Using MDA to Improve Naïve Bayes Classification for Students Performance Prediction

Abstract

Adoption of information technology in education sector made data grow exponentially in this field. There are lot of data produce by education institution such as registration, teaching and learning, administration, and examination. Those data can be more useful if we can turn them into knowledge. Data mining is tools used to uncover pattern hidden in data and turn them into knowledge. Naïve bayes classifier is a classification algorithm based on naïve theorem. This algorithm has high accuracy and fast. However, Naïve bayes has no ability of select the best features since all attributes in this naïve bayes theorem consider as equal. However, it is common in the real data that there are attributes that higher dependency degree than others and many data have attributes that considered as superfluous or redundant, hence this paper proposes Maximum Dependency Attribute (MDA) to tackle that problem. MDA is feature selection technique based on rough set that is used to select the best features and Naïve bayes is used to predict student performance. Based on the experiment show that this proposed model has accuracy 79%. The result has improvement compare to Naïve bayes without MDA with accuracy 68%.