Dissimilarity Measure Properties of Binary Test Patterns of Controlled Random Tests

Abstract

The problem of application of the dissimilarity measures for binary test sequences is investigated. Their relevance in generating controlled random tests is substantiated. Dissimilarity measure AD(Ti, Tk) between test sets Ti and Tk is considered, using the characteristic of the distance D(ti,j, tk,r) between ti,j and tk,r, which is based on the determination of independent pairs of identical data ti,j = tk,r belonging to two patterns Ti and Tk. This measure AD(Ti, Tk) allows us to estimate the degree of difference between two test sets Ti and Tk, which may be indistinguishable when using other difference measures, including the Hamming distance. Upper and lower estimates for the measurement of dissimilarity are obtained for the case of inverse test patterns and arbitrary test patterns Ti and Tk with different combinations of their weights wi and wk. Examples of calculating the boundary values of the specified dissimilarity measure and the ratio of their values are given. Experimental results confirm the correctness of the obtained boundary values of the indicated dissimilarity measure AD(Ti, Tk) and show the possibility of their application for its evaluation.

Keywords

• dissimilarity measure
• hamming distance
• levenshtein distance
• test
• test pattern
• controlled random tests