KEEL: CURRENT TRENDS AND NEW CHALLENGES
Juan Carlos Fernández
Pedro Antonio Gutiérrez
José Raúl Romero
The project KEEL-CTNC focuses on the extraction of knowledge based on genetic and evolutionary learning algorithms. This project integrates the construction and use of specific modules collecting the algorithms that make the state of the art on specific topics.
This general objective can be broken down into the following sub-objectives:
- To continue the development of the KEEL software tool (http://www.keel.es) that integrates the construction and use of specific modules collecting the algorithms that make the state of the art on specific topics, such as module for fuzzy systems, learning algorithms module from low quality data (data vague and incomplete), unbalanced dataset learning algorithms, etc.
- To establish new evolutionary learning models and/or their improvement and adaptation to specific contexts associated to the current trends in the development of knowledge extraction algorithms based on evolutionary learning: evolutionary fuzzy systems, evolutionary neural networks, evolutionary models based on coding Michigan, evolutionary learning from low quality data, evolutionary learning algorithms for unbalanced data bases, multiple instances based learning, evaluation of evolutionary learning algorithms based on measures of complexity on databases, etc.
- To develop studies on new challenges in extracting knowledge based on evolutionary learning, fixing our attention to the challenge ”scaling up for high dimensional databases”, the scaling up and high dimensional problems.
- To characterize specific actual problems and apply our evolutionary learning algorithms on them: market analysis, chemical and agroalimentary problems, mobile robotics, educational data mining (e-education) and web mining.
KEEL-CTNC is a coordinated project consists of five groups of Spanish researchers. Our main contribution lies in:
- Development of evolutionary models for learning association rules and subgroup discovery.
- Development of evolutionary learning models from unbalanced datasets.
- Development of evolutionary learning models for multi-instance data.
- Development of a multi-instance learning module for KEEL system.
- Participation in the development of a learning module using unbalanced data.
- Development in the dataset evaluation module (KEEL-dataset).
- Application of the developed models to web-pages categorization problems.
- Application of the developed models to web mining.
- Application of the developed models to problems of educational data mining.