According to the fractal market hypothesis, the current share prices do not absorb all the historical market information, we think that the price series information is a sort of grouped incomplete fuzzy information. The theorem that the information volume is a constant under the linear information distribution function is proved. We set up a pattern recognition model, based on the fuzzy information distribution theory, to predict the direction of fluctuation of Shanghai stock market composite index day time series. By empirical analysis it is found that this model is very effective and can also give the possible distribution of logarithmic stock return. Finally the framework of decision support system for pattern recognition of short-term share price fluctuation is put forth