An Implementation of Support Vector Machine Classification for Developer Academy Acceptance Prediction Model

Abstract

In order to prepare graduates with work readiness in the IT industry, specifically in mobile apps development, one of its ways is to create a Developer Academy where final year students are prepared in an intensive program for two consecutive semesters to learn the stages of mobile apps development. To ensure the quality of participants in the Developer Academy, a set of selection procedures needs to be prepared, consisting of Aptitude Test, Portfolio Showcase, and Individual Interview. The problem arises when applicants are far more than the class capacity. Hence selection procedures take a longer time. The Developer Academy registration team record showed a ratio of 1:12, which overburdens the team when it comes to selecting the applicants. More effective procedures are needed with the help of machine learning tools to help with decision making. This study aims to produce a prediction model for developer academy applicants. Several classification algorithms such as k-nearest neighbors, support vector machine, decision tree, and random forest were analyzed. Data was collected from 527 valid applicant's data which submit complete documents based on due date, other applicants who did not submit complete documents were not included in the analysis. Preliminary findings from the study show that the Support Vector Machine algorithm performs best with an accuracy of 86% and this score was then increased by applying oversampling and kernel tricks to get an accuracy rate of 98%. Hence it can be concluded that the prediction model has excellent performance