Estimating Feature Weights For Distance-Based Classification.
José Martínez Sotoca, J. Salvador Sánchez, Filiberto Pla · 2003
Abstract: This paper presents a new feature weighting method for distance-based classifiers. It is based on a generalized least squares minimization of a criterion function to estimate a feature relevance metric. Experiments over both artificial and real data sets illustrate the behaviour of this algorithm when irrelevant attributes and/or features with varying relevance are present. Effectiveness of the proposed technique is compared with that of other weighting methods. We also provide an empirical study on the effect of the training set size on performance of feature weighting models. 1