MAGIC: A Hybrid Modeling Approach for Optimizing Inspection Cost-Effectiveness.

Software inspections are a key technology for achieving quality at a reasonable cost. Yet, inspections do not always result in substantial benefits, as they are affected by a number of technical and managerial factors. Therefore, measurement needs to be applied in an organization in order to assess the economic impact of inspections to understand the factors influencing it and to optimize these factors accordingly.
This thesis presents a hybrid modeling approach for optimizing inspection cost-effectiveness. First it proposes and evaluates a refined cost-effectiveness model and a corresponding assessment procedure combining expert opinion and project data. Second it proposes and evaluates the data-mining technique MARS for building effectiveness prediction models as well as an approach for deriving planning guidelines from these models.
The validation of the approach shows; (i) the proposed cost-effectiveness model is more accurate than existing models, (ii) using expert opinion for assessing cost-effectiveness is a practical and feasible approach, (iii) MARS models are more accurate than regression models, and (iv) the derived planning guidelines are plausible.