Abstract: A new visual clustering method for high dimensional data is presented, and a general visualization system VisNN is implemented in this paper. Classical visual method for clustering high dimension data is to project high dimensional data into two-dimension or three-dimension space by using dimension reduction method. In the method for each axis of two-dimension X-Y space, a record series, instead of a reduced attribute, are used. When the order of records is sorted by some attributes that users concern about, the distance relationship of two records in high dimensional space could be kept to some extent in the X-Y space. The experiments show that VisNN can help users discover interesting clusters and abnormality easily.