A Machine-Learning-Based Early Warning System Boosted by Topological Data Analysis

23 Pages Posted: 12 Jun 2019

See all articles by Devraj Basu

Devraj Basu

University of Strathclyde - Department of Accounting and Finance

Tieqiang Li

affiliation not provided to SSRN

Date Written: May 27, 2019

Abstract

We propose a novel early warning system for detecting financial market crashes that utilizes the information extracted from the shape of financial market movement. Our system incorporates Topological Data Analysis (TDA), a new set of data analytics techniques specialised in profiling the shape of data, into a more traditional machine learning framework. Incorporating TDA leads to substantial improvements in timely detecting the onset of a sharp market decline. Our framework is both able to generate new features and also unlock more value from existing factors. Our results illustrate the importance of understanding the shape of financial market data and suggest that incorporating TDA into a machine learning framework could be beneficial in a number of financial market settings.

Keywords: financial crashes, equity markets, machine learning, Topological Data Analysis

JEL Classification: C02, C18, C45, C58, G01

Suggested Citation

Basu, Devraj and Li, Tieqiang, A Machine-Learning-Based Early Warning System Boosted by Topological Data Analysis (May 27, 2019). Available at SSRN: https://ssrn.com/abstract=3394704 or http://dx.doi.org/10.2139/ssrn.3394704

Devraj Basu (Contact Author)

University of Strathclyde - Department of Accounting and Finance ( email )

Curran Building
100 Cathedral Street
Glasgow G4 0LN
United Kingdom

Tieqiang Li

affiliation not provided to SSRN

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
322
Abstract Views
1,337
Rank
168,707
PlumX Metrics