A Hierarchy Cluster Method for Functional Data

An analysis is made of the mechanism of Functional Principal Component Analysis, and a new method of hierarchy clustering for functional data is presented, with which visualizing the original high dimensional datum in a low dimensional space is realizable. A definition of distance between functions is given. The relationship between the argued form of distance and the Euclidean distance is investigated. To prove validity, the proposed method is applied to classify the quality of cookies from curves representing the resistance of dough observed during the kneading process.