Earthquake magnitude prediction with advanced machine learning models

Earthquakes are an integral part of life, yet they are natural disasters that significantly impact human life. This thesis aims to contribute to the understanding of this crucial subject through a comprehensive evaluation of earthquakes from geological, seismological, and engineering perspectives. By assessing the effects of plate tectonics, volcanic activities, and human-induced triggers on earthquakes, the hypotheses formed analyze the process of earthquake occurrence and their risk profiles. The strategic importance of determining risk profiles for geographic regions based on their earthquake potential is emphasized, with a focus on earthquake zones and hazard analyses. Based on their performance in estimating earthquake magnitude, machine learning methods were comparatively evaluated. The dataset used includes earthquakes that have occurred since 1900, and various machine learning models have been applied to this dataset. These models include Random Forest, Gradient Boosting, XGBoost, Linear Regression, Ridge Regression, Lasso Regression, ElasticNet Regression, KNeighbors Regression, AdaBoost Regression, Decision Tree Regressor, Bagging Regressor ve Support Vector Regression. The performances of these models were compared with various metrics and the most effective model was determined. The results demonstrate that machine learning models hold significant potential for predicting earthquake magnitude. This study aims to evaluate the role of exploratory data analysis using artificial intelligence in earthquake risk analysis and prediction, thereby contributing to society's earthquake preparedness efforts. By providing a multifaceted observation of the behavior of machine learning models related to earthquakes, this work makes a substantial contribution to the academic literature. Keywords: Earthquake, Machine Learning, Risk Analysis, Artificial Intelligence