Classification of Study Program Competitiveness Levels in SNBT Using the Random Forest Method Based on Capacity and Regional Distribution Data

Abstract

The high level of competition in the Selection Nasional Berdasarkan Tes (SNBT) creates uncertainty for prospective students in choosing appropriate study programs, potentially leading to selection failure and uneven distribution of applicants across regions. This study aims to build a classification model to determine the competitiveness level of study programs in the SNBT by integrating data on capacity quotas and regional distributions using the Random Forest algorithm. The methodology follows a quantitative approach with a predictive computational workflow, utilizing secondary data from official national selection records. The dataset was automatically categorized into three tier Low, Medium, and High competitiveness using a quantile-based approach derived from available seats and application numbers. The predictive performance evaluated through a multi-class confusion matrix demonstrated an overall classification accuracy, precision, recall, and F1-score of 68.00% across all competitiveness tiers, indicating a balanced decision threshold free from class bias. Furthermore, feature importance analysis based on Mean Decrease Impurity revealed that institutional seat capacity is the primary computational driver, contributing 53.60% of the decision weight, followed by historical applicant volume at 29.40%, while localized geographic variables (provinces and regions) account for a combined 17.00%. These empirical findings confirm that while market supply-and-demand indicators dictate the baseline of SNBT competition, geographic factors function as critical spatial catalysts that mathematically refine student distribution choices.