Towards Using KDD for an Interactive Software Architecture Extraction

Abstract

Knowledge Discovery in Databases (KDD) tools emerged as a solution to discover information and knowledge from rapidly growing data; they allow the execution of interactive and iterative processes in order to obtain new useful information from data. In this work, we present a KDD tool extension that allows the execution of interactive software architecture extraction processes. Actually, even though every system has an architecture, not every system has a reliable architecture description that represents the system parts and their relationships. To overcome this problem that puts the system maintenance and reuse in danger, the extraction of the system architecture is proposed. The extraction of a software architecture from an existing system consists at analyzing the source code in order to discover a software architecture description (the system parts and their relationships). While several extraction approaches proved to be efficient, there is no approach allowing an architect to integrate his knowledge and interact with the extraction process. To deal with this lack, we suggest the use of KDD tools to extract a software architecture. Unfortunately, the KDD tools allow the execution of processes with user interaction but they do not support the execution of software architecture extraction processes. Thus, we provide in this work a KDD tool extension that enables the execution of software architecture extraction processes. To achieve our goal, we have firstly elaborated a process model for software architecture extraction, then we have implemented an extension for a KDD tool (called KNIME) that supports the proposed model.