Experimental designs in software engineering: d-optimal designs and covering arrays

Abstract

For over a century, Design of Experiment (DOE) techniques have been applied to testing in large problem domains such as agriculture, chemistry, medicine, and industrial design. Recently, the application of DOE has appeared in component-based software testing. This is a natural extension, as software testing is a complex problem that suffers from a combinatorial explosion. Exhaustive testing is not possible in most systems. In this paper, we focus on three areas: (1) the application of DOE techniques to software testing, (2) improved algorithms for screening tests involving categorical factors, and (3) construction methods for generating covering arrays.