xi. IMPLEMENTATION OF TUMOR CLASSIFICATION. BASED ON FITUR
SELECTION USING. FARNEM ALGORITHM. Name. : Rindang Nisa Prasasti.
NRP.
IMPLEMENTATION OF TUMOR CLASSIFICATION BASED ON FITUR SELECTION USING FARNEM ALGORITHM Name NRP Department Advisor I Advisor II
: Rindang Nisa Prasasti : 5107 100 166 : Teknik Informatika, FTIF-ITS : Prof. Ir. Handayani Tjandrasa, M.Sc, Ph.D : Dr. Nanik Suciati, S.Kom, M.Kom
Abstract The problem of analyzing gene expression data is its characteristics of high dimensionality and small sample size, and the great number of redundant genes not related to tumor phenotype. An alternative to solve such problems is by applying pattern recognition method that can analyze data with number of samples which is relatively small than its dimensions. Feature selection is viewed as an important preprocessing step for pattern recognition. In this experiment, FARNeM algorithm is introduced to deal with these problems. FARNeM algorithm is constructed based on neighborhood rough set (NRS) model. NRS is the generalization of classical rough set model with neighborhood relation. NRS used to reduce features by assigning different thresholds. Effectiveness of the use of the algorithm is proved by classification process of original features (i.e. features is used as input feature selection) and optimal features (i.e features obtained by using FARNeM algorithm). Two kinds of classifier, Support Vector Machine (SVM) and K-Nearest Neighbor (KNN), are used to classify the features. The experimental results show that the average accuracy obtained using optimal features is higher than the original features. The accuracy of original features is 69.52% (SVM) and 74.92% (KNN) and the accuracy of optimal features is 88.57%% (SVM) and 79.94% (KNN).
Keywords :, Genes Expression, Fitur Selection, FARNe M Algorithm, Neighborhood Rough set
