In the preliminary design stage of ships, the ship's characteristics are estimated by the statistic correlation between particulars. This situation is the same in the case where the computer aided design techniques are adopted. Therefore it is very important to have the reliable correlation data of particulars with the rational background. This can only be made by the reasonable classification of ships. The classification of ships has usually been made according to the mission of ships, cargo boats, bulk carriers and so on. However, there would be another method to classify ships which provides us with more reliable statistic correlation depending on the specified purposes. The cluster analysis and the principal component analysis were adopted in this paper for the data set of 628 existing ships. The grouping of ships was made automatically by the cluster analysis connecting a group to the nearest neighbour. The obtained results were displayed on the two dimmensional plane which was spanned by the first and second components. The classification of ships obtained by this method were as follows : 1. Car ferry boats 2. Container ships and cargo liners 3. Cargo boats and bulk carriers 4. Medium size bulk carriers and are carriers 5. Very large tankers and combination carriers These results seem to be a conventional classification. However, the characteristics of these group can be explained very well and the boundary of two groups were made fairly clear by this method.