Authors:
Iñaki Fernández de Viana
;
Pedro J. Abad
;
José L. Álvarez
and
José L. Arjona
Affiliation:
Universidad de Huelva, Spain
Keyword(s):
Outlier detection, One class classification, Novelty recognition, Web wrapper, Enterprise application integration.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Cross-Feeding between Data and Software Engineering
;
Enterprise Information Systems
;
Enterprise Integration Strategies and Patterns
;
Knowledge Management and Information Sharing
;
Knowledge-Based Systems
;
Software Engineering
;
Symbolic Systems
Abstract:
Reducing maintenance costs of Enterprise Application Integration (EAI) solutions becomes a challenge when you are trying to integrate friendly web applications. This problem can be solved if we use automated systems which allow navigating, extracting, structuring and verifying relevant information. The verification task aims to check if the information is correct. In this work we intend to solve the verify problem regarding One Class Classification Problem. One Class Classification Problems are classification problems where the training set
contains classes that have either no instances at all or very few. During training, in the verify problem, we only have instances of the classes we know. Therefore, the One Class Classifier techniques could be applied. In order to evaluate the performance of these methods we use different databases proposed in the current literature. Statistical analyses of the results obtained by some basic One Class Classification techniques will be described.