| Summary: | 碩士 === 國立高雄第一科技大學 === 金融系碩士班 === 106 === This paper applies a deep learning model to explore stock price prediction by utilizing the non-linearity of said deep learning model to achieve better accuracy and rate of return as compared to traditional prediction methods. It also overcomes the arbitrary subjectiveness of market sentiment through quantitative trading and reduces the probability of investment failure in the stock market. The primary methodology used in this study is a long short-term memory model, and technical analysis of data variables is used to study how the variables for the previous three periods from June 1, 2015 to May 30, 2018 can be used to predict the closing prices of the Taiwan 50 index in the following period, and to test accuracy and rate of return. The results show that, in terms of accuracy, Root-mean-square deviation RMSE is 1.619 and R is 0.55, while Theil’s U is 0.01. The annual rate of return is 10.62%, which surpassed the performances of the deep neural networks and the corresponding period, showing that this model can be a profitable tool for investors.
|
|---|
