By Kaelynn Rose
The objective of this project was to design and compare two types of deep learning models for earthquake detection and prediction of earthquake characteristics, using seismic signals from the STanford EArthquake Dataset (STEAD). I trained and tested CNN and LSTM classification and regression models to predict whether seismic signals were earthquakes or noise, and predict earthquake magnitude, p-wave arrival time, and s-wave arrival time. The best performing classification model was a CNN model, which I containerized and deployed using an AWS Lambda function. I then wrote a script to fetch live data from a seismometer on Kilauea volcano in Hawaii which produces many earthquakes per day, and ran the model on the live stream of data using Lambda to predict whether seismic signals were earthquakes or noise in near-real-time.
LinkedIn: [ Ссылка ]
Project Repo: [ Ссылка ]
