English abstract
This thesis aims to develop an ATAP-BKT (Absence rate, Time response, AI detection, and Performance) AI model to predict the students’ performance in Computer programming courses at the undergraduate level in Oman. It employs a quantitative research method with a descriptive design incorporating universal literature to suggest design and develop the ATAP-BKT model to answer the thesis questions.
The ATAP-BKT model was developed to enhance the existing Bayesian Knowledge Tracing (BKT) algorithm, designed to address specific challenges in tracking and predicting student performance in computer programming courses. The development process involved adapting and extending the BKT algorithm by introducing new parameters that reflect crucial factors influencing student learning. The final model comprises five phases: the initial assessment phase, ATAP-BKT monitoring phase, Adaptive practice phase, final assessment phase, and analysis phase. Also, it introduces five new parameters: (1) time response P(dt), (2) the structure of questions P(Q), (3) use of AI tools to answer P(Ai), (4) Attendance P(At), and (5) Activities P(Ac).
The sample consisted of 29 ILT undergraduate students in the Education college at Sultan Qaboos University (SQU) who enrolled in Computer programming courses. Accordingly, data were collected using four instruments: initial and final tests, Learner experience questionnaire, classroom settings, and ATAP-BKT data. Logistics metrics (Root Mean Square Error (RMSE), Area Under the Curve (AUC), and Accuracy) were used to analyze data.
The main finding of this thesis indicated that the ATAP-BKT model tracked academic performance by identifying students' skills, calculating factors influencing the learning process, and predicting future performance with 89% accuracy. This is considered reasonable compared to previous studies where BKT models typically achieved 75% to 85% accuracy. This result was validated using cross-validation techniques to ensure the model robustness and generalizability to new data. The result also illustrates that developing new parameters on the ATAP-BKT model has enhanced prediction accuracy. The thesis recommended integrating the ATAP-BKT model with emerging educational technologies. Further, it suggested that future research areas use ATAP-BKT in international studies, such as TIMSS, PIRLS, and STEM.
