Abstract: An original time series signal can be decomposed into a trend-cycle， including seasonal cycle, component and an irregular component. The trend-cycle part is defined as underlying level of the series, and is a manifestation of medium-long term variation influenced by fluctuation and cycles referring to generally deterministic or predictable change of a series. The irregular component contains the residual variation and random abrupt changes, etc., being unpredictable. Knowing the trend-cycle component, the irregular part can be calculated. Based on the investigation of the faults associated with the 1975 Haicheng MS7.3 earthquake, this study reasonably determined the area window of the Haicheng earthquake series. Then the time series of monthly earthquakes in the Haicheng seismicity window was decomposed. The trend-cycle component of the series was determined using ARIMA (atuo regression integrated moving average) model and the irregular variation was also extracted. The reaction of the anomalous abrupt variation to the MSge;6.0 earthquakes in North China and MSge;5.0 earthquakes near the seismicity window was analyzed. The result shows that the anomaly of abrupt seismicity variation may be taken as an indicator with prediction ability. This is of significance in earthquake prediction.