Identifying Important Attributes for Secondary School Student Performance Prediction

In traditional teaching, teachers cannot effectively provide targeted guidance to students to improve thier performance. Moreover, the current student performance prediction model is complicated in design, and there are a lot of irrelevant data input. In this paper, taking secondary school students as an example, we have selected a number of important attributes that can influence the prediction of student performance from the attribute set. The attribute set includes past school grades, and several attributes related to individuals, society and school. The two core courses (i.e. mathematics and portuguese) were modeled under Binary/Five-level classification and regression tasks. Three feature selection methods are tested under three input configurations, the accuracy of the model is verified by decision tree algorithm. This makes it more possible for teachers to provide professional guidance to students, and can effectively reduce the dimension of the dataset of the student performence prediction model to improve the prediction efficiency.