Abstract: South-North Seismic Belt is a main active belt in China's mainland region. There were many big earthquakes occurred along this belt since twentieth century. To study the variation pattern of the seismic activity along the South-North Seismic Belt and to make a medium-term forecast in it, a forecasting mode with the technology of artificial neural network is made in this paper. The study was co nducted on the evolution of sequence of maximum earthquake mag nitude along the South-North Seismic Belt. Besides, a new applied way to fix on the size of network input window, which enables us to obtain better forecasting effect, is advanced. It is indicated that the st rong-weak cycle of seismic activity is an important feature of the South-North Seismic Belt, and the error between forecasting value of NN and practical value is less, therefore this modeling can be used as the forecast modeling for the South-North Seismic Belt.

Key words: Seismic activity strength, Strong-weak earthquake cycle, Artificial neural network, Modeling, Forecast