A Web-Based Machine Learning System for Early Detection of ADHD Symptoms in Peruvian High School Students

Curi J.; Farro A.; Felix R. · International Conference on Computer Supported Education, CSEDU - Proceedings (2026) · DOI: 10.5220/0014675300004021

Resumen

Attention Deficit Hyperactivity Disorder (ADHD) affects a substantial proportion of school-age populations; however, many cases remain undetected in school environments where traditional assessment is time-intensive and difficult to scale. This study presents a web-based application designed to support early ADHD screening in secondary school students through three cognitive tasks: Stroop Task, CPT, and SST. The study was conducted in three phases. First, a real-world pilot deployment (D1, n = 324, ages 12–13) evaluated feasibility, usability, and classroom scalability under school supervision and was not used for model training or performance reporting due to the lack of complete diagnostic ground truth access. Second, the Voting Classifier (Random Forest, SVM, Gradient Boosting; hard voting) was trained and internally evaluated using a labeled dataset provided through institutional psychopedagogical records (D2, n = 349, ages 12–17). Third, an external validation phase assessed sensitivity on confirmed ADHD cases (D3, n = 45, ages 14–17), yielding TP = 39 and FN = 6 (recall = 86.7%). The platform is intended as a decision-support screening tool for educational professionals in coordination with school psychopedagogical services, complementing—not replacing—clinical diagnosis.

Instituciones: Universidad Peruana de Ciencias Aplicadas