Ensemble Learning Applied to Quant Equity: Gradient Boosting in a Multifactor Framework

Tony Guida and Guillaume Coqueret ()
Additional contact information
Tony Guida: EDHEC - EDHEC Business School - UCL - Université catholique de Lille
Guillaume Coqueret: Groupe Sup de Co Montpellier (GSCM) - Montpellier Business School

Post-Print from HAL

Abstract: This chapter proposes to benefit from the advantages of machine learning (ML) in general and boosted trees in particular, e.g. non‐linearity, regularization and good generalization results, scaling up well with lots of data. It gives a mildly technical introduction to boosted trees. The chapter introduces the construction of the dataset with the feature and labels engineering, and the calibration of the ML applying rigorous protocol established by the computer science community. It describes the data used and the empirical protocol for the ML model. The chapter also introduces the concept of confusion matrix and all the related metrics in order to precisely assess a ML model's quality. It provides guidance on how to tune, train and test an ML‐based model using traditional financial characteristics such as valuation and profitability metrics, but also price momentum, risk estimates, volume and liquidity characteristic.

Date: 2019-01-01
Note: View the original document on HAL open archive server: https://hal.science/hal-02311104v1
References: View references in EconPapers View complete reference list from CitEc
Citations:

Published in Big Data and Machine Learning in Quantitative Investment, John Wiley & Sons, 129-148 p., 2019, 9781119522195. ⟨10.1002/9781119522225.ch7⟩

Downloads: (external link)
https://hal.science/hal-02311104v1/document (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:hal:journl:hal-02311104

DOI: 10.1002/9781119522225.ch7

Access Statistics for this paper

More papers in Post-Print from HAL
Bibliographic data for series maintained by CCSD ().