A predictive model for chinese children with developmental dyslexia—Based on a genetic algorithm optimized back-propagation neural network

Abstract

The identification or the diagnosis of developmental dyslexia has long been a difficult issue, and traditional logistic regression predictive models have some defects. This study established a genetic algorithm optimized back-propagation neural network model to predict whether Chinese children have dyslexia based on data from 399 children (187 children with dyslexia and 212 typically developing children, 3rd–6th graders, aged 7–13 years). The model achieved an overall prediction accuracy of approximately 94%. Moreover, reading accuracy was the strongest factor in predicting Chinese dyslexic children, and phonological awareness, the accuracy rate of pseudocharacters, morphological awareness, reading fluency, rapid digit naming, and the reaction times of noncharacters also made important contributions to the prediction. In summary, the model we established in this study had an excellent predictive capability regarding Chinese children with/without developmental dyslexia. Furthermore, the genetic algorithm optimized back-propagation neural network model that substantially improves the prediction accuracy of Chinese dyslexia, has the potential to direct more targeted prevention and treatment strategies, and lay the foundation for the artificial intelligence expert diagnosis system for Chinese dyslexia.