On Using Conventional Machine Learning for Detecting Microseisms at Llaima Volcano

Volcanic eruptions represent a formidable geological force capable of causing widespread devastation and the loss of human lives. Institutions in charge of monitoring volcanoes want to provide timely and invaluable information on volcanic activity, ultimately safeguarding lives in the face of imminent volcanic disasters. To do that, the analysis and interpretation of large volumes of data collected by monitoring efforts pose formidable challenges. The process is complex, labor intensive, and often subject to human bias. To address these issues, this research introduces an intelligent algorithm rooted in traditional Machine Learning to detect microseisms. This innovative approach aims to streamline the detection of microseismic events that occur within 20-min data records, using a comprehensive database comprising 3592 meticulously recorded microseismic events from the LAV station at the Llaima volcano. The results obtained from this effort are truly remarkable. We determined Decision Tree presents the best results with an Accuracy of 99.6%, and a Balanced Error Rate of 0.006 in the test phase. These results underline the transformative potential of this innovative approach, which represents a significant step towards more efficient and reliable microseismic detection.