YEREVAN, Armenia — A new study published in the IEEE Access, explores how explainable artificial intelligence can help engineers identify, understand, and troubleshoot performance problems in cloud computing environments. The paper, titled “Explaining Performance Issues of Cloud Applications From Logs Using Rule Induction and Dempster–Shafer Theory,” was developed through an international collaboration involving researchers from Armenia and Chile, including Dr. Nelson Baloian, associate professor at the University of Chile and a longtime collaborator and visiting professor at the American University of Armenia (AUA).
The research team includes Dr. Ashot Harutyunyan from Yerevan State University, Dr. Arnak Poghosyan from the Institute of Mathematics of the National Academy of Sciences of Armenia, Dr. Nelson Baloian, alumni Edgar Davtyan (BSCS ’21), and Karen Petrosyan (MSCIS ’25). The work was supported by the ADVANCE Research Grants program of the Foundation for Armenian Science and Technology (FAST).
Cloud computing systems generate vast amounts of operational data every day, and when applications slow down, experience performance degradation or fail entirely, engineers must determine the underlying cause. This process, known as root cause analysis, becomes increasingly difficult as cloud infrastructures grow more complex and produce larger volumes of logs, metrics, and monitoring information. The authors address this challenge by developing a framework that combines machine learning, log analytics, and uncertainty modeling to provide more transparent and interpretable diagnostics.
Unlike many AI systems that can detect anomalies without explaining their conclusions, the proposed framework focuses on interpretability, learns human-readable rules that connect patterns found in system logs, and performance metrics with specific types of failures. The researchers also incorporated the Dempster–Shafer theory, a mathematical framework for reasoning under uncertainty, enabling the system to assess the confidence of its diagnoses, and distinguish between strong evidence and ambiguous situations.
To evaluate the approach, the team tested the framework using data collected from cloud infrastructure hosts operating in a private cloud environment. The experiments examined several common failure scenarios, including processor overload, network congestion, and storage bottlenecks. According to the study, the framework successfully identified the causes of performance degradation while providing explanations that could be interpreted and validated by system operators. The results demonstrated that combining log data with performance metrics significantly improves both diagnostic accuracy and interpretability.
Dr. Baloian’s participation in the project reflects his long-standing research interest in interpretable artificial intelligence and collaborative technologies, where he currently serves as a visiting professor at AUA and authored more than 120 conference papers and over 40 journal articles. His research focuses on interpretable AI models, collaborative computer-supported learning, and digital systems.
Speaking about the impact of the project, Dr. Baloian remarked “I am very thankful for the opportunity I had to perform a research project with the help of the FAST-ADVANCE program and under the umbrella of AUA. In this project I worked with students who never had the opportunity to do serious research, nor participate as a co-author in an indexed journal article. Now we have five students from the AUA who have their works published in different conferences and journals. This will open the doors to graduate studies, research opportunities, and professional careers where scientific publication and research experience are highly valued. More importantly, it gives them the confidence to contribute to the scientific community and pursue ambitious projects in the future.”
The publication also highlights the contributions of Edgar Davtyan and Karen Petrosyan. Davtyan is currently pursuing a master’s degree in Data Science at the IU International University of Applied Sciences, and his research interests include causal inference and interpretable machine learning. He has previously worked as a researcher with FAST ADVANCE and as a Data Scientist at Picsart.
Petrosyan is currently pursuing a master’s degree in Computer and Information Science at AUA and alongside his studies, he works as a computer vision engineer at Pixeria Laboratory. His research interests include computer vision, generative models, and explainable artificial intelligence.
According to the authors, explainable AI is becoming increasingly important in modern cloud operations. While many machine learning models can identify anomalies, engineers often need to understand why a particular decision was made before taking corrective action. The framework presented in this study aims to bridge that gap by providing explanations that connect observed system behavior with potential root causes.
The researchers conclude that combining interpretable machine learning methods with uncertainty-aware reasoning can improve the reliability of automated diagnostics in cloud environments. They also note that the approach is designed to be applicable beyond the specific systems studied and could be adapted to other large-scale computing environments that rely on logs and performance monitoring data.
The publication demonstrates the value of international research collaboration, and highlights the contributions of AUA-affiliated researchers to ongoing advances in artificial intelligence, data science, and cloud computing.
Founded in 1991, the American University of Armenia (AUA) is a private, independent university located in Yerevan, Armenia, affiliated with the University of California, and accredited by the WASC Senior College and University Commission (WSCUC) in the United States. AUA provides local and international students with Western-style education through top-quality undergraduate and graduate degree and certificate programs, promotes research and innovation, encourages civic engagement and community service, and fosters democratic values.