Çok boyutlu kutu kapsama yöntemi ve mikrodizi analizine uygulanması

Data mining is an important tool employed in many bioinformatics domains includinggenomics, proteomics, evolution, systems biology, and microarray analysis. A recentlydeveloped classifier, hyper-box enclosure (HBE) algorithm is an efficient method forclassification problems, and it does not require parameter optimization depending on data typefor higher prediction accuracy. The goal of this thesis is to understand, improve HBEalgorithm, and apply it for microarray analysis which is an important bioinformatics problem.The most important use of data obtained from microarray measurements is theclassification of tumor types with respect to specific genes that are either up or downregulated in specific cancer types. However, almost all classification algorithms used inmicroarray analysis usually require optimization to obtain accurate results depending on thedata type. Additionally, it is highly critical to find an optimal set of markers among those upor down regulated genes that can be clinically utilized to build assays for the diagnosis or tofollow progression of specific cancer types. On the base of these necessities, we employ amixed integer programming based classification algorithm named hyper-box enclosuremethod (HBE) for the classification of some cancer types with a minimal set of predictorgenes. This method, a user friendly and efficient classifier, may also allow the clinicians todiagnose and follow progression of certain cancer types.