Feature subset selection problem on microarray data

摘要: Recent advance of technology gave birth to tools such as microarray chips. The use chips enabled the scientists measure amount protein production from their genes in a cell, known gene expression data. classification cell samples by means data is hot research area. used for analysis massive and therefore features, i.e., genes, must be reduced reasonable level due computational cost experiments possibility misleading irrelevant genes. Therefore, usually, based on includes feature subset selection phase. This thesis aims develop tool that can during phase analyses. Three novel algorithms are proposed problem basic association rule mining. first algorithm starts with fuzzy partitioning discovers highly confident IF-THEN rules enable sample tissues. second search possible IFTHEN heuristic pruning approach which beam algorithm. Finally, third focuses hierarchical information carried through expressions constructing decision trees different performance measures. We found satisfactory results Leukemia Dataset. In addition, colon cancer dataset, construction showed good performance.