Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network
碩士 === 華梵大學 === 資訊管理學系碩士班 === 98 === This study attempted to predict trends of Taiwan Weighted Stock Index (TWSI) by two kinds of different professional technologies - Back-propagation Network (BPN) and Recurrent Neural Network (RNN), and then compared prediction results. The use of neural network i...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2010
|
Online Access: | http://ndltd.ncl.edu.tw/handle/75503610214195977932 |
id |
ndltd-TW-098HCHT0396040 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-098HCHT03960402015-10-13T18:20:59Z http://ndltd.ncl.edu.tw/handle/75503610214195977932 Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network 以回饋式類神經網路於台灣加權股價指數預測之研究 Kuang-Ting Hsiao 蕭光廷 碩士 華梵大學 資訊管理學系碩士班 98 This study attempted to predict trends of Taiwan Weighted Stock Index (TWSI) by two kinds of different professional technologies - Back-propagation Network (BPN) and Recurrent Neural Network (RNN), and then compared prediction results. The use of neural network in predicting trends of stock prices is one of the frequently used methods; in particular, the use of BPN prevails. Being a dynamic neural network, RNN can, in the dominant presentation manner, directly present "time factor" in network structure by loop method, and has the advantage function suitable for process of time sequence of stock price trends. This study took data of 1580 trading days as training test sample. The time period is six years and five months. Training period: from January 2, 2003 to February 20, 2009. After the training phase, data was backward divided into 5 days, 20 days and 60 days, as test data. Input variables adopted by this study totaled 45 items: 16 items of technical indices, 7 items of international stock markets, 7 items of macroeconomics and 15 items of post trade information. Empirical results suggested that both partially recurrent neural network (PRNN) and fully recurrent neural network (FRNN) were better than BPN; and the shorter the time after training period, the better the predicated values. Yeou-Ren Shiue 薛友仁 2010 學位論文 ; thesis 63 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 華梵大學 === 資訊管理學系碩士班 === 98 === This study attempted to predict trends of Taiwan Weighted Stock Index (TWSI) by two kinds of different professional technologies - Back-propagation Network (BPN) and Recurrent Neural Network (RNN), and then compared prediction results. The use of neural network in predicting trends of stock prices is one of the frequently used methods; in particular, the use of BPN prevails. Being a dynamic neural network, RNN can, in the dominant presentation manner, directly present "time factor" in network structure by loop method, and has the advantage function suitable for process of time sequence of stock price trends.
This study took data of 1580 trading days as training test sample. The time period is six years and five months. Training period: from January 2, 2003 to February 20, 2009. After the training phase, data was backward divided into 5 days, 20 days and 60 days, as test data. Input variables adopted by this study totaled 45 items: 16 items of technical indices, 7 items of international stock markets, 7 items of macroeconomics and 15 items of post trade information. Empirical results suggested that both partially recurrent neural network (PRNN) and fully recurrent neural network (FRNN) were better than BPN; and the shorter the time after training period, the better the predicated values.
|
author2 |
Yeou-Ren Shiue |
author_facet |
Yeou-Ren Shiue Kuang-Ting Hsiao 蕭光廷 |
author |
Kuang-Ting Hsiao 蕭光廷 |
spellingShingle |
Kuang-Ting Hsiao 蕭光廷 Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network |
author_sort |
Kuang-Ting Hsiao |
title |
Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network |
title_short |
Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network |
title_full |
Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network |
title_fullStr |
Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network |
title_full_unstemmed |
Study on predicting Taiwan Weighted Stock Index by use of Recurrent Neural Network |
title_sort |
study on predicting taiwan weighted stock index by use of recurrent neural network |
publishDate |
2010 |
url |
http://ndltd.ncl.edu.tw/handle/75503610214195977932 |
work_keys_str_mv |
AT kuangtinghsiao studyonpredictingtaiwanweightedstockindexbyuseofrecurrentneuralnetwork AT xiāoguāngtíng studyonpredictingtaiwanweightedstockindexbyuseofrecurrentneuralnetwork AT kuangtinghsiao yǐhuíkuìshìlèishénjīngwǎnglùyútáiwānjiāquángǔjiàzhǐshùyùcèzhīyánjiū AT xiāoguāngtíng yǐhuíkuìshìlèishénjīngwǎnglùyútáiwānjiāquángǔjiàzhǐshùyùcèzhīyánjiū |
_version_ |
1718030372543922176 |