Abstract: As is well known, before most strong earthquakes some properties of seismicity appear. Taking several statistical indexes of seismicity, the ISODATA fuzzy clustering analysis is used to do comprehencive judgement of earthquake hazard. This method has two advantages. One is a certain accuracy can be fulfil, the other is it is probably impossible to classify the predicted sample of the forthcoming period into an independent class. If each selected index carries some fuzzy information, a reasonable judgement for earthquake hazard could be made according to whether the predicted sample falls into a class having seismic risk or a class without seismic risk.Two actual examples, one a -month and the other a-half-year-prediction, are calculated.