Latent semantic analysis for business protocol discovery using log files

Abstract

Recently, business protocol discovery has great importance in the field of web services. This field of research which aims at discovering the dynamic behaviour of the web service is mainly based on the execution history and reverse engineering techniques. On the other hand, latent semantic analysis LSA is a mathematical approach that aims at extracting relationships between large sets of data with regard to the context of data occurrence. In this paper, a new LSA-based approach for business protocol discovery is presented. A new microscopic/macroscopic pattern is defined and used for allowing incremental composition of protocols. The proposed approach uses synthetically generated log files, with the presence of noise and incomplete information. The results obtained from applying the proposed approach have proved its efficiency for business protocol discovery.