Prediction of Peak Ground Acceleration (PGA) in Java Using Artificial Neural Network Method

Abstract

Java is one of the islands in Indonesia that frequently experiences earthquakes. Earthquakes can cause significant ground motion that can damage buildings and threaten human life. Peak Ground Acceleration (PGA) is a measure of the maximum ground acceleration that occurs during an earthquake and is an important factor that must be considered at every construction site to assess the potential damage that can be caused by an earthquake. The parameters considered in determining PGA predictions are earthquake parameters, such as magnitude and hypocenter distance. In addition, the PGA value is also influenced by local site conditions. With advances in information technology and artificial intelligence, especially in the development of Artificial Neural Networks (ANN), research on PGA prediction needs to be conducted as one of the efforts in reducing the risk of earthquakes. The purpose of this research is to obtain the best network architecture in predicting PGA values. The criteria for selecting the best network architecture is done by comparing the error value of each possible architecture formed. The best prediction results are obtained in the model with 3-15-1 architecture with a correlation value of 0.67.