Innovative Data Science Approaches to Automatic Aftershock Detection

Abstract

The aim of this contract was to explore innovative data mining techniques for automatic aftershock detection in order to intelligently monitor for nuclear explosions. In a nuclear explosion monitoring system, a burst of aftershocks (small earthquakes that occur for days to years following a large earthquake) is uninteresting and could be mislabeled as the target events. Such a burst of uninteresting events overloads the human analysts of the monitoring system. To reduce the load, at the onset of a sequence of events (e.g., aftershocks), a human analyst can label a few of these events and start an online classifier to filter out subsequent uninteresting events. In this project, we develop an online classification framework for aftershocks. Our specific technique uses a learned model to classify an event as an aftershock and exploits another classifier to decide if the model should be re-trained with the newly detected events. The framework has been tested on two large events: the 7.8Mw earthquake in Gorkha, Nepal on April 2015 and the 8.2Mw earthquake in Chiapas, Mexico on September 2017. We measure that a tolerable false-positive rate of 5 percent will allow us to automatically remove 83 percent of the Nepal aftershocks and 90 percent of the Chiapas aftershocks using the data from only one three-component station. This result shows that the proposed framework can save enough human effort to outweigh the effort needed to carefully judge the false positives. We demonstrate that our method, named FewSig, learns better than existing data science methods from only a few confirmed aftershock instances. We demonstrate that the performance of the framework improves with more confirmed instances and with better-placed stations.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 28, 2022
Accession Number
AD1195980

Entities

People

  • Abdullah Mueen
  • Shen Zhong

Organizations

  • Department of Computer Science, University of Oxford

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Artificial Intelligence
  • Artificial Intelligence Software
  • Automata Theory
  • Computer Science
  • Data Analysis
  • Data Mining
  • Data Preprocessing
  • Data Science
  • Databases
  • Deep Learning
  • Detection
  • Dimensionality Reduction
  • Information Processing
  • Information Science
  • Machine Learning
  • Network Science
  • Semi-Supervised Learning
  • Supervised Machine Learning

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Neural Networks