DysDetectV1: Dyslexia-dysgraphia severity classification using handwriting image and machine learning model

The existence of dyslexia-dysgraphia symptoms is assessed by a specialist or evaluator through observation of the candidate using prescribed techniques and a scoring system. Nonetheless, this methodology presents some limitations, including time consumption, the potential for activities to be incomplete, and ultimately, the risk of yielding biased outcomes. The growth of technology has significantly evolved ways for automatic dysgraphia detection, leading to a diversity of handwritingbased methodologies. Nevertheless, limited research highlights the detection of handwritten images using a handwritten image collection, particularly in the development of machine learning models. Therefore, this study introduces DysDetectV1, which is an innovative approach to classifying the severity of dyslexia-dysgraphia using handwritten samples. It uses Residual network feature extraction and a bagged tree model to classify the severity of dyslexia-dysgraphia levels. The model was then evaluated using a multi-class confusion matrix and represented accuracy, precision, and recall measurements. This research has the potential to enhance handwriting classification for individuals with dyslexia-dysgraphia by utilizing handwritten images and a machine learning model. This comprehensive approach is crucial for developing reliable tools that facilitate the early identification and support of children experiencing learning difficulties, in line with the overarching goal of improving educational outcomes. This research aligns with the primary objective of enhancing handwritten recognition for educational purposes, thereby fostering early childhood development in relation to SDG 4 (Quality Education).