Design and implementation of a knowledge-based decision support system for estimating software development work-effort

Abstract

Up to now, the assessment of work-effort in software engineering is based on statistical methods. Among the best known are COCOMO (Boehm [2]) or SPQR (Jones [6]). Nevertheless it is generally recognized that many qualitative factors enter into the cost of development, such as effectiveness of the team, user's motivation, and accuracy of the specifications. We have designed a Decision Support System (DSS) for estimating the work-effort, in which the processing of the qualitative data is made by an expert system while a function points analysis provides the theoretical work-effort according to the type of software and the past experience. The evaluation is performed at two levels: global and detailed. The global evaluation is made at the beginning of the development according to the data that are, at this moment, available. The detailed evaluation takes place when the design of the software becomes more precise. The software manager can follow the evolution of the changes at the detailed level during the development.

In software development, project leaders mostly reason by using their past experience. It therefore follows that a DSS must contain a learning process. We have accordingly designed our system to record the data of the completed developments. These data serve for the new evaluations. At the end of each project, the learning module examines to what extent the already-recorded information must be updated. Thus our system combines statistic data and knowledge-based reasonings.