Abstract: A fuzzy time series is set up to forecast the securities price, which is a trapezoidal fuzzy number, and the forecasting of return on investment is measured in the ratio of forecasted value to purchasing price of securities. The model is solved by compromise programming based on the double-object model and compared with the effect of portfolio in terms of mean-absolute deviations. An empirical analysis is therefore made for 15 stocks picked out from SSE 50 indexing stocks. The results revealed that the compromise programming is available for decision-making of investment in accordance to the market quotation trend. The compromise programming also avoids the invalid solution to the problem that the double-objective model in terms of mean-absolute deviations under strict restriction on data.