Comparison of Different Heuristics Integrated with Neural Networks: A Case Study for Earthquake Damage Estimation

Abstract

Earthquakes are among the most challenging natural phenomena to predict. Most of these unpredictable earthquakes result in the loss of human lives and property. Seismologists can estimate the probable location and magnitude of such earthquakes. However, the actual time and extent of their impact remain unknown. If the effects of possible earthquakes can be predicted, quick and accurate decisions can be made. For this purpose, developing predictive models about earthquakes is a prevalent and vital issue in the literature. In this study, various Machine Learning (ML) algorithms were compared on a public dataset of earthquakes, which had occurred worldwide and had a local magnitude Ml >= 3, and the algorithm with the highest performance was selected and optimized with various other algorithms. The performances of the models were compared using different performance evaluation metrics such as accuracy, Mean Square Error, Root-Mean Square Error, precision, recall, and f1 score. As a result, it was observed that the Artificial Neural Network (ANN) algorithm optimized with the Particle Swarm Optimization (PSO) algorithm produced the most successful result with an accuracy value of 0.82. Based on the obtained results, it is believed that this model can be used in different earthquake damage prediction studies and as a guide in emergency planning