Measuring Quality of Decision Rules Through Ranking of Conditional Attributes
One of the reasons for the wide popularity of rule classification systems is their ability to enhance understanding of mined data, and structures present in it. The discovered patterns are stated explicitly, which allows for more transparent descriptions of learned knowledge. To reach this goal of good descriptive and generalisation properties, induced rules need to be of a certain quality, which is typically measured by the predictive accuracy of the rule classifier. The paper presents research dedicated to measuring qualities of the inferred rules by taking into account a ranking of considered conditional attributes. Calculated quality measures along with supports of rules lead to construction of new classifiers, with improved parameters. The process is illustrated by a case of binary authorship attribution based on recognition of writing styles.
KeywordsDecision rule Quality measure Conditional attribute Ranking of attributes Weights of attributes
4eMka Software used for DRSA processing was developed at the Poznan University of Technology (http://www-idss.cs.put.poznan.pl/). The research works presented in the paper were performed at the Silesian University of Technology, Gliwice, Poland within the project BK/RAu2/2016.
- 3.Bayardo, Jr., R., Agrawal, R.: Mining the most interesting rules. In: Proceedings of the 5th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. pp. 145–154 (1999)Google Scholar
- 11.Mansoori, E.: Using statistical measures for feature ranking. Int. J. Pattern Recogn. Artif. Intell. 27(1), 1350003–14 (2013)Google Scholar
- 13.Sikora, M., Wróbel, L.: Data-driven adaptive selection of rule quality measures for improving the rule induction algorithm. LNCS 6743, 279–287 (2011)Google Scholar
- 14.Slowiński, R., Greco, S., Matarazzo, B.: Dominance-based rough set approach to reasoning about ordinal data. In: Kryszkiewicz, M., Peters, J., Rybiński, H., Skowron, A. (eds.) Rough Sets and Intelligent Systems Paradigms. LNCS (LNAI), vol. 4585, pp. 5–11. Springer, Berlin (2007)CrossRefGoogle Scholar
- 15.Stańczyk, U.: Dominance-based rough set approach employed in search of authorial invariants. In: Kurzyński, M., Woźniak, M. (eds.) Computer Recognition Systems 3, AISC, vol. 57, pp. 315–323. Springer, Berlin (2009)Google Scholar
- 16.Stańczyk, U.: On performance of DRSA-ANN classifier. In: Corchado, E., Kurzyński, M., Woźniak, M. (eds.) Hybrid Artificial Intelligence Systems. Part 2, LNCS (LNAI), vol. 6679, pp. 172–179. Springer, Berlin (2011)Google Scholar
- 19.Zielosko, B.: Optimization of decision rules relative to coverage—comparative study. In: Kryszkiewicz, M., Cornelis, C., Ciucci, D., Medina-Moreno, J., Motoda, H., Raś, Z. (eds.) Rough Sets and Intelligent Systems Paradigms, LNCS, vol. 8537, pp. 237–247. Springer, Berlin (2014)Google Scholar