Stock investing is an exciting and challenging monetary activity, and forecasting stock trend and price plays an important role in stock market. The stock investors could have a chance to make much money in stock returns with wise decisions; however the most investors keep a pessimistic image with a heavy loss of money in stock investment. Due to stock market behavior is nonlinearity and non-stationary, and the stock price fluctuation is extremely hard to predict correctly if without have experienced or expert knowledge. Up to date, it is difficult to build a general model for forecasting stock price accurately. Nevertheless, many researchers continue to establish feasible model for approximating stock market behaviors. Forecasting activities play an important role in our daily life; the goal of forecasting activities is to increase accuracy and profit, such as chasing predict [38] and control prediction. In financial engineering, Kuo et al. [17] have demonstrated that the general techniques used for stock market prediction are mathematical and statistical models. In time-series analysis [14], there are many time series models such as ARIMA (Autoregressive Integrated Moving Average model, [3]) and GARCH (Generalized Autoregressive Conditional Heteroskedasticity, [2]), these models have been applied to forecast stock price and trends in the financial market. However, statistical models usually deal with linear forecasting model and variables must obey statistical normal distribution for better forecasting performance. If the research data are represented by linguistic values (also named it, “linguistic intervals” such as linguistic values of age is very young, young, old) or the number of sample data is very little, the traditional forecasting methods maybe generate the bias of forecast or poor results. Therefore, many researchers have proposed different forecasting models based on fuzzy theory [37] to solve time-series problems with linguistic values. Song and Chissom [27] first proposed a fuzzy time-series model to forecast the enrollments at University of Alabama; the fuzzy time-series model constructed the fuzzy relation R and used a Max–Min composition operator to calculate forecasting values. Chen [4] proposed a fuzzy time-series model which used equal interval lengths to partition the universe of discourse and generate forecasting rules with a simplified calculation process. However, in stock forecasting, Huarng [12] extended Chen’s model with additional heuristic forecasting rules to produce forecasts. In a subsequent study, Huarng proposed another model to define interval length with distribution based length and average-based length (2001). And Yu [35] proposed a weighted fuzzy time-series model with recurrent fuzzy relationships to produce forecasts. Sun et al. [30] did a prediction of Chinese stock index (CSI) future prices using fuzzy sets and multivariate fuzzy time series method. Aladag˘ et al. [1] proposed a partial high order Fuzzy lagged variable selection in fuzzy time series with genetic algorithms.