Performance Improvement for Frequent Term-based Text Clustering Algorithm

Frequent term-based text clustering [2] is a recently introduced text clustering technique, which uses frequent term sets and dramatically decreases the dimensionality of the document vector space, thus especially addressing itself to the problems of text clustering: very high dimensionality of the date and very large size of the databases [2]. Moreover, frequent term sets provide understandable meanings for clusters. Frequent term-based text clustering algorithm (FTC) has shown significant efficiency comparing to some well-known text clustering methods [2], but the quality of clustering still needs further enhancement. This report points out the problems of the overlap calculations in FTC and introduces polished algorithms that aim at improvements on both the running time and the cluster quality. The performances of FTC before and after improvements are compared on the basis of the experiments on classical text documents as well as on web documents. At last, an evaluation on the clustering procedure may provide a clue for further work. General Terms Algorithms, Performance, Experimentation.