One of the most popular pre-processing techniques in data mining and classification is feature selection. If the dataset is continuous, it might be challenging to pick the most important features from the vast amount of data that is available. This study presents a novel feature selection method for continuous data based on an overlapping class strategy (having normal and non-normal distributions). The classification uncertainty of an instance increases along with the area of class overlap between two classes, and it decreases along with the area. Three possible cases of class overlapping, i.e., complete class overlap, partial class overlap, and no class overlap, are caused by the uncertainty problem resulting from overlapping areas in classification. Therefore, the k-best feature selection in continuous features could make use of this variant overlapping area-based concept.