Explainable Priority Assessment of Software-Defects using Categorical Features at SAP HANA

Abstract

We want to automate priority assessment of software defects. To do so we provide a tool which uses an explainability-driven framework and classical machine learning algorithms to keep the decisions transparent. Differing from other approaches we only use objective and categorical fields from the bug tracking system as features. This makes our approach lightweight and extremely fast. We perform binary classification with priority labels corresponding to deadlines. Additionally, we evaluate the tool on real data to ensure good performance in the practical use case. CCS CONCEPTS • Software and its engineering → Software maintenance tools; • Computing methodologies → Feature selection.

Testing Very Large Database Management Systems: The Case of SAP HANA

“…Assessing the risk of a software defect manually via its report is an expensive task and the assessment results are unreliable. Therefore, we apply, among others, machine learning (ML) techniques (support vector machines, decision trees) to automate the risk assessment based on available data about bugs [51]. Similarly, we use ML-based approaches to classify issues for root cause analysis while replaying recorded workload data [47].…”