An Effort Prediction Model for Data-Centred Fourth-Generation-Language Software Development

Accurate effort prediction is often an important factor for successful software development. However, the diversity of software development tools observed today has resulted in a situation where existing effort prediction models’ applicability appears to be limited. Data-centred fourth-generation-language (4GL) software development provides one such difficulty. This paper aims to construct an accurate effort prediction model for data-centred 4GL development where a specific tool suite is used. Using historical data collected from 17 systems developed in the target environment, several linear regression models are constructed and evaluated in terms of two commonly used prediction accuracy measures, namely the mean magnitude of relative error (MMRE) and pred measures. In addition, R, the maximum value of MRE, and statistics of the absolute residuals are used for comparing the models. The results show that models consisting of specification-based software size metrics, which were derived from Entity Relationship Diagrams (ERDs) and Function Hierarchy Diagrams (FHDs), achieve good prediction accuracy in the target environment. The models’ good effort prediction ability is particularly beneficial because specification-based metrics usually become available at an early stage of development. This paper also investigates the effect of developers’ productivity on effort prediction and has found that inclusion of productivity improves the models’ prediction accuracy further. However, additional studies will be required in order to establish the best productivity inclusive effort prediction model.