Directions for impoving the effectiveness of the CBUA method with dynamic test suites

DOI: 10.31673/2412-9070.2026.028108

Abstract

In the context of increasing complexity in modern software systems, ensuring their reliability and quality remains one of the key challenges in software engineering. One of the promising directions is the use of mutation testing, which allows evaluating the quality of test suites by creating modified versions of the program — mutants. However, applying this approach under conditions of a dynamic test suite, where test cases are generated or extended during execution, makes it impossible to use a number of traditional optimizations. In response to these challenges, approaches for predicting mutation testing outcomes without actually executing the mutants have been proposed, including Predictive Mutation Testing (PMT). At the same time, the Coverage-Based Unsupervised Approach (CBUA), which maintains mutation score when the test suite changes, proves potentially more effective in such conditions.
The aim of this study is to systematize and analyze modern approaches to improving the effectiveness of methods for predicting mutation testing outcomes, with a focus on their applicability within the CBUA method. Special attention is given to analyzing strategies for enhancing the effectiveness of such methods, including the selection of relevant features for machine learning algorithms, the construction of adequate models of mutation testing, and the identification of threats to the validity of the results.
As a result of the study, a group of features promising for improving prediction models was identified, major threats to the validity of experiments were outlined, and directions for further research were defined. In particular, it is recommended to explore historical features in the context of mutation testing prediction methods, to utilize more effective static and dynamic features, and to develop more accurate mathematical models that take temporal characteristics into account to improve the overall
effectiveness of the CBUA method.

Keywords: software quality, software testing, mutation testing, test outcome prediction.