An integrated DEA–machine learning–system dynamics framework for faculty-level graduate employability efficiency measurement

Abstract

Graduate employability is an important outcome of higher education, but its evaluation often focuses primarily on employment rates or other individual performance indicators. Such measures may not provide a complete representation of faculty performance because faculties operate with different levels of institutional resources while producing multidimensional graduate outcomes. This concept paper proposes an integrated three-phase framework combining Data Envelopment Analysis (DEA), machine learning (ML), and system dynamics (SD) for faculty-level graduate employability evaluation, with Universiti Teknologi MARA (UiTM) as the intended context. In Phase 1, an output-oriented Multiple-Layer DEA approach is proposed to evaluate how faculties transform academic resources into hierarchically structured graduate employability outcomes. In Phase 2, ML-based clustering will identify faculties with similar efficiency characteristics and improvement patterns. In Phase 3, SD will examine the potential long-term effects of alternative improvement strategies while accounting for feedback relationships and time delays. Faculty-level secondary information from the Graduate Tracer Study and relevant university records will support the proposed analysis, while comparable historical information may be used to support the calibration and validation of the dynamic model. Existing studies have primarily linked two analytical stages, including DEA with regression, clustering or ML, and SD with policy evaluation. The proposed framework extends these approaches by establishing a sequential connection between efficiency diagnosis, faculty profiling, and dynamic policy simulation. The framework is expected to provide a more comprehensive basis for faculty benchmarking and university policy planning.