Loading...
Hybrid rough sets-population based system
Author(s)
Pattaraintakorn, Puntip
Cercone, Nick
Date Issued
January 1, 2007
Type
Conference Paper
Abstract
The integration of mathematical and statistical data analysis research can engender a novel and better approach, especially for survival analysis. This paper is devoted to Professor Pawlak and his ideas about rough sets and its applications. We propose MULTIHYRIS, an alternative hybrid intelligent system with a rough sets and population based approach for survival analysis. MULTIHYRIS is designed to increase the versatility and efficiency of survival analysis techniques. The MULTIHYRIS architecture incorporates mathematics - rough sets (with discernibility relations and individual patient consideration) - with statistics - Kaplan-Meier and Cox methods (with population estimates). The central idea behind MULTIHYRIS is to perform univariate analysis by using rough sets, database management and the Kaplan-Meier method with soft computing. All results from the univariate analysis are subsequently used in further mulitvariate analysis. In this stage, we provide two optional approaches to serve different requirements; rough sets integrated with database management and the Cox method. The former approach is able to produce decision rules while the latter generates a Cox model. Furthermore, set operations are used to unite these two outcomes and generate new reducts - hybrid reducts based on our rough sets-population based system. The informativeness of the rules and models can be verified within this analysis by validation processes and statistical tests. To demonstrate MULTIHYRIS, we have implemented it on a real-world geriatric data set, collected from the Dalhousie Medical School. © Springer-Verlag Berlin Heidelberg 2007.
Citation
Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics, 4400 LNCS(PART 2), 190-205, 2007