Towards adaptive soft computing based software effort prediction

Algorithmic effort prediction models are limited by their inability to cope with imprecision present in software projects early in the development life cycle. A critical survey carried out by Satin and Ahmed [17] reveals the lack of adaptive soft computing based effort prediction systems that provide complete transparency to the prediction system building strategies. In addition, efforts to model the imprecision problem in one of the most widely used algorithmic model, COCOMO, have not been appropriate [17]. The components of COCOMO model were addressed independently. Integrating the individual component into a single prediction system remains an open question. In this paper, we present a transparent and adaptive fuzzy logic framework for effort prediction based on COCOMO. The training strategies we have implemented in the framework tolerates imprecision, explains prediction rationale through rules, incorporates experts knowledge, and could adapt to a new environment as new data becomes available. Validation of the framework potentials has been carried out on artificial datasets and the COCOMO public database of completed projects.