A hybrid method based on MLFS approach to analyze students' academic achievement

How to get the information we need from the existing data, even extract the hidden message, and then transform them into knowledge, is an important skill we must learn in this era. Data mining technology in recent years is increasing attention in various fields, because of its extraordinary process. Usually, when we find more unexpected information, it may have the higher the value. This paper proposed Machine Learning Feature Selection (MLFS) combined with Support Vector Machine (SVM) method to extract and verify the key features that influence students' academic achievement in a Taiwan elementary school, and compare with the listing machine learning algorithms. The collected dataset include the students' graduated grades as decision features, and 15 condition features come from two databases “Student Profile” and “Tutorship Record”. After experiment, we found that when deleting five low influential features the quality and effectiveness of model is better than other conditions. Therefore, the proposed method is effective to enhance the accuracy and the quality of model. Finally, we use other classification algorithms to compare with the proposed method, and having the more accuracy (92.39%).