Weighted Proportional Sampling : AGeneralization for Sampling Strategies in Software Testing

Abstract

Current activities to measure the quality of software products rely on software testing. The size and complexity of software systems make it almost impossible to perform complete coverage testing. During the past several years, many techniques to improve the test effectiveness (i.e., the ability to find faults) have been proposed to address this issue. Two examples of such strategies are random testing and partition testing. Both strategies follow an input domain sampling to perform the testing process. The procedure and assumptions for selecting these points seem to be different for both strategies: random testing considers only the probability of each sub-domain (i.e. uniform sampling) while partition testing considers only the sampling rate of each sub-domain (i.e., proportional sampling). This paper describes a more general sampling strategy, named weighted proportional sampling strategy. This strategy unifies both strategies into a general model that encompasses both of them as special cases. This paper also proposes an optimization model to determine the number of sampled points depending on the sampling strategy