Abstract:
"This study investigated the efficacy of a Feedforward Neural Network (FNN) Multi-Layer
Perceptron (MLP) in predicting dyslexia in children. The MLP model achieved a notable
accuracy of 78%, representing a significant improvement compared to the baseline MLP model
without transfer learning. The utilization of transfer learning was pivotal in enhancing the
predictive capacity of the model. Transfer learning involves leveraging pre-existing knowledge
from broader datasets and adapting it to the specific task at hand. In this case, incorporating
transfer learning methodologies allowed the model to better understand and predict dyslexia
indicators in children. The study also highlights the importance of early detection and
intervention in managing learning disorders such as dyslexia. By accurately predicting dyslexia
at an early age, interventions and support can be initiated promptly, potentially mitigating the
long-term effects of the disorder on a child's academic and personal development. Furthermore,
the integration of a language-independent game framework offers a non-invasive and engaging
means of screening for dyslexia symptoms. By analyzing gameplay data alongside traditional
assessment measures, a comprehensive predictive model is formulated, enhancing the accuracy
and reliability of dyslexia detection. Overall, these findings underscore the potential of neural
network models, particularly when augmented with transfer learning techniques, in facilitating
early detection and intervention for learning disorders in children."
