Abstract: There exist a lot of uncertainties during drilling operation, with high investment and high risks. Up to now, there is no quantitative risk probability prediction model for drilling operation site. This paper proposes the ARIMA model for quantitative prediction of drilling operation risks that may take place in future. Taking the field risk probability recorded from Longgang M well as samples of time sequence, the statistical characteristics are extracted followed by ADF inspection and pure random test with SAS software. Combining the ACF and PACF charts with the criterions of AIC, SBC and standard error, the final ARIMA (2,1,3) model has been developed, with the error of simulation of 4.477 4%. The prediction results show that the model is more suitable for drilling operation site in deep carbonate fractured reservoirs, and the short?term prediction results are better than the long?term prediction ones