Building a software cost estimation model based on categorical data

Abstract

The paper explores the possibility of generating a multi-organisational software cost estimation model by analysing the software cost data collected by the International Software Benchmarking Standards Group. This database contains data about recently developed projects characterised mostly by attributes of categorical nature such as the project business area, organisation type, application domain and usage of certain tools or methods. The generation of the model is based on a statistical technique which has been proposed as alternative to the standard regression approach, namely the categorical regression or regression with optimal scaling. This technique starts with the quantification of the qualitative attributes (expressed either on nominal or ordinal scale), that appear frequently within such data, and proceeds by using the obtained scores as independent variables of a regression model. The generated model is validated by measuring certain indicators of accuracy.