Abstract: Knowledge fusion is a new concept proposed at the end of last century. The fused knowledge is referred as method, experience and idea as well as data and information. A new knowledge fusion method based on D-S(Dempster-Shafer) evidence theory was presented and applied to corporate failure prediction. There were three steps in the method: mathematic modeling, knowledge fusing and discriminating. The implementation of these steps was developed. The information entropy theory was employed to determine the basic probability assignment of knowledge. The knowledge fusing equation was induced following D-S evidence combination rule. A concise discriminating regulation was proposed to analyze the fusion result. The data of two listed companies was used to show the validity of the presented method. The results indicate this method can reduce the uncertainty of the prediction model effectively.