Type-2 fuzzy logic in breast cancer relapse time prediction using genetic algorithm

摘要

Microarray analysis and gene expression profile have been widely used in tumor classification, survival analysis and ER (Estrogen Receptor) status of breast cancer. Sample discrimination as well as identification of significant genes, has been the focus of most previous studies. Fuzzy Partitioning Method (FPM) has been introduced in our previous study which generates a rule base for fuzzy systems based on multi-level partitioning of gene expression profiles. The aim of this research is to improve proposed method by optimizing the size of selected genes in a procedure which has been statistically analyzed in this study. To handle the existence of uncertainties in proposed model, the TSK (Takagi-Sugeno-Kang) model of Interval Type-2 Fuzzy Logic System (IT2-FLS) has been used to predict the relapse time of breast cancer. In addition, a new method has been developed to consider the uncertainties of the fuzzy rules generated in the model. The results with 95% confidence interval show an improvement in the relapse time prediction with respect to the previous works.