouses and Data Mining
;
e-Business
;
Enterprise Information Systems
Abstract:
The aim of the paper is to develop a new learning by examples PCA-based algorithm for extracting skeleton information from data to assure both good recognition performances, and generalization capabilities. Here the generalization capabilities are viewed twofold, on one hand to identify the right class for new samples coming from one of the classes taken into account and, on the other hand, to identify the samples coming from a new class. The classes are represented in the measurement/feature space by continuous repartitions, that is the model is given by the family of density functions (fh) hϵH, where H stands for the finite set of hypothesis (classes). The basis of the learning process is represented by samples of possible different sizes coming from the considered classes. The skeleton of each class is given by the principal components obtained for the corresponding sample.