European Journal of Biomedical and Pharmaceutical Sciences

ROUGH SET BASED RULE INDUCTION APPROACH FOR SURVIVAL ANALYSIS

*Perumal Venkatesan

ABSTRACT

The boundaries in medical diagnosis and treatment are usually vague and imprecise. With vague information on signs and symptoms the physicians diagnose a patient and decide on the best way to cure them. Rough set is a leading soft computing method and its theory provides methods for knowledge extraction from imperfect data. This paper deals with a rough set based new approach for survival analysis. The decision problem is formulated over database on spinal tuberculosis patients treated under a randomized trial. The rule based method expresses the difference between the expected outcomes under the different treatment groups. The rules for different survival tendencies are also formulated using a prognostic index. The new frame work is combined with semi-parametric survival models to identify survival patterns. The usefulness of the method is demonstrated using outcomes of the trial.

Keywords: Rough set, rule induction, survival analysis, Kaplan-Meier, Cox model, spinal tuberculosis.