Implementation of the Neural Network Algorithm in Monitoring Child Development to Screen for Developmental Disorders at an Early Age

Abstract

This research aims to implement a Neural Network (NN) in monitoring children's development, especially to detect developmental disorders from an early age. The data used includes variables such as Age, Height, and Weight, which have been normalized to have a uniform scale. The modeling process begins with the use of Convolutional Layers to extract important features from numerical data, which are then passed to the ReLU activation layer to introduce non-linearity to the model, enabling the detection of more complex patterns. After that, Max Pooling is carried out to reduce data dimensions and increase computing efficiency. This model was trained using 100 normalized data, and continued with the use of fully connected layers to process further information. In the output layer, a sigmoid activation function is used to generate probability predictions, allowing binary classification (whether a developmental disorder is present or not). Evaluation results show that this model has an accuracy of 85%, which indicates its effectiveness in detecting child developmental disorders based on available data. Although the results are promising, there is still room for improvement, especially in improving the model's accuracy and ability to handle more complex data. Overall, this research shows that Neural Networks can be a useful tool in the early detection of childhood developmental disorders, with potential for broad applications in the fields of children's health and education.