Abstract: This paper proposes a novel approach to forecast technology development based on a knowledge network. Taking the data of 1101 nanotechnology-related patents from 2011 to 2015 as an example, using the patents’ knowledge combinatorial potential, combinatorial opportunities, diversity and uniqueness as indexes, we employ K-means clustering to divide the patents into two classes. By analyzing the content of the patent cluster with high knowledge combinatorial potential, combinatorial opportunities, diversity and uniqueness, hot fields and development trends of nanotechnology research are extracted. The effectiveness of the method is tested by using the data of nanotechnology-related patents in 2017.