The financial time series analysis pays attention to the fast prediction of trend.An algorithm for fast discovering frequent pattern in financial time series data is proposed based on data mining.Comparing with traditional methods,the algorithm requires no special demand on the distribution and stability of raw data,is not based on any hypothesis,and can fast discover the frequent patterns from time series data.The discovered frequent patterns can be used for prediction of financial trend by matching patterns.In addition,the experiments on exchange rate data demonstrate the utility of this approach.