Leveraging of recurrent neural networks architectures and SMOTE for dyslexia prediction optimization in children

Abstract

Learning disorders in children (dyslexia) have become a severe problem that needs attention. If this is not immediately detected and treated early on, bad habits due to dyslexia will carry over into adulthood. Many studies have been carried out on developing analytical methods for detecting/predicting dyslexia, both conventionally and based on machine learning. However, many prediction systems that have been proposed previously still do not focus on solving the problem of inequality in the data classes of people with dyslexia and ordinary people who are used in the training and testing process. Therefore, we are trying to build a system using RNN architectures that can quickly and accurately predict the possibility of a child having dyslexia. To overcome the data imbalance between dyslexics and non-dyslexics, we also apply the SMOTE method to the dataset. SMOTE will synthesize dyslexic data to balance the numbers with non-dyslexic data. This study used a dataset of 3640 participants (392 dyslexic and 3248 non-dyslexics). For the process of predicting dyslexia, several algorithms such as Simple RNN, LSTM, and GRU are used. As a result, there is an increase in prediction accuracy when SMOTE is applied (compared to without SMOTE) in the dyslexia forecasting process using RNN (92.68% for training and 91.16% for testing), LSTM (94.81% for training and 93.16% for testing), and GRU (96.43 % for training and 92.24% for testing). Using SMOTE+RNN architecture in this research increased the accuracy of dyslexia prediction by up to 5% compared to without SMOTE.