Prediction of Attention Deficit Hyperactivity Disorder and Autism Spectrum Disorder Patients Emotion Recognition using Machine Learning Models

To design a machine learning-based approach to predict and differentiate emotional recognition patterns in patients with ASD and ADHD, thus aiding in precise diagnosis and personalized care strategies. Current approaches of group 1 emotional recognition in ADHD and ASD rely mainly on traditional statistical methods and a relatively small number of datasets. Such approaches group 2 often consider only simple facial expression analysis and vocal tone evaluations without integrating the multi-modal features, leading to lower classification accuracy. Conversely, the method under this study employed machine learning algorithms, namely Random Forests and Deep Neural Networks, to analyze multi-modal data including facial expressions, vocal tones, and physiological signals of 200 samples from ADHD and ASD patients. The Random Forest model was 92 % accurate in classification compared to existing methods, and the Deep Neural Networks further improved detection of subtle variations in emotions and proved to be a more robust and reliable method for distinguishing the patterns of emotional recognition between both groups. The proposed machine learning framework is thus a strong, data-driven approach to improving diagnostic accuracy and personalizing interventions for ASD and ADHD. This is, therefore, promising in clinical application in the management of these neurodevelopmental conditions on time and effectively.