| Summary: | 碩士 === 雲林科技大學 === 資訊管理系碩士班 === 96 === In stock market, many systems or methods are used to predict stock prices such as neural network, genetic algorithm, and traditional statistic. However, there are some problems for investors among these methods such as no rules for making decision, high time complexity in computing outcome, or some strict mathematic distribution assumption for datasets. Therefore, this dissertation has proposed a new method which combines fuzzy theory and granular rough set algorithm in forecasting process and use technical indicators to produce effective rules for forecasts. Four main procedures are contained in the method: (1) transfer the basic index of dataset (time , open index, high index, low index, close index, and volume) into technical indicators(MA, RSI, PSY, STOD, VR, OBV, DIS, AR, ROC ); (2) use MEPA (Minimize Entropy Principle Approach)and CPDA(Cumulative probability distribution approach) to granulate the technical indicators; (3) employ rough set theory to extract effective rules from the granulated dataset of technical indicators for forecasting; and (4) use the extracted rules to produce forecasts and evaluate the performance of the forecasting model with RMSE(Root Mean Square Error). From model verification and comparison, the new forecasting method surpasses in accuracy the listing conventional fuzzy time-series models referred in this dissertation.
|
|---|
