Applications of machine learning in biomedical research has been an important is- sue in computer science since the 1960s. The complexity of data to be analyzed has been increasing dramatically since more accurate medical approaches and advanced techniques are used [Morik 2010]. Since nowadays there are tools (more details in Chapter 2) that allow us to analyze genetic expression data of patients suffering under various diseases, finding genetic biomarkers has become an useful approach to predict patients’ prognosis. Current researches for neuroblastoma focus on finding biomarkers for neuroblastoma cancer to understand its nature [Schramm et al. 2009; Oberthuer et al. 2010; Takita et al. 2011; North et al. 1997]. Since the considered data is very high dimensional, different methods of feature selection can be applied to reduce the number of dimensions [Ng 2004; Zou and Hastie 2003; Schowe 2011]. Feature selection filters those variables which contain no or less important informa- tion concerning the learning task. Considering this data we are aiming in comparing two different approaches of feature selection in order to distinguish a better approach to decrease the number of attributes and increase the prediction accuracy on the neuroblastoma data.