Abstract: Huge volumes of datasets with relatively higher number of dimensions are being collected by medical practitioners to identify the relevant features that cause a disease, which gives rise to an important technique, called feature selection, as the pre-processing strategy in obtaining knowledge and information from datasets. Feature selection is important when machine learning algorithms are applied on medical datasets to make the model easy to understand. Feature section techniques in medical domain should be model independent and at the same time should come with less number of features. Filter feature selection is independent of any model and helps in solving the curse of dimensionality. In this paper different types of filter feature selection algorithms are applied to A.P Liver dataset and performance is evaluated using sensitivity and specificity analysis.

Keywords: Feature Selection, Liver Diagnosis, Data Mining, A.P. Liver Dataset, Wrapper, Filter.