Academic Research

R. Akay, A. N. Toprak, O. Sahin

2nd THEQC International Conference on Quality Assurance and Accreditation

Quality assurance depends not only on program-level evaluation but also on recognising at-risk students in time to intervene. This paper presents a hybrid, machine learning-enhanced student risk analytics model built on learning-outcome-based data structures, together with its implementation on the UniMetrik platform.

• accreditation
• risk analysis

R. Akay, O. Sahin, A. N. Toprak

2nd THEQC International Conference on Quality Assurance and Accreditation

Accreditation processes still rely largely on manually prepared documentation, fragmented data and evaluation criteria whose interpretation varies from person to person. This reduces consistency in accreditation work and increases the workload. This paper proposes an approach that embeds large language models into the accreditation process as a decision-support layer and supports evidence-based quality assurance in a scalable way.

• accreditation
• large language models

O. Sahin, M. Zhang, A. Poth, O. Rrjolli, A. Faes, P. Teng, K. Xue, W. Ma, A. Arcuri

arXiv preprint (cs.SE)

In black-box testing of REST APIs, relying on OpenAPI schemas alone falls short, while tool-specific supplementary data definitions create vendor lock-in and cognitive load. To overcome this, the study adopts the OpenAPI Overlay standard, which standardises test inputs, and extends the advanced testing tool EvoMaster to support it. The practical applicability of the proposed solution was tested on the real APIs of five companies, including the Fortune 500 firms Volkswagen and Meituan as well as Unimetrik. The results demonstrate that the Overlay standard gives industrial REST API testing a more effective, standardised and tool-independent structure.

• software testing
• rest api
• fuzzing