Prediction of Students' Final Results using JMP Pro 12
Oklahoma State University
There are a host of factors that can affect school results of teenagers. By understanding those factors, schools and parents can provide adequately timely supports, which may help students improve their school grades, and have continued better performances in future. We have obtained data of 1,044 students in two Portugal public secondary schools (“GP” - Gabriel Pereira and "MS" - Mousinho da Silveira), from UCI Machine Learning Repository. The original dataset contains 33 input variables and 1 nominal target variable identified as “G” which contains the average final grade of students. In Portuguese middle school, a student passes a class if he gets 50% or more in the final result. For the purposes of our analysis, we have created a binary target out of the average final grades from the original data set. The new target variable is called “Final” and describes if the students have pass or failed the class. There are no missing values in the dataset; no multicollinearity was observed among input variables, and important input variables for model building were selected using chi-square test.
In this project, we use JMP 12 Pro to develop models (Regression, Decision Tree, and Neural Network) to predict if students pass or fail their classes, based on their personal characteristics, habits, family and demographic characteristics, and school performances. Such models will not only help predict students’ results but also provide insights about how each specific factor impacts the performance of students.