A Rule-Based Approach to Developing Software Development Prediction Models

• Authors: Prodromos D. Chatzoglou;Linda A. Macaulay
• Venue: Automated Software Engineering
• Year: 1998

Managers of software development projects increasinglyrecognize the importance of planning and estimation and nowhave many sophisticated tools at their disposal. Despitethis many systems are still delivered way behind schedule,cost far more to produce than original budget estimates andfail to meet user requirements.It is the contention of the authors that many existing toolsare inadequate because they fail to embrace the significantbody of knowledge accumulated by past and present projectmanagers.This paper presents a new approach to planning which enablesproject managers to learn from the experience of others. Theauthors have adopted a bottom-up approach to planning whichgoes from the specific (planning the requirements captureand analysis process—RCA) to the general (planning thewhole development process). A model, called MARCS, wasconstructed to give predictions of the resources (time,effort, cost, people) needed for the completion of andoutcomes of the RCA process. Based on the predictions aboutthe RCA process, the model then attempts to predict theresources and outcomes of the whole development process.MARCS is a combination of rule-based models and its mainadvantage is that it incorporates both qualitative andquantitative factors that can be easily identified andmeasured in the beginning of the development process.Empirical data concerning 107 projects developed by morethan 70 organizations within UK, gathered through a two-stage mail survey was used for the construction andvalidation of the MARCS planning model.