Akademik Veri Yönetim Sistemi
NATURAL HAZARDS, 2025 (SCI-Expanded)
Predicting earthquake risk areas and risk levels is vital in minimizing the loss of life. In this study, earthquake risk assessment has been conducted by producing predictions for both five-class and two-class risk levels. The methods were tested on Izmir province. For this purpose, the city was divided into 28 zones. Twenty-two different evaluation criteria were assessed using geographic information systems. Risky areas were predicted using Support Vector Machines, k-Nearest Neighbors, Naive Bayes, Decision Trees, and Ensemble classifiers. It has been concluded that the F1 score results, the highest prediction success in training is ensemble classifier with 96%, and tests is decision tree methods with 45% for five classes. In addition, the training results is the ensemble classifier with 98%, and the test results is the decision tree methods with 76% for two classes. When all machine learning results were examined together, test prediction success on data labeled with two-classes was found to be significantly more successful than on data labeled with five classes. As a result of this study, it has been observed that Multi-Criteria Decision Making and machine learning give significant results in the area-based earthquake vulnerability analysis performed together. In addition, this study provides a practical contribution to urban planning and the improvement of development strategies in & Idot;zmir by identifying high-risk areas to mitigate seismic risks. Furthermore, the findings offer a data-driven framework for enhancing disaster management policies, enabling authorities to effectively plan emergency responses in vulnerable regions, implement appropriate construction techniques in high-risk areas, and optimize resource allocation.