Akademik Veri Yönetim Sistemi
5th International Conference on Informatics and Software Engineering, IISEC 2026, Ankara, Türkiye, 5 - 06 Şubat 2026, ss.473-478, (Tam Metin Bildiri)
This paper proposes a multi-stage Time Series (TS), Machine Learning (ML), and Deep Learning (DL) framework for classifying ADHD and control subjects using the HYPERAKTIV activity-level dataset. TS analyses and decomposition examine the influence of gender, Attention Deficit Hyperactivity Disorder (ADHD) medication, and antidepressant use on heart rate variability and response times, revealing more frequent residual outliers and longer reaction times in the ADHD group, as well as higher ADHD and antidepressant use prevalence among males. Activity sequences are then modeled with Random Forest (RF), Support Vector Classifier (SVC), and Gradient Boosting (GF) algorithms using ten hourly activity inputs, generating labeled activity-pattern images for subsequent DL experiments. Six custom image-based deep architectures, including CNN, SwinTransformerV2 (SwinV2), ConvNextV2, CoAtNet, and two hybrids that fuse SwinV2 with ConvNextV2 and CoAtNet are trained on these images, achieving the best results at 0.9894 and 0.9824 accuracy, respectively, in binary ADHD versus non-ADHD classification. These findings highlight the diagnostic potential of activity-level representations and hybrid deep architectures for automated ADHD screening, while pointing to future work on improving transformer-based models and designing more efficient hybrid networks.