Early identification of learning and developmental difficulties in children can help teachers and parents provide suitable support at the right stage. However, identifying conditions such as dyslexia, dysgraphia, ADHD, and autism can be difficult because children may show different academic, behavioural, and learning patterns. This paper presents Smart Diagnosis, an AI-assisted web application developed to support the early screening of these difficulties in children aged 6–10 years. The system brings together real research datasets covering handwriting and structured screening information and uses machine learning techniques suitable for different data types. For the experimental evaluation, HOG features with an RBF-based Support Vector Machine were used for handwriting-based dyslexia and dysgraphia datasets, while Logistic Regression and Random Forest were evaluated on the autism screening dataset. The evaluated models achieved accuracies of 78.36% for dyslexia, 88.00% for dysgraphia, and 98.31% for autism on the respective test sets. The system also provides separate dashboards for teachers, parents, and students, along with assessment support, handwriting evidence where available, personalised learning recommendations, progress information, and an AI-based educational assistant. Sensitive screening information is not exposed through the student dashboard. The proposed platform therefore combines data-driven screening support with educational assistance in a single application. The results are specific to the datasets and test splits used in this study and are not intended to represent clinical diagnostic accuracy. Further validation using larger and more diverse datasets is required before practical deployment.

earning screening, dyslexia, dysgraphia, ADHD, autism spectrum disorder, machine learning, handwriting analysis, educational technology