Abstract: Deep channel sand body is characterized by deep burial, small thickness and weak seismic reflection, and it is difficult for conventional seismic attribute prediction technology to identify deep thin reservoirs. Based on the stratigraphic classification and correlation and the seismic reflection analysis, the paper establishes the seismic response characteristics of the channel sand bodies by using a forward model, studies the attributes sensitive to seismic targets, extracts reasonable attributes of small time window for geological targets, then performs clustering constrained fusion for these small time windows in different areas and conducts detailed characterization for favorable reservoirs, develops a grading sensitive attribute fusion method and finally establishes a reservoir prediction technology for deep channel sand bodies, which can guide the exploration for deep channel sand bodies