Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm
碩士 === 國立高雄應用科技大學 === 資訊管理研究所碩士班 === 103 === In the field of financial investment and financial management, there are many types of investment, such as stocks, funds, and futures. Among them, many investors choose stock for it is a high-risk and high-yield investment. Knowing how to effectively pred...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2015
|
Online Access: | http://ndltd.ncl.edu.tw/handle/a73b7c |
id |
ndltd-TW-103KUAS0396009 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-103KUAS03960092019-05-15T22:08:04Z http://ndltd.ncl.edu.tw/handle/a73b7c Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm 利用技術指標於股價預測:使用SVM與特徵選取演算法 Jun-Wei Huang 黃鈞偉 碩士 國立高雄應用科技大學 資訊管理研究所碩士班 103 In the field of financial investment and financial management, there are many types of investment, such as stocks, funds, and futures. Among them, many investors choose stock for it is a high-risk and high-yield investment. Knowing how to effectively predict the change of stock priceis what investors care about, and many investors use technical analysis indicators as major methods. In this research Support Vector Machines (SVM) were to build models for prediction of stock market. The training data and testing data sets were collected from Taiwan Economic Journals (TEJ) and were based on stock exchange.The parameters for prediction models weretechnical analysis indicators, such as Psychological Line (PSY), Williams %R (WMS%R), and Relative Strength Index (RSI), setting the change of stock price as 0.5% ( Changes over ±5% are considered rise and fall, and the rest are flat) to train with different response time (1 day, 3 days, and 5 days). In this research two types of feature selection method were conducted, which were Stepwise Feature Selection and Sequential Feature Selection. By using feature selection with cross validation, the best feature combination was obtained, which allowed us to know the best combination of technical analysis indicators, and to enhance the accuracy and effectiveness of stock market prediction models. The results show that technical analysis in the setting of 5-day response time has the highest accuracy, and is better than other classification algorithms,while in the setting of one-dayand 3-day response time, the accuracy is affected by the excessive amount of flat trainingdata. Furthermore, the use of feature selection reveals that in the setting of one-day and 3-day response time, the cross-validation rate of bias ratio (BIAS) and stochastic oscillator (KD) is respectively highest and second highest, and in the setting of 5-day response time, the cross-validation rate of RSI is the highest and slightly higher than ones of BIAS and KD. According to the results above, RSI, BIAS and KD are applicable to the response time of one day, three days, and five days respectively. Keywords: Feature Selection,Support Vector Machine, Stock Price Prediction Pei-Yi Hao 郝沛毅 2015 學位論文 ; thesis 102 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立高雄應用科技大學 === 資訊管理研究所碩士班 === 103 === In the field of financial investment and financial management, there are many types of investment, such as stocks, funds, and futures. Among them, many investors choose stock for it is a high-risk and high-yield investment. Knowing how to effectively predict the change of stock priceis what investors care about, and many investors use technical analysis indicators as major methods.
In this research Support Vector Machines (SVM) were to build models for prediction of stock market. The training data and testing data sets were collected from Taiwan Economic Journals (TEJ) and were based on stock exchange.The parameters for prediction models weretechnical analysis indicators, such as Psychological Line (PSY), Williams %R (WMS%R), and Relative Strength Index (RSI), setting the change of stock price as 0.5% ( Changes over ±5% are considered rise and fall, and the rest are flat) to train with different response time (1 day, 3 days, and 5 days).
In this research two types of feature selection method were conducted, which were Stepwise Feature Selection and Sequential Feature Selection. By using feature selection with cross validation, the best feature combination was obtained, which allowed us to know the best combination of technical analysis indicators, and to enhance the accuracy and effectiveness of stock market prediction models.
The results show that technical analysis in the setting of 5-day response time has the highest accuracy, and is better than other classification algorithms,while in the setting of one-dayand 3-day response time, the accuracy is affected by the excessive amount of flat trainingdata. Furthermore, the use of feature selection reveals that in the setting of one-day and 3-day response time, the cross-validation rate of bias ratio (BIAS) and stochastic oscillator (KD) is respectively highest and second highest, and in the setting of 5-day response time, the cross-validation rate of RSI is the highest and slightly higher than ones of BIAS and KD. According to the results above, RSI, BIAS and KD are applicable to the response time of one day, three days, and five days respectively.
Keywords: Feature Selection,Support Vector Machine, Stock Price Prediction
|
author2 |
Pei-Yi Hao |
author_facet |
Pei-Yi Hao Jun-Wei Huang 黃鈞偉 |
author |
Jun-Wei Huang 黃鈞偉 |
spellingShingle |
Jun-Wei Huang 黃鈞偉 Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm |
author_sort |
Jun-Wei Huang |
title |
Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm |
title_short |
Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm |
title_full |
Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm |
title_fullStr |
Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm |
title_full_unstemmed |
Using Technical Indicator for Stock Price Forecast: Using SVM and Feature Selection Algorithm |
title_sort |
using technical indicator for stock price forecast: using svm and feature selection algorithm |
publishDate |
2015 |
url |
http://ndltd.ncl.edu.tw/handle/a73b7c |
work_keys_str_mv |
AT junweihuang usingtechnicalindicatorforstockpriceforecastusingsvmandfeatureselectionalgorithm AT huángjūnwěi usingtechnicalindicatorforstockpriceforecastusingsvmandfeatureselectionalgorithm AT junweihuang lìyòngjìshùzhǐbiāoyúgǔjiàyùcèshǐyòngsvmyǔtèzhēngxuǎnqǔyǎnsuànfǎ AT huángjūnwěi lìyòngjìshùzhǐbiāoyúgǔjiàyùcèshǐyòngsvmyǔtèzhēngxuǎnqǔyǎnsuànfǎ |
_version_ |
1719124970463821824 |