INTRODUCTION: To develop a model for forecasting the mortality of stroke in Tianjin, China. The time series of stroke mortality from 1999 Jan. to 2006 Dec. in Tianjin city were subjected. Circle distribution analysis was used to verify the trend of time concentration. Multiple seasonal autoregressive integrated moving average model [ARIMA (p, d, q) (P, D, Q) s], based on model identification, estimation and verification of parameter, and analysis of the fitting of model, was established. Most of the deaths from stroke occurred in January and had a cycle of 12 months. An ARIMA model (0, 1, 0) x (0, 1, 1)12 was established (1--B) (1-- B12) lnx(i) = 0.001 + (1--0.537 B12)epsilon(t). CONCLUSION: ARIMA & Circle Distribution analysis is an important tool for stroke mortality analysis. Potentially it has a high practical value on the surveillance, forecasting and prevention of stroke mortality.