Abstract: This study deals with how to efficiently rank features of datasets. As we may know well, reducing the dimensionality of datasets (i.e., feature reduction) is an important step in pattern recognition tasks and exploratory data analysis. Quite often, feature ranking is required before completing feature reduction. In this study, a novel classifier-free feature ranking approach based on the combination of both weighting features and ISE (Integrated Squared Error) criterion is proposed. ISE is measured in terms of the modified non-parametric Parzen window density estimator in this study. The advantage of the proposed approach is that it allows us to make an efficient and effective non-parametric implementation and requires no prior assumption. The experimental results demonstrate that the proposed approach here is very promising.