A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh
The motivation behind the present study is to determine whether stock indices futures possess long-term memory and entropy, and whether they follow the adaptive market hypothesis (AMH). Another motivation is to specify whether innovatively combined new methods such as wavelets, principal component a...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Published: |
2018
|
Subjects: | |
Online Access: | http://studentsrepo.um.edu.my/14536/1/Mohammadali_Mehralizadeh.pdf http://studentsrepo.um.edu.my/14536/ |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
id |
my.um.stud.14536 |
---|---|
record_format |
eprints |
spelling |
my.um.stud.145362023-06-26T18:37:41Z A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh Mohammadali , Mehralizadeh HG Finance The motivation behind the present study is to determine whether stock indices futures possess long-term memory and entropy, and whether they follow the adaptive market hypothesis (AMH). Another motivation is to specify whether innovatively combined new methods such as wavelets, principal component analysis (PCA), and artificial neural networks (ANNs) produce returns in excess of the threshold buy-andhold strategy advocated by the random walk hypothesis (RWH). In addition, this study tries to determine if a novel hybrid forecasting model produces higher returns than traditional and intelligent technical trading strategies, such as the pure ANN forecasting method and combination of wavelets and ANN, in the current increasingly difficult and volatile futures markets. In accordance with the literature review and the limitations of ANNs in the context of noisy time series, this study proposes a hybrid ANN with wavelet transforms, which is a special tool for denoising signal, namely wavelet neural network (WNN). Moreover, in accordance with an enhancement on denoising process with wavelet transforms using principal component analysis, a novel hybrid intelligent system, namely wavelet principal component analysis neural network (WPCA-NN) model, is developed in order to denoise financial time series carefully and forecast them with greater accuracy. The study shows evidence drawn from 2005 to 2014 of the predictive ability and profitability of the WPCA-NN model regarding the contracts of Hong Kong’s Hang Seng futures, Japan’s NIKKEI 225 futures, Malaysia’s Kuala Lumpur Composite Index (KLCI) futures, Singapore’s Morgan Stanley Capital International (MSCI) futures, the Korea Composite Stock Price Index 200 (KOSPI 200) futures, the Standard & Poor’s 500 (S&P 500), and Taiwan’s Stock Exchange TAIEX futures markets. Moreover, this study employs many technical analysis indicators, consisting of the relative strength index (RSI), moving average convergence/divergence (MACD), MACD signal, stochastic fast %K, stochastic slow %K, stochastic %D. In addition, this study uses three years of historical data to train a network in order to forecast an incoming quarter, which we call the evaluation period. Because the investment trading community and fund managers evaluate their portfolio performances quarterly, and change their trading parameters for the next quarter, this study is conducted on a quarterly basis. There is also a quarterly evaluation for 10 years, from 2005 to 2014, in order to measure trading performance and trading profitability. The advantage of the sustainable returns of the WPCA-NN model to threshold buyand- hold strategy and other forecasting methods, would offer existence of the inefficiency in selected markets and will give suitable tools for weary traders to forecast financial markets. In addition, this superiority would support the adaptive market hypothesis (AMH). Moreover, the offered method of denoising can be considered an enhancement to the univariate wavelet denoising, not only in financial domain, but also in other fields of study. 2018-11 Thesis NonPeerReviewed application/pdf http://studentsrepo.um.edu.my/14536/1/Mohammadali_Mehralizadeh.pdf Mohammadali , Mehralizadeh (2018) A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh. PhD thesis, Universiti Malaya. http://studentsrepo.um.edu.my/14536/ |
institution |
Universiti Malaya |
building |
UM Library |
collection |
Institutional Repository |
continent |
Asia |
country |
Malaysia |
content_provider |
Universiti Malaya |
content_source |
UM Student Repository |
url_provider |
http://studentsrepo.um.edu.my/ |
topic |
HG Finance |
spellingShingle |
HG Finance Mohammadali , Mehralizadeh A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh |
description |
The motivation behind the present study is to determine whether stock indices futures possess long-term memory and entropy, and whether they follow the adaptive market hypothesis (AMH). Another motivation is to specify whether innovatively combined new methods such as wavelets, principal component analysis (PCA), and artificial neural networks (ANNs) produce returns in excess of the threshold buy-andhold
strategy advocated by the random walk hypothesis (RWH). In addition, this study
tries to determine if a novel hybrid forecasting model produces higher returns than
traditional and intelligent technical trading strategies, such as the pure ANN
forecasting method and combination of wavelets and ANN, in the current increasingly
difficult and volatile futures markets. In accordance with the literature review and the
limitations of ANNs in the context of noisy time series, this study proposes a hybrid
ANN with wavelet transforms, which is a special tool for denoising signal, namely
wavelet neural network (WNN). Moreover, in accordance with an enhancement on
denoising process with wavelet transforms using principal component analysis, a novel
hybrid intelligent system, namely wavelet principal component analysis neural
network (WPCA-NN) model, is developed in order to denoise financial time series
carefully and forecast them with greater accuracy.
The study shows evidence drawn from 2005 to 2014 of the predictive ability and
profitability of the WPCA-NN model regarding the contracts of Hong Kong’s Hang
Seng futures, Japan’s NIKKEI 225 futures, Malaysia’s Kuala Lumpur Composite
Index (KLCI) futures, Singapore’s Morgan Stanley Capital International (MSCI)
futures, the Korea Composite Stock Price Index 200 (KOSPI 200) futures, the
Standard & Poor’s 500 (S&P 500), and Taiwan’s Stock Exchange TAIEX futures
markets. Moreover, this study employs many technical analysis indicators, consisting of the relative strength index (RSI), moving average convergence/divergence
(MACD), MACD signal, stochastic fast %K, stochastic slow %K, stochastic %D. In
addition, this study uses three years of historical data to train a network in order to
forecast an incoming quarter, which we call the evaluation period. Because the
investment trading community and fund managers evaluate their portfolio
performances quarterly, and change their trading parameters for the next quarter, this
study is conducted on a quarterly basis. There is also a quarterly evaluation for 10
years, from 2005 to 2014, in order to measure trading performance and trading
profitability.
The advantage of the sustainable returns of the WPCA-NN model to threshold buyand-
hold strategy and other forecasting methods, would offer existence of the
inefficiency in selected markets and will give suitable tools for weary traders to
forecast financial markets. In addition, this superiority would support the adaptive
market hypothesis (AMH). Moreover, the offered method of denoising can be
considered an enhancement to the univariate wavelet denoising, not only in financial
domain, but also in other fields of study.
|
format |
Thesis |
author |
Mohammadali , Mehralizadeh |
author_facet |
Mohammadali , Mehralizadeh |
author_sort |
Mohammadali , Mehralizadeh |
title |
A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh |
title_short |
A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh |
title_full |
A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh |
title_fullStr |
A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh |
title_full_unstemmed |
A new hybrid forecasting model using wavelet-PCA and artificial neural network for futures markets / Mohammadali Mehralizadeh |
title_sort |
new hybrid forecasting model using wavelet-pca and artificial neural network for futures markets / mohammadali mehralizadeh |
publishDate |
2018 |
url |
http://studentsrepo.um.edu.my/14536/1/Mohammadali_Mehralizadeh.pdf http://studentsrepo.um.edu.my/14536/ |
_version_ |
1770551697109680128 |
score |
13.18916 |