Predictability of gold price using a modified EMD-based artificial neural network model
碩士 === 國立臺北大學 === 統計學系 === 105 === The demand for gold has been increasing since 2004 due to rising oil price, subprime mortgage crisis, financial crisis of 2007-2008, European debt crisis, Brexit announced by UK officially, a series of policy from the U.S. president Donald Trump, and the Federal Re...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2017
|
Online Access: | http://ndltd.ncl.edu.tw/handle/34957909391508410280 |
id |
ndltd-TW-105NTPU0337026 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-105NTPU03370262017-09-01T04:30:02Z http://ndltd.ncl.edu.tw/handle/34957909391508410280 Predictability of gold price using a modified EMD-based artificial neural network model 利用經驗模態分解法的類神經網路模型對黃金價格之預測 LIN, FANG-YI 林芳儀 碩士 國立臺北大學 統計學系 105 The demand for gold has been increasing since 2004 due to rising oil price, subprime mortgage crisis, financial crisis of 2007-2008, European debt crisis, Brexit announced by UK officially, a series of policy from the U.S. president Donald Trump, and the Federal Reserve decided to raise interest rates, etc. How to establish model that can predict the future direction of the spot gold price change is what investor concern about. The purpose of the thesis is to examine the gold price. The research uses the sample used for the analysis is the fortnightly data up to 442 records from January 1st, 2000 to December 30, 2016 to predict the future direction of the spot gold price change. The past study about gold price forecast model mostly decomposed nonstationary data into tendency, season, short-term circle and random items. This study used the Empirical Mode Decomposition (EMD) which was widely used in various fields. Three different setups of combinations or individuals of intrinsic mode functions (IMF) components resulting from EMD are considered as inputs for artificial neural network (ANN) model. Also compared the forecasting performance of conventional Autoregressive Integrated Moving Average model (ARIMA) and ANN model (without EMD progress). The experimental results indicate that EMD-ANN model performs better than the ANN model (without EMD progress) either with in-sample forecasting, or with out-of-sample forecasting. If compare the ANN model (without EMD progress) with the ARIMA in-sample forecasting, the performance of ARIMA model was more excellent. But in out-of-sample forecasting both models were at about the same performance. Therefore the ARIMA model predict the gold price for high accuracy rate than the ANN model (without EMD progress). In summary, the model with calculated IMF resulting from EMD as inputs for ANN got the minimize Mean Squared Error (MSE), and therefore the better performance of gold price forecast could be obtained by EMD-ANN model. LIN, TSAIR-CHUAN 林財川 2017 學位論文 ; thesis 54 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立臺北大學 === 統計學系 === 105 === The demand for gold has been increasing since 2004 due to rising oil price, subprime mortgage crisis, financial crisis of 2007-2008, European debt crisis, Brexit announced by UK officially, a series of policy from the U.S. president Donald Trump, and the Federal Reserve decided to raise interest rates, etc. How to establish model that can predict the future direction of the spot gold price change is what investor concern about.
The purpose of the thesis is to examine the gold price. The research uses the sample used for the analysis is the fortnightly data up to 442 records from January 1st, 2000 to December 30, 2016 to predict the future direction of the spot gold price change. The past study about gold price forecast model mostly decomposed nonstationary data into tendency, season, short-term circle and random items. This study used the Empirical Mode Decomposition (EMD) which was widely used in various fields. Three different setups of combinations or individuals of intrinsic mode functions (IMF) components resulting from EMD are considered as inputs for artificial neural network (ANN) model. Also compared the forecasting performance of conventional Autoregressive Integrated Moving Average model (ARIMA) and ANN model (without EMD progress).
The experimental results indicate that EMD-ANN model performs better than the ANN model (without EMD progress) either with in-sample forecasting, or with out-of-sample forecasting. If compare the ANN model (without EMD progress) with the ARIMA in-sample forecasting, the performance of ARIMA model was more excellent. But in out-of-sample forecasting both models were at about the same performance. Therefore the ARIMA model predict the gold price for high accuracy rate than the ANN model (without EMD progress). In summary, the model with calculated IMF resulting from EMD as inputs for ANN got the minimize Mean Squared Error (MSE), and therefore the better performance of gold price forecast could be obtained by EMD-ANN model.
|
author2 |
LIN, TSAIR-CHUAN |
author_facet |
LIN, TSAIR-CHUAN LIN, FANG-YI 林芳儀 |
author |
LIN, FANG-YI 林芳儀 |
spellingShingle |
LIN, FANG-YI 林芳儀 Predictability of gold price using a modified EMD-based artificial neural network model |
author_sort |
LIN, FANG-YI |
title |
Predictability of gold price using a modified EMD-based artificial neural network model |
title_short |
Predictability of gold price using a modified EMD-based artificial neural network model |
title_full |
Predictability of gold price using a modified EMD-based artificial neural network model |
title_fullStr |
Predictability of gold price using a modified EMD-based artificial neural network model |
title_full_unstemmed |
Predictability of gold price using a modified EMD-based artificial neural network model |
title_sort |
predictability of gold price using a modified emd-based artificial neural network model |
publishDate |
2017 |
url |
http://ndltd.ncl.edu.tw/handle/34957909391508410280 |
work_keys_str_mv |
AT linfangyi predictabilityofgoldpriceusingamodifiedemdbasedartificialneuralnetworkmodel AT línfāngyí predictabilityofgoldpriceusingamodifiedemdbasedartificialneuralnetworkmodel AT linfangyi lìyòngjīngyànmótàifēnjiěfǎdelèishénjīngwǎnglùmóxíngduìhuángjīnjiàgézhīyùcè AT línfāngyí lìyòngjīngyànmótàifēnjiěfǎdelèishénjīngwǎnglùmóxíngduìhuángjīnjiàgézhīyùcè |
_version_ |
1718523838621286400 |