International Journal of Computer Techniques Volume 12 Issue 4 | Testing Machine Learning Algorithms: Software QA Strategies for Non-Deterministic and Data-Driven System
Testing Machine Learning Algorithms: Software QA Strategies for Non-Deterministic and Data-Driven Systems
Author: Oyindamola Adebayo
Wilmington University, Delaware, USA
Contact: +1 (484) 626-6150
Journal: International Journal of Computer Techniques (IJCT)
Volume: 9 | Issue: 4 | Publication Date: July – August 2022
ISSN: 2394-2231 | Journal URL: https://ijctjournal.org/
Abstract
This paper investigates the challenges of software quality assurance (QA) in machine learning systems, which are inherently non-deterministic and data-driven. It reviews limitations of traditional testing and introduces specialized QA strategies such as metamorphic testing, adversarial testing, drift monitoring, and explainability checks. Case studies from autonomous systems, finance, and healthcare illustrate practical applications.
Keywords
Machine Learning Testing, Software Quality Assurance (QA), Non-Deterministic Systems, Data-Driven Software, Model Validation Techniques
Conclusion
ML systems demand new QA paradigms due to their stochastic nature and evolving behavior. This paper outlines a comprehensive framework combining traditional software testing with ML-specific evaluation techniques. Future directions include ethical QA automation, standardized ML QA protocols, and interdisciplinary AI quality engineering roles.
References
Includes 15+ references from IEEE, Elsevier, Springer, and peer-reviewed journals covering ML testing, QA frameworks, and non-deterministic system validation.
