In software development, to implement given functional specifications, developers generally use class files from well-known class libraries. Therefore, a set of library classes used by software seems to have a strong relationship with the functionality of the software. In this paper we propose a method to classify software from their functional viewpoints based on used classes. In the proposed method, the distance between two software is calculated based on the number of uses of each library class, and, software is classified by hierarchical clustering. As a result of an experimental evaluation, the most appropriate classification was obtained when the tf-idf method and Euclidean distance was employed to calculate the distance.