Abstract:

The paper discusses the issue of using random methods for the purposes of generating a set of test cases, which is the basis for the implementation of the software testing process. Two methods, most often used in practice, were characterized: random testing and partition testing. Presented characteristics include a description of the conditions for which one of the above-mentioned random testing methods is better than the other, in terms of the probability of detecting at least one error using equally numerous test data sets created on their basis. Particular attention was focused on the analysis of how to generate a set of test cases for the partition testing method. Contrary to the majority of works related to this issue, the presented considerations include the cost of determining such a set. The main result of this work is to obtain the method of determining the partition testing strategy, which can be the basis for the implementation of the program testing process. This strategy is obtained as a result of solving the problem of two-criteria optimization, with the following component criteria taken into account as the component criteria: probability of an event that the set of test cases will allow to detect at least one error and the cost of extracting a certain number of partitions from the domain of the tested program and drawing some sets of test cases from them.