Abstract:
This article presents a new way to enhance the prediction of software attributes in the first stages of software life cycle through the study of the relationships among the elements of models used in different development phases. A system to predict the cost of verifying the design derived from data flow diagrams (DFD) was defined using this approach in structured analysis and merise (SAM) a formalized methodology. The construction and validation of measures for featuring design attributes is based on a solid theoretical analysis of the properties that should be found in the measured object. Prediction validity is achieved as a logical consequence of development relationships between and among the involved models. Moreover, some empirical data is included to enable observation of the application of the system. The article concludes that this approach can also lead to prediction enhancement through process improvement in other methodologies like SSADM or the object-oriented (OO) ones.