A Novel Support-Vector-Machine-Based Prediction System with Modified Grey Fourier Series and Fuzzy Time Series
| Autor: | Yu-Tung Lin, 林玉堂 |
|---|---|
| Rok vydání: | 2017 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 105 The main purpose of this study is to improve the performance of grey theory in non-smooth time series, so that after modification, the theory can predict stock value more accurate both in the trend or non-trend, especially Taiwan’s weighted share price index. In the trend, we use the modified GM (1,1) in which the initial value of the background is set in the near-term data, and use the Fourier series and exponential smoothing method to perform primary and secondary residual correction. In the non-trend, we mainly use the fixed range characteristics to find out the technical index of high relevance. Through the predicted technical index placement value predicted by the fuzzy time series of defuzzification center of gravity method, we reverse the stock value. Finally, through the hidden Markov forecast non-trend segment value, we predict actual forecast stocks. Through analysis and decision-making process provided by the support vector machine, grey theory integrated the advantages of both trend and non-trend system and will be able to work effectively in terms of stock value prediction. The use of Markov model helps to find out in which section stock prices may fall, correct the real final stock price forecast, and obtain the value of system prediction. It is proved that the modified GM (1,1) can effectively improve the traditional GM (1,1) prediction when faced with market reversal and randomness. Combined with the hidden Markov model, the use of technical index conducting fuzzy time series can effectively find non-trend market turning point. The integration of modified GM (1,1) and technical index systems makes more accurate predictions in both trend and non-trend. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|