Since official operation in 2019,CENC Automatic Aftershock Forecasting System(abbr. CAAFs) has provided both timely data and technical support for post-earthquake emergency responses and post-earthquake trend analysis. This article summarizes the data output of the CAAFs for 4.5-year operation,and evaluates the aftershock prediction results produced by the system from both national and regional areas. Our results show that:①the accuracy of upper-limit earthquake magnitude prediction (i.e. the happening probability of earthquakes with a certain magnitude or above is not high) is mostly ≥ 90 %,which is significantly better than the interval earthquake magnitude prediction(i.e. the happening possibility of earthquakes with the same magnitude or a certain magnitude occurring with an error of ±0.5). There is not much difference in the predicted results in different seismic zones; ②The predicted magnitudes in the interval are generally higher than the actual earthquake magnitude values. The proportion of earthquakes with magnitude deviation(actual maximum aftershock magnitude within 7 days minus predicted magnitude)falling into[-0.5,0.5] is about 44 %,falling into[-1,1] is about 69 %. The closer the magnitude deviation is to the[-0.5,0.5] interval,the higher the proportion of earthquakes are; ③The result of interval prediction in southwestern China is better than that in the northwestern,southern,and north-northeastern China. The predicted results of M4.0~5.0 earthquakes are not as good as those of other magnitude ranges,which may be related to the large range of changes in the maximum aftershock magnitude of earthquakes in this magnitude range.