Enhancing input value selection in parametric software cost estimation models through second level cost drivers

Abstract

Parametric cost estimation models are widely used effort prediction tools for software development projects. These models are based on mathematical models that use as inputs specific values for relevant cost drivers. The selection of these inputs is, in many cases, driven by public prescriptive rules that determine the selection of the values. Nonetheless, such selection may in some cases be restrictive and somewhat contradictory with empirical evidence, in other cases the selection procedure is somewhat subject to ambiguity. This paper presents an approach to improve the quality of the selection of adequate cost driver values in parametric models through a process of adjustment to bodies of empirical evidence. The approach has two essential elements. Firstly, it proceeds by analyzing the diverse factors potentially affecting the values a cost driver input might adopt for a given project. And secondly, an aggregation mechanism device for the selection of input variables based on existing data is explicitly devised. This paper describes the rationale for the overall approach and provides evidence of its appropriateness through a concrete empirical study that analyses the COCOMO II DOCU cost driver.