Predictive Model with Machine Learning for Academic Performance

Abstract

Academic achievement (AP) in recent years has shown minimal progress with a difference of 0.05%, according to the report made by the Program for International Student Assessment (PISA). For this reason, the objective of this research is to build a predictive multiclass classification model for the AP of students in an elementary school. It was conducted with a dataset of 218 third-year high school students. The Cross Industry Standard Process for Data Mining (CRISP-DM) methodology was used to create the model, which consists of 6 phases and is effective in data mining (DM) projects. The random forest (RF) algorithm was also used. The results indicated that the RF model obtained the highest prediction rates compared to other studies, with an accuracy of 95% of the model, respectively. Finally, it is observed that the attributes that mostly influence prediction are the scores of Ability 02 end of I bimester, Positive Impression, Ability 01 end of I bimester, Ability 03 end of I bimester, and Adaptability. Thus, it is concluded that academic attributes are more relevant than psychological attributes in predicting RF.