An earthquake forecasting system based on sequential data assimilation of slip on the plate boundary is constructed. Sequential Importance Sampling, a type of particle filter, is used for the data assimilation. A forward calculation is made using an earthquake generation cycle simulation. From the simulation results, we estimate crustal deformation that can be compared with observations of the GNSS Earth Observation Network System (GEONET) on land and the Dense Ocean Floor Network for Earthquakes and Tsunamis (DONET) on the seafloor. We demonstrate a numerical test of this system using synthetic data of seafloor deformation before virtual Nankai earthquakes. We also use a real GEONET data set to compare the results of earthquake generation cycles.