Research on the clustering algorithm of component based on the grade strategy

Abstract

With the development of software component technology and the rapid increase in the number of components, reasonable component classification is the foundation to achieve fast and effective retrieval. Currently, the faceted classification method often has subjective factors. And in traditional component clustering, the similarity between components is calculated by average thought which is unable to well satisfy user requirements. So, from the point of user requirements, grade strategy is introduced, which gives each facet different grade weight; and the similarity between components is objectively calculated by synthesis account so as to overcome the irrationality of traditional component similarity calculation method, and then a component clustering algorithm based on the grade strategy is proposed. Experiments of component clustering based on Vector Space Model and Latent Semantic Analysis with the grade strategy have been made. They prove that component clustering algorithm based on the grade strategy makes the component clustering result more humanized and better serve user requirements. Thereby, the quality of component classification can be improved effectively.