Feature Selection Technique for Predicting Retention and Dropout Risk in the Alternative Learning System Using Principal Component Analysis
Ace C. Lagman
a
,
Maribel L. Campo
a
,
Fernand T. Layug
b
,
Roman M. De Angel
a
,
Richmon L. Carabeo
c
,
Jayson M. Victoriano
d
a College of Computer Studies and Multimedia Arts, FEU Institute of Technology, Philippines
b College of Computer Studies, Santa Rita College of Pampanga, Philippines
c Bataan Peninsula State University, Philippines
d Bulacan State University, Philippines
2024 IEEE 16th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management (HNICEM), (2024), pp. 1-5
Abstract: This study aims to identify the most critical attributes influencing retention and dropout risk in the Alternative Learning System (ALS) by analyzing various demographic, socio-economic, academic, and behavioral factors. Using Gradient Boosting Decision Trees (GBDT) for predictive modeling, the research explores feature importance scores to rank and prioritize the key attributes. The researcher used Knowledge Discovery in Databases as analytics methodology. Using principal component analysis, it was identified that regular attendance, availability, financial support, parental cohabitation (living together), and internet access positively influence retention. Furthermore, attending public schools, having a widowed parent, and possibly other features like distance to school are linked to increased dropout risk. The results provide insights into the main factors affecting student success, enabling more focused and data-driven interventions. The findings can help ALS administrators and educators develop personalized support plans for at-risk students and allocate resources more effectively.