Abstract:

A prediction mechanism is essential and effective in recognizing weak student and potential failure particularly in the prior step ahead of the final assessment yields a great chance for any necessary remedial help and support. The introduced fusion mechanism can be viewed as an early warning decision support system for identifying at-risk students and is important in strengthening, improving and sustaining their academic achievements throughout their academic years, thus eliminating the dropout percentages in school or universities. The decision support system introduced in this study can also help teachers or lecturers to prepare a course's teaching and learning materials and can be an alternative for placement examinations as well. The mechanism not only helps the school administration to deliver suitable and appropriate education courses and to increase the possibility for generating excellence in students, but is also beneficial to the administrative duties of the institutes. Motivated by the aforementioned reasons, this study constructs a fusion mechanism that integrates many machine learning/artificial intelligence techniques for teachers and lectures to understand those elements that contribute to academic achievement and to learn new methods at strengthening academic achievement. The fusion mechanism combines Multidimensional scaling (MDS), extreme learning machine (ELM) and knowledge visualization technique. The multidimensional scaling is implemented to overcome the Curse of dimensionality and to eliminate the calculation complexity. ELM has been demonstrated its superior generalization ability and superior forecasting performance. However, lacking of comprehensibility is one of the critical defect of black-box mechanism (that is, ELM) as well as impedes its practical application. Thus, decision tree (DT) is used to extract the inherent knowledge from ELM and the knowledge can be visualized in human readable format. The proposed fusion mechanism was examined by real data under three assessing criteria, namely forecasting accuracy, sensitivity and specificity. The proposed mechanism succeeded 89.65%, 89.12%, and 90.03% for three assessing criteria respectively. In addition, to examine the effectiveness of the proposed mechanism, this study compared it with other classifiers, namely support vector machine (SVM), random forest (RF) and rough set theory (RST). The results stated that the proposed mechanism still outperformed than other three models and it is a promising alternative for predicting student academic achievement. Furthermore, the teachers or lecturers can take this mechanism as a decision support system to implement subsequent educational interventions to improve student learning as well as strengthen their academic achievement.

Keywords:

Academic achievement, decision support system, decision making, knowledge visualization.