Ground Acceleration Clustering Using Self-Organizing Map Method

Peak Ground Acceleration (PGA) is an important parameter in seismic studies because it is directly related to the level of shaking felt on the earth's surface. Analysis of ground acceleration data is needed to identify patterns, group regions based on their seismic characteristics, and support earthquake disaster mitigation efforts. This study uses the Self-Organizing Map (SOM) method, which is an unsupervised learning approach based on artificial neural networks that can map high-dimensional data into a two-dimensional map representation without losing its topological structure. The ground acceleration dataset used in this study consists of key seismic parameters such as depth, magnitude, source distance, and PGA values. The SOM learning process is carried out iteratively to produce a cluster map that groups earthquake data into several groups with different ground acceleration characteristics. The results show that the SOM method is able to identify ground acceleration distribution patterns more clearly than conventional methods, by producing clusters that represent variations in PGA from low to high. These findings can provide important contributions to earthquake risk mapping, regional spatial planning, and the formulation of more accurate disaster mitigation strategies.