Lowering the Gradient Error on Neural Network Using Backpropagation for Diagnosing Psychological Problems among Children

Abstract

Children’s problems are sometimes little known to parents. To be able to recognize problems that occur in children, their parents will consult with child psychology. However, not all parents have spare time to consult. There is now also an application-based expert system to solve problems. Unfortunately, the use of expert systems with various methods has been too many and few are built using Neural Networks. Neural Networks is one application of artificial intelligence that is able to solve unstructured and complicated problems, and able to process input data without having to have a target. One model is the BackPropagation network. This network model is then used to identify child psychological problems. By taking some of the symptoms generated, then set up some network architecture and followed by conducting training using training data and testing using testing data on each network architecture that is formed. From the results of the training process and testing the architecture, then obtained the best network that has the lowest gradient error. This result is obtained by increasing the number of hidden layer on the built network.