Interpreting Decision Tree model using ELI5 for modelling student performance: Case of a structural analysis graduate attribute-based teaching course

Decision tree is used in this study as a classification machine learning to model year mark, using mainly formative assessments. Binary digits are used to classify the different marks, which illustrate performing versus non-performing students when considering four assessments (e.g. 1 to 4). A structural analysis course is used for the application of the decision tree model and is part of the Advanced Diploma programme of engineering technologists. This fairly new programme is mainly designed and developed on engineering competency namely graduate attributes. The results revealed a higher an accuracy, precision and f1-score relatively high during the testing period. Although Decision tree model can be easily interpreted through its tree structure, the interpretability of the model can be enhanced through ELI5, which could reveal the ranking of assessments as influencing factors for the whole dataset (globally) and for each student performance prediction (locally), during testing. Hence, the decision tree algorithm was proven elegantly to be a suitable candidate for modelling the competence level of students and could guide teaching staff on student fate, in an academic environment