Abstract:
This research presents the ATML system (Adaptive Testing
using Machine Learning: ATML) with the following objectives: (1) to
develop and evaluate the effectiveness of the ATML system, and (2) to
assess programming knowledge and skills through the ATML system.
The system development employed the K-Means clustering technique in
conjunction with the Elbow Method and Silhouette Score to categorize
the difficulty levels of test items for adaptive testing aimed at grouping
learners by ability levels. The sample group comprised 23 students from
the Computer Technology program at a university in Ubon Ratchathani
Province. The results indicated that: (1) the system effectively grouped
test items to classify the examinees' ability levels, and the expert
evaluation ( by three experts) rated the system at a good level of
performance; and (2) the experimental evaluation of the sample group
through the ATML system showed that most students had very low to
low knowledge and skills (48.73%), moderate level (34.78%), and only
17. 39% demonstrated high to very high levels of proficiency. These
results indicate that the system can appropriately classify examinee
levels. The outcomes reflect the existing need for skill development
among learners. Furthermore, the ATML system demonstrates the
potential to provide effective, time efficient assessment that appropriately
adapts to individual learner abilities. It can be applied to support efficient
teaching and learning evaluations in future educational settings.