A Machine Learning Approach to Identify the Feature Importance for Admission in the National Military High Schools

Abstract

Abstract The article provides the impact of different averages (feature importance) within the admission exam for the national military high schools using and testing three supervised machine learning algorithms: logistic regression, K-Nearest Neighbors, and random forest. For this purpose, I have used the list with the results of candidates compounded by 430 rows, an unclassified document posted on the national military high school website, with details about: the final admission grade, the general grade for graduating of the secondary school, the general grade obtained at the national assessment, the mark obtained at admission test from Romanian language and mathematics items, etc. From the machine learning perspective, I have built a Jupyter notebook, a Python code using the specialized ML libraries (numpy, pandas, matplotlib, sklearn). In conclusion, the logistic regression algorithm identified the ‘feature importance’ (how each variable contributes to the predicted model) for admission in the national military high school: the mark obtained at admission test from Romanian language and Mathematics items - 4.821834, the general average obtained at the national assessment - 0.584434, the general average for graduating of the secondary school - 0.285446, etc. These are the expected results based on the admission methodology for the national military high schools.