A Comparison of Forecasting Approaches for Capital Markets

Scott McDonald, SA Coleman, TM McGinnity, Yuhua Li, Ammar Belatreche

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

In recent years, machine learning algorithms havebecome increasingly popular in financial forecasting. Theirflexible, data-driven nature makes them ideal candidates fordealing with complex financial data. This paper investigates theeffectiveness of a number of machine learning algorithms, andcombinations of these algorithms, at generating one-step aheadforecasts of a number of financial time series. We find thathybrid models consisting of a linear statistical model and a nonlinearmachine learning algorithm are effective at forecastingfuture values of the series, particularly in terms of the futuredirection of the series.
LanguageEnglish
Title of host publicationUnknown Host Publication
Pages32-39
Number of pages8
Publication statusPublished - 27 Mar 2014
EventIEEE Computational Intelligence for Financial Engineering and Economics -
Duration: 27 Mar 2014 → …

Conference

ConferenceIEEE Computational Intelligence for Financial Engineering and Economics
Period27/03/14 → …

Fingerprint

Learning algorithms
Learning systems
Time series
Financial markets
Statistical Models

Cite this

McDonald, S., Coleman, SA., McGinnity, TM., Li, Y., & Belatreche, A. (2014). A Comparison of Forecasting Approaches for Capital Markets. In Unknown Host Publication (pp. 32-39)
McDonald, Scott ; Coleman, SA ; McGinnity, TM ; Li, Yuhua ; Belatreche, Ammar. / A Comparison of Forecasting Approaches for Capital Markets. Unknown Host Publication. 2014. pp. 32-39
@inproceedings{1f4babe462c349fbb56e1886a7baa244,
title = "A Comparison of Forecasting Approaches for Capital Markets",
abstract = "In recent years, machine learning algorithms havebecome increasingly popular in financial forecasting. Theirflexible, data-driven nature makes them ideal candidates fordealing with complex financial data. This paper investigates theeffectiveness of a number of machine learning algorithms, andcombinations of these algorithms, at generating one-step aheadforecasts of a number of financial time series. We find thathybrid models consisting of a linear statistical model and a nonlinearmachine learning algorithm are effective at forecastingfuture values of the series, particularly in terms of the futuredirection of the series.",
author = "Scott McDonald and SA Coleman and TM McGinnity and Yuhua Li and Ammar Belatreche",
year = "2014",
month = "3",
day = "27",
language = "English",
pages = "32--39",
booktitle = "Unknown Host Publication",

}

McDonald, S, Coleman, SA, McGinnity, TM, Li, Y & Belatreche, A 2014, A Comparison of Forecasting Approaches for Capital Markets. in Unknown Host Publication. pp. 32-39, IEEE Computational Intelligence for Financial Engineering and Economics, 27/03/14.

A Comparison of Forecasting Approaches for Capital Markets. / McDonald, Scott; Coleman, SA; McGinnity, TM; Li, Yuhua; Belatreche, Ammar.

Unknown Host Publication. 2014. p. 32-39.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - A Comparison of Forecasting Approaches for Capital Markets

AU - McDonald, Scott

AU - Coleman, SA

AU - McGinnity, TM

AU - Li, Yuhua

AU - Belatreche, Ammar

PY - 2014/3/27

Y1 - 2014/3/27

N2 - In recent years, machine learning algorithms havebecome increasingly popular in financial forecasting. Theirflexible, data-driven nature makes them ideal candidates fordealing with complex financial data. This paper investigates theeffectiveness of a number of machine learning algorithms, andcombinations of these algorithms, at generating one-step aheadforecasts of a number of financial time series. We find thathybrid models consisting of a linear statistical model and a nonlinearmachine learning algorithm are effective at forecastingfuture values of the series, particularly in terms of the futuredirection of the series.

AB - In recent years, machine learning algorithms havebecome increasingly popular in financial forecasting. Theirflexible, data-driven nature makes them ideal candidates fordealing with complex financial data. This paper investigates theeffectiveness of a number of machine learning algorithms, andcombinations of these algorithms, at generating one-step aheadforecasts of a number of financial time series. We find thathybrid models consisting of a linear statistical model and a nonlinearmachine learning algorithm are effective at forecastingfuture values of the series, particularly in terms of the futuredirection of the series.

M3 - Conference contribution

SP - 32

EP - 39

BT - Unknown Host Publication

ER -

McDonald S, Coleman SA, McGinnity TM, Li Y, Belatreche A. A Comparison of Forecasting Approaches for Capital Markets. In Unknown Host Publication. 2014. p. 32-39