Diagnosing ADHD: Unleashing Fully Behavioral Attribute Insights with Supervised Learning

In recent years, the medical community has increasingly focused on Attention Deficit Hyper-activity Disorder (ADHD) due to its profound impact on individuals, marked by impulsivity, hyperactivity, and inattention. This attention has prompted the exploration of advanced computerized detection methods. This study introduces an innovative approach to ADHD detection by harnessing fully behavioral attribute data. The methodology emphasizes optimizing features to extract maximum information while minimizing their quantity for more efficient ADHD detection. Fully behavioral attributes comprise checklist features used for diagnosis, readily observable and compiled to aid in ADHD detection. Techniques like missing value replacement and min-max normalization further refine these attributes, resulting in heightened accuracy and robustness compared to methods reliant on datasets and methodologies. The study validates its simulation outcomes against existing literature and conducts experiments using supervised learning with various machine learning algorithms, including Gradient Boost (GB), Decision Tree (DT), K Nearest Neighbor (KNN), Naive Bayes (NB), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), all demonstrating promising results. The research highlights the effectiveness of fully behavioral attributes in addressing constraints and challenges, presenting a carefully curated selection of features classified through eight machine learning techniques. Notably, the GB algorithm emerges as a standout performer in supervised learning, achieving an impressive accuracy rate of 89.7%.