Automated learning of decision rules for text categorization
ACM Transactions on Information Systems (TOIS)
Unifying instance-based and rule-based induction
Machine Learning
Machine learning in automated text categorization
ACM Computing Surveys (CSUR)
Information Retrieval
Introduction to Modern Information Retrieval
Introduction to Modern Information Retrieval
Machine Learning
A Comparative Study on Feature Selection in Text Categorization
ICML '97 Proceedings of the Fourteenth International Conference on Machine Learning
Machine Learning Usefulness Relies on Accuracy and Self-Maintenance
IEA/AIE '98 Proceedings of the 11th International Conference on Industrial and Engineering Applications of Artificial In telligence and Expert Systems: Tasks and Methods in Applied Artificial Intelligence
A study of instance-based algorithms for supervised learning tasks: mathematical, empirical, and psychological evaluations
Improved heterogeneous distance functions
Journal of Artificial Intelligence Research
A Misuse Detection Agent for Intrusion Detection in a Multi-agent Architecture
KES-AMSTA '07 Proceedings of the 1st KES International Symposium on Agent and Multi-Agent Systems: Technologies and Applications
A Snort-based agent for a JADE multi-agent intrusion detection system
International Journal of Intelligent Information and Database Systems
Hi-index | 0.00 |
In this paper a family of rule learners whose application is carried out according to a partial-matching criterion based on different purity measures is presented. The behavior of these rule learners is tested by solving a Text Categorisation problem. To illustrate the advantages of each learner, the MDL-based method of C4-5 is replaced by a pruning process whose performance relies on an estimation of the quality of the rules. Empirical results show that, in general, inducing partial-matching rules yields more compact rule sets without degrading performance measured in terms of microaveraged F1 which is one of the most common performance measure in Information Retrieval tasks. The experiments show that there are some purity measures which produces a number of rules significantly lesser than C4-5 meanwhile the performance measured with F1 is not degraded.