Técnicas de predicción de terremotos usando machine learning: Una revisión sistemática de la literatura

Abstract

Technological development has awakened interest on the part of the scientific community in researching to predict earthquakes. The article's objective is to know what variables, techniques, tools, and methodologies have been used in the different studies to predict earthquakes using machine learning techniques. To carry out the study, the Kitchenham methodology was used, which consists of three development phases: review planning, conducting, and reporting. In the planning phase, four research questions were posed; for this purpose, an exhaustive literature search was carried out. After carrying out the selection and exclusion criteria, the questions posed were developed, of which it was found that 15% of the variables to predict earthquakes were latitude, longitude, and depth. In comparison, 13% were the seismic magnitude. 17% of the most used techniques were Random Forest, followed by Artificial Neural Networks with 17%. 65% used Python to develop algorithms, followed by MATLAB and R at 14%. 50% implemented the CRISP-DM methodology for data mining projects, followed by KDD with 33%.