Artificial intelligence (AI) has become a buzzword in nearly every sector of society, including education, where it has been utilized in various forms, such as automating administrative tasks, supporting students’ needs, and enabling intelligent tutoring systems that provide real-time pedagogical guidance. Additionally, AI has been employed for course personalization, often through data-driven methods that recommend courses to enhance students' competencies.

We explored the personalization of higher education course modules based on students’ competencies. We proposed QUAIMOP–Quiz and AI-based Module Personalization system. It comprises two components: QuizBot, a tool that asks questions based on students’ degree path and background knowledge on the domain, and assesses students’ competencies using a rule-based algorithm; and a Random Forest (RF) machine learning algorithm that provides module recommendations based on students’ competencies. The novelty of this approach is the use of dynamic questions in QuizBot rather than relying on a static questionnaire.

We demonstrated QUAIMOP in a Software Engineering Project course, which included several modules, each consisting of a set of questions for the QuizBot, used to assess students’ competencies. Students with varying capabilities receive different module recommendations based on their quiz performance. This approach enables students to save time, enhance their competencies, and avoid reviewing content they already understand.

QUAIMOP was developed using React, Python, MariaDB, and Node.js technologies. We utilized MariaDB as our database solution to enable tracking students’ performance and facilitating data analysis. Backend interaction with the user interface (UI) was handled using Node.js with the Express framework.

We generated sample inputs to evaluate the accuracy of QUAIMOP and the appropriateness of the module recommendation. We designed 20 test scenarios to assess whether the returned value of the test case is based on the core function's logic. The results indicated that 95.65% of functional tests passed successfully. Additionally, 20% of the dataset was reserved for testing the RF algorithm. This test data was used to evaluate the algorithm’s ability to accurately recommend the potential module selection.

We tested the UI’s adaptiveness to the recommended modules using Cypress. In addition to automatic testing, we conducted manual system testing to verify the accuracy of the recommendations and the UI’s responsiveness and adaptiveness to the module changes. Overall, the evaluation results indicated that the algorithm performs as expected, providing students with appropriate module recommendations through an adaptive UI.

As the next step, we aim to integrate QUAIMOP into the university's Moodle platform for the upcoming semester and evaluate its performance in a real-world environment.

The study demonstrates that course modules can be tailored to meet the needs of diverse students by assessing their knowledge levels before the course commencement. As a result, both modules and the UI can be tailored to each student’s needs, allowing for more efficient use of time and supporting successful course completion. Educators can adopt this approach to provide courses for multidisciplinary audiences with varying levels of prior knowledge.

Keywords: Quizbot, Personalized learning.