Using Machine Learning to Forecast Future Earnings

Earnings prediction has always been an important subject in accounting research given the proven relationship between accurate earnings prediction and excess investment return (Beaver, Journal of Accounting Research, 1968). Apart from the development of accounting and fnance subjects, advances in st...

Full description

Saved in:
Bibliographic Details
Published inAtlantic economic journal Vol. 48; no. 4; pp. 543 - 545
Main Authors Xinyue, Cui, Zhaoyu, Xu, Yue, Zhou
Format Journal Article
LanguageEnglish
Published New York Springer US 01.12.2020
Springer Nature B.V
Subjects
Accounting
Anthology
Companies
Computer science
Digital currencies
Earnings
Economics
Economics and Finance
Finance
International Economics
Internet
Machine learning
Macroeconomics
Macroeconomics/Monetary Economics//Financial Economics
Microeconomics
Predictions
Prices
Public Finance
Securities markets
Statistics
G10
Online AccessGet full text

Cover

More Information
Summary:Earnings prediction has always been an important subject in accounting research given the proven relationship between accurate earnings prediction and excess investment return (Beaver, Journal of Accounting Research, 1968). Apart from the development of accounting and fnance subjects, advances in statistics and computer science have also contributed to advances in earnings prediction methods. Numerous studies have highlighted the surprising potential of machine learning models in earnings prediction. This paper is aimed to recommend an applicable approach for earnings prediction with a state-of-the-art machine learning model, LightGBM (Online Supplemental Appendix Table 1), which has shown noteworthy efciency in other prediction tasks such as cryptocurrency pricing (Sun et al., Finance Research Letters, 2018), but has not been extensively studied for earnings prediction. In this paper, the model was constructed using LightGBM to predict accounting earnings growth with fnancial, macroeconomic, and market variables. The samples were selected from the quarterly reports of 3,000 companies with the highest market capitalization in the U.S. equity market from 1988 to 2018, eliminating companies with share prices below $1, companies from the utility and fnance sectors and companies whose fscal year-end changed or was not March, June, September, or December
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0197-4254
1573-9678
DOI:10.1007/s11293-020-09691-1