Artificial intelligence for assessing the planets' positions as a precursor to earthquake events

Abstract

Questions about interconnection possibilities between planets’ positions and seismic events on the earth have emerged recently in TV channels, social media, etc. In this study, an Artificial Neural Network (ANN) and Random Forest Regression (RFR) are used to predict the number of earthquakes that can occur on Earth, depending on the Earth's position relative to other planets and solar positions. Our new integration dataset contains 9809 observations and nine features firstly from the global earthquake archive, which is an authoritative layer by Esri, and secondly from the accurate data web portal “theskylive.com.”. The results obtained from RFR and ANN prove the partial influence of planets positions on sesimic activity on the earth. In other words, quantitatively through the ANN that gets an accuracy of 68.27 %, MAE of 5.36, MSE of 52.78, RMSE of 7.26, R-Squared of 0.65, and also through the RFR that gets an accuracy of 65.06 %, MAE of 5.60, MSE of 58.21, RMSE of 7.63, R-Squared of 0.67, prove the partial influence on one hand. Qualitatively through the curve of the training phase of the ANN, which is a decreasing and convex function, reinforces the aforementioned proof on the other hand. For these reasons, it can be deduced that there is a possible connection between tectonic stress triggers and the positions of the planets in the solar system. Our dataset was uploaded to the github(https://github.com/mouddentarik/Earthquake01.) as well as the code will be publicly available at the github(https://github.com/mouddentarik/PythonCode_Earthquakes-.) to share our results.