Machine Learning and Factor-Based Portfolio Optimization

89 Pages Posted: 20 Jul 2021 Last revised: 22 Jul 2021

See all articles by Thomas Conlon

Thomas Conlon

University College Dublin

John Cotter

University College Dublin

Iason Kynigakis

University College Dublin

Date Written: July 8, 2021

Abstract

We examine machine learning and factor-based portfolio optimization. We find that factors based on autoencoder neural networks exhibit a weaker relationship with commonly used characteristic-sorted portfolios than popular dimensionality reduction techniques. Machine learning methods also lead to covariance and portfolio weight structures that diverge from simpler estimators. Minimum-variance portfolios using latent factors derived from autoencoders and sparse methods outperform simpler benchmarks in terms of risk minimization. These effects are amplified for investors with an increased sensitivity to risk-adjusted returns, during high volatility periods or when accounting for tail risk.

Keywords: Autoencoder, covariance matrix, dimensionality reduction, factor models, machine learning, minimum-variance, principal component analysis, Partial least squares, portfolio optimization, sparse principal component analysis, sparse partial least squares

JEL Classification: C38, C4, C45, C5, C58, G1, G11

Suggested Citation

Conlon, Thomas and Cotter, John and Kynigakis, Iason, Machine Learning and Factor-Based Portfolio Optimization (July 8, 2021). Michael J. Brennan Irish Finance Working Paper Series Research Paper No. 21-6, Available at SSRN: https://ssrn.com/abstract=3889459 or http://dx.doi.org/10.2139/ssrn.3889459

Thomas Conlon

University College Dublin ( email )

Smurfit Graduate Business School
Blackrock
Co. Dublin, n/a
Ireland

HOME PAGE: http://www.ucd.ie/bankingfinance/staff/drthomasconlon/

John Cotter

University College Dublin ( email )

School of Business, Carysfort Avenue
Blackrock, Co. Dublin
Ireland
353 1 716 8900 (Phone)
353 1 283 5482 (Fax)

HOME PAGE: http://https://johncotter.org/

Iason Kynigakis (Contact Author)

University College Dublin ( email )

Blackrock, Co. Dublin
Ireland

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

Paper statistics

Downloads
1,191
Abstract Views
3,246
Rank
32,284
PlumX Metrics