Machine Learning Approach for Classifying Natural and Anthropogenic Seismic Events

Authors: Saeid Zarifikoliaee and Mark Grmek

This repository provides code and resources for training and evaluating two machine learning models—a Convolutional Neural Network (CNN) and a PCA+SVM model—designed to classify seismic events according to their origin, distinguishing between anthropogenic and natural sources.

All the methods and resources used for this project are described in detail in the Project Report

Environment Setup

It is highly recommended to use Conda to manage the Python environment and dependencies. To create the environment from the provided YAML file:

conda env create -f environment.yml
conda activate project_c_env

Data Preparation

• The training data is obtained using the script waveform_fetch.py
• The original seismic event dataset is filtered using filter_events.py (only needs to be run once)

Model Training

• CNN Model: Training and evaluation are done in CNN.ipynb
• PCA+SVM Model: Training and evaluation are done in PCA_SVM.ipynb