Arithmetic proficiency at the primary level is foundational for long-term academic success, yet data-driven approaches to predicting arithmetic outcomes are still scarce in the Democratic Republic of the Congo. This study investigates the cognitive, behavioral, and socio-environmental determinants of arithmetic performance among 10,994 primary school pupils across 35 schools in the 24 municipalities of Kinshasa. Eleven variables were recorded, including personality type, intelligence type (Gardner), behavioral traits, socioeconomic status, French proficiency, learning environment, concentration, tutor availability, school affiliation, and response time. Response time was operationalized as the manually recorded time taken to complete arithmetic tasks during supervised assessment sessions; possible observer bias is acknowledged as a limitation. Four classifiers were evaluated using a stratified 80/20 train-test split and 5-fold cross-validation: Decision Tree, Random Forest, Gradient Boosting, and Logistic Regression. Random Forest achieved the highest test accuracy (94.72%), likely reflecting non-linear interactions among learner and school-context variables, while Logistic Regression showed comparatively stable generalization (CV mean = 81.22%, SD = 5.72%). Feature importance identified response time (48.6%), school affiliation (9.6%), learning environment (7.9%), personality type (6.1%), and intelligence type (6.1%) as the strongest predictors. The outcome distribution was 75.2% Good, 15.5% Average, and 9.3% Poor. These findings demonstrate the potential of machine learning for early screening and targeted educational support in resource-constrained settings and represent the first large-scale ML study of primary arithmetic performance in Kinshasa, DRC.

Authors:
Vogel Kiketa, University of Kinshasa, DR Congo
Landry Mate, Université Pédagogique Nationale, DR Congo
Shambel Arega, Ethiopian Artificial Intelligence Institute, Ethiopia
David Kutangila, University of Kinshasa, DR Congo

About the Presenter(s)
Dr. Vogel Salvador is an Assistant Professor at the University of Kinshasa and Université Protestante au Congo. His interests include AI, cloud–edge computing, and digital education systems. He currently leads the development of CISNET, a university

Connect on Linkedin
https://www.linkedin.com/in/vogel-kiketa-a2b581146/

See this presentation on the full schedule – On Demand Schedule

Bookmark