Using Par-v-SVC For Stock Price Forecasting

• Main Authors: Min-Wei Chiang, 江旻緯
• Other Authors: Pei-Yi Hao
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/96607272986322083315

碩士 === 國立高雄應用科技大學 === 資訊管理研究所碩士班 === 102 === Stock has been a higher rate of return but a higher risk on investment in the market,so investors focus that how to get a great price forecasting model.In this study,we apply Support Vector Machine(par-v-SVC)to build prediction model by technical indicators (BIAS,PSY,RSI…).There are two parts of research in this study.First, dividing the reaction time of technical indicators into three parts(one,three,five days) and dividing the change rate of stock price into two parts(1%,3%).Second, comparing the accuracy with algorithms of different classification(Neural Networks, Naive Bayes Classifier, Decision tree).The result obtained by par-v-SVC is the best compared to the other classification algorithms. Experimental results show the forecast accuracy of change rate 3% is better than 1%.The reaction time of one day has the worst forecast accuracy in change rate 1% and the reaction time of one day has the best forecast accuracy in change rate 3%.