Gen ifade miktarı verisi analizi için yinelemeli öbek eliminasyon yöntemlerinin iyileştirilmesi

The computational and interpretational difficulties caused by the ever-increasing dimensionality of biological data generated by new technologies pose a significant challenge. Feature selection (FS) methods aim to reduce the dimension, and feature grouping has emerged as a foundation for FS techniques that seek to detect strong correlations among features and identify irrelevant features. In this thesis, methods that utilize feature grouping in a supervised context were developed. We initially tested the effects of different clustering algorithms on SVM-RCE and observed the best performance with K-means. In the first developed method, Recursive Cluster Elimination with Intra-cluster Feature Elimination (RCE-IFE), both cluster and intra-cluster elimination is performed recursively in each cluster reduction step. Our experimental findings imply that RCE-IFE provides robust classifier performance and significantly reduces feature size while maintaining feature relevance and consistency. In the second developed Grouping – Scoring – Model (G-S-M) based study, G-S-M_Rep, we use prior knowledge to form disease groups, and select top features as representative of each group. These representative features are learnt by the model in a cumulative manner. Results show that G-S-M_Rep attains satisfactory model performance with a small number of features. Consequently, this thesis presents methods based on feature grouping and focuses on improving feature reduction capability, classification performance, feature relevancy, and feature consistency.