Features selection approach for non-invasive evaluation of liver fibrosis

In many domains, a range of input variables are considered, not clearly which of them are most useful, or indeed are needed at all. Data are often collected on variables that are not only correlated, but also are large in number. This makes the data process, interpretation and detection of its structure difficult. Feature selection is a pattern recognition approach to choose important variables according to some criteria, in order to improve the decision process by removing the redundant information. The intent of this work is to provide a feature selection approach, based on the analysis of correlations between the explanatory (input) variables and the outcome variables, to improve the classification process of the liver fibrosis stages, using both the naive Bayes classifier and the probabilistic neural network model.