Neurodevelopmental disorders significantly impact individuals’ learning, working abilities and social interactions with Attention Deficit Hyperactivity Disorder (ADHD) being one of the most common and challenging to diagnose accurately due to the reliance on subjective questionnaires in clinical assessment [Amado-Caballero et al., Artif. Intell. Med. 143:102630 (2023)]. ADHD is a common neurodevelopmental disorder that is experienced by people across the world and has its effects on learning, work, and interpersonal relationships. Deep learning and machine learning in ADHD diagnosis is a topic of continual research in this paper, based on information received from ECG, actigraphy, fMRI, and neuropsychological tests. In this study, the effectiveness of machine learning classifiers for ADHD diagnosis was investigated using the HYPERAKTIV dataset and several classifiers, such as Random Forest, LightGBM, and SVM. In the present work, we show that using multiple datasets, such as adding Conners Continuous Performance Test II (CPT-II), improves model performance; Random Forest achieves the highest accuracy (0.708) on the merged data. The findings stress the need for implementing a more inclusive data integration solution and indicate that there is still a lack of understanding of these phenomena on larger samples.