How to Make Sense of Empirical Software Engineering Data. An Integrated Approach

Discovery, representation, and reuse of software engineering know-how are of crucial importance for realizing learning software organizations. Maximum insight from the systematically established industrial experiments can only be achieved from synergy of formal and semi-formal methods and techniques on the one hand, and human expert knowledge on the other hand. The paper describes an integrated and interactive knowledge discovery approach from empirical software engineering data called INTEGERgqm. It can be seen as measurement-related implementation of the Quality Improvement Paradigm and is fundamentally based on goal-oriented measurement and encompasses reuse of software artifacts. Abstraction sheets, feedback sessions, and rough set analysis are applied as supplementary techniques. Special emphasis is dedicated to demonstrate how the different parts of the overall approach support acquisition of explicit and implicit knowledge based on interaction with human experts. The overall app roach was successfully applied for two consecutive and related pilot projects at Allianz Life, market leader of life insurer in Germany. The paper describes main experiences and results from these pilot projects incorporating the insurance product adjustments in the years 1996 and 1997 into the existing software products. Improvement in software development was essentially based on achieving improvements in requirements engineering and reuse of know how.