Hepatocellular carcinoma (HCC) is one of the common malignant tumors.Accurate diagnosing of HCC is of great importance.k-NFN (k-Nearest and Farthest Neighbors)[1]is an ensemble classifying method which labels the input samples based on the voting of both the nearest neighbors and the farthest neighbors.This paper proposes an improved k-NFN (I-k-NFN) method based on sequential forward selection (SFS).The N-feature subset and F-feature subset are determined by SFS.Hence,for an input sample,the nearest neighbors and farthest neighbors are calculated based on N-feature subset and F-feature subset,respectively,and then final voting from the nearest neighbors and the farthest neighbors is fused.