Time series segmentation of paleoclimate tipping points by an evolutionary algorithm

Research areas:

• Methodology - Machine learning

Year:

2014

Type of Publication:

In Proceedings

Keywords:

Time series segmentation, genetic algorithms, clustering, paleoclimate data, tipping points, abrupt climate change

Authors:

• Pérez-Ortiz, María
• Gutiérrez, Pedro Antonio
• Sánchez-Monedero, Javier
• Hervás-Martínez, César
• Nikolaou, Athanasia
• Dicaire, Isabelle
• Fernandez-Navarro, Francisco

Volume:

8480

Book title:

9th International Conference on Hybrid Artificial Intelligence Systems (HAIS2014)

Series:

Lecture Notes in Computer Science

Pages:

318-329

Organization:

Salamanca (Spain)

Month:

11th-13th September

ISBN:

978-3-319-07616-4

ISSN:

0302-9743

Abstract:

Recent studies propose that some dynamical systems, such as climate, ecological and financial systems, among others, present critical transition points named to as tipping points (TP). Climate TPs can severely affect millions of lives on Earth so that an active scientific community is working on finding early warning signals. This paper deals with the segmentation of a paleoclimate time series to find segments sharing common patterns with the purpose of finding one or more kinds of segments corresponding to TPs. Due to the limitations of classical statistical methods, we propose the use of a genetic algorithm to automatically segment the series together with a method to perform time series segmentation comparisons. Without a priori information, the method clusters together most of the TPs and avoids false positives, which is a promising result given the challenging nature of the problem.