The evaluation of forecasting performance with Machine Learning
碩士 === 國立高雄科技大學 === 財務管理系 === 107 === In recent years, more and more scientist research how to create a system which can predict the direction of the underlying price correctly in the financial domain. Fundamental analysis and technical analysis have been verified by the prior study that has super...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2019
|
Online Access: | http://ndltd.ncl.edu.tw/handle/cdm87h |
id |
ndltd-TW-107NKUS0305016 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-107NKUS03050162019-07-16T03:45:10Z http://ndltd.ncl.edu.tw/handle/cdm87h The evaluation of forecasting performance with Machine Learning 以機器學習預測台灣50指數漲跌幅並探討模型準確性之研究 許庭瑀 HSU, TING-YU 許庭瑀 碩士 國立高雄科技大學 財務管理系 107 In recent years, more and more scientist research how to create a system which can predict the direction of the underlying price correctly in the financial domain. Fundamental analysis and technical analysis have been verified by the prior study that has superior ability to predict the movement of prices. Furthermore, some of the researchers employed financial ratios and technical indicators as parameters of the machine learning algorithm to forecast the future movement of the financial instruments. In this research, we used technical indicators as our input features and the forecast horizon for each indicator divided into 5, 10, 15, 20. In the financial market, we always consider five days as one trading week; and on this basis, 20 days considered as one trading month. This paper aims to evaluate the performance of algorithms of machine learning models which utilized technical indicators as their input. In our research, when the input window length is 20 days, the performance of each model is the highest. The consequence is consistent with the prior study’s result (Shynkevich et al., 2017). LEE, WEN-YI LIN, TSAI-YIN 李文毅 林財印 2019 學位論文 ; thesis 31 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立高雄科技大學 === 財務管理系 === 107 === In recent years, more and more scientist research how to create a system which can predict the direction of the underlying price correctly in the financial domain. Fundamental analysis and technical analysis have been verified by the prior study that has superior ability to predict the movement of prices. Furthermore, some of the researchers employed financial ratios and technical indicators as parameters of the machine learning algorithm to forecast the future movement of the financial instruments. In this research, we used technical indicators as our input features and the forecast horizon for each indicator divided into 5, 10, 15, 20. In the financial market, we always consider five days as one trading week; and on this basis, 20 days considered as one trading month. This paper aims to evaluate the performance of algorithms of machine learning models which utilized technical indicators as their input. In our research, when the input window length is 20 days, the performance of each model is the highest. The consequence is consistent with the prior study’s result (Shynkevich et al., 2017).
|
author2 |
LEE, WEN-YI |
author_facet |
LEE, WEN-YI 許庭瑀 HSU, TING-YU 許庭瑀 |
author |
許庭瑀 HSU, TING-YU 許庭瑀 |
spellingShingle |
許庭瑀 HSU, TING-YU 許庭瑀 The evaluation of forecasting performance with Machine Learning |
author_sort |
許庭瑀 HSU, TING-YU |
title |
The evaluation of forecasting performance with Machine Learning |
title_short |
The evaluation of forecasting performance with Machine Learning |
title_full |
The evaluation of forecasting performance with Machine Learning |
title_fullStr |
The evaluation of forecasting performance with Machine Learning |
title_full_unstemmed |
The evaluation of forecasting performance with Machine Learning |
title_sort |
evaluation of forecasting performance with machine learning |
publishDate |
2019 |
url |
http://ndltd.ncl.edu.tw/handle/cdm87h |
work_keys_str_mv |
AT xǔtíngyǔhsutingyu theevaluationofforecastingperformancewithmachinelearning AT xǔtíngyǔ theevaluationofforecastingperformancewithmachinelearning AT xǔtíngyǔhsutingyu yǐjīqìxuéxíyùcètáiwān50zhǐshùzhǎngdiēfúbìngtàntǎomóxíngzhǔnquèxìngzhīyánjiū AT xǔtíngyǔ yǐjīqìxuéxíyùcètáiwān50zhǐshùzhǎngdiēfúbìngtàntǎomóxíngzhǔnquèxìngzhīyánjiū AT xǔtíngyǔhsutingyu evaluationofforecastingperformancewithmachinelearning AT xǔtíngyǔ evaluationofforecastingperformancewithmachinelearning |
_version_ |
1719223850890166272 |