IMPROVED HETEROGENEOUS DISTANCE FUNCTIONS

Instance-based learning techniques typically handle continuous and lin ear input values well, but often do not handle nominal input attribute s appropriately. The Value Difference Metric (VDM) was designed to fin d reasonable distance values between nominal attribute values, but it largely ignores continuous attributes, requiring discretization to map continuous values into nominal values. This paper proposes three new heterogeneous distance functions, called the Heterogeneous Value Diffe rence Metric (HVDM), the Interpolated Value Difference Metric (IVDM), and the Windowed Value Difference Metric (WVDM). These new distance fu nctions are designed to handle applications with nominal attributes, c ontinuous attributes, or both. In experiments on 48 applications the n ew distance metrics achieve higher classification accuracy on average than three previous distance functions on those datasets that have bot h nominal and continuous attributes.