Leveraging LLMs and MLPs in Designing a Computer Science Placement Test System

Introductory Computer Science (CS) programs at higher education institutions include a variety of introductory courses commonly known as CS1 and CS2; some institutions offer additional variations (e.g., CS0, CS1.5). Accurate placement of students in these courses is vital for students' success. This paper introduces a machine learning-based CS placement test system using LLMs and classifiers, incorporating a concept inventory (CI) assessment. Five LLMs were evaluated, and the best one was chosen to generate questions. The questions were stored in a database, allowing customized placement tests through an interactive assessment system. After the test, students' performance data were used to train three simple MLP classifiers, achieving over 83% accuracy in a case study with 46 participants.