Metrics for classifying heterogeneous objects
Marc Bezem, K. Blok, Maarten Keijzer · Data Archiving and Networked Services (DANS) · 1998
We propose and discuss distance measures to compare objects that have heterogeneous sets of characteristics, such as encountered in, for example, medical diagnosis and information retrieval. We treat both boolean-valued and (scaled) real-valued characteristics. Weighting of characteristics is accomodated. The paper is a modified and extended version of [1]. 1991 Mathematics Subject Classification: 54E35, 54E50 1991 Computing Reviews Classification System: H.3.3 Keywords and Phrases: Metric, Classification, Information Retrieval 1. Introduction Consider two bitstrings ~s 1 and ~s 2 of equal length, say n. The Hamming distance d(~s 1 ; ~s 2 ) between ~s 1 and ~s 2 is by definition the number of positions in which ~s 1 and ~s 2 differ. It is well-known (and, moreover, easily verified) that d(~s 1 ; ~s 2 ) satisfies the following general conditions axiomatising a realvalued metric on a set S, in the special case here the set f0; 1g n of bitstrings of length n: (i) 8x; y 2 S (d(x...