An acquisition unit acquires feature vectors (the number is N). A first selection unit selects first neighbor features (k (1≦k≦N)) of each feature vector in order of higher similarity from the feature vectors. A second selection unit generates a plurality of groups each including similar feature vectors from the feature vectors, and selects second neighbor features (u (1≦k+u≦N−2)) of each feature vector in order of higher similarity. Each of the second neighbor features is differently included in a group. A determination unit calculates a density of each feature vector by using a threshold, the first neighbor features and the second neighbor features, and determines feature vectors to be classified into the same class as each feature vector by using the density and a threshold. A classification unit classifies the feature vectors into a plurality of classes by using the selection result. A control unit controls each threshold.