Integrations distinct sources databases to improve the estimation models
by Francisco Valdés-Souto
Often organizations have problems in initiating improvement programs, in metric-based estimations. One issue is the lack of historical IT project data. Otherwise, if data exists, the numbers of projects is not statistically sufficient.
Many studies, that have developed estimation models, are based on databases that are not always available. Even if they are available, they do not always represent the behavior of the organization that is implementing the estimation improvement program.
In this conference, a solution that we have applied to generate reference databases with a greater amount of data is presented. This integrates different databases, including that of ISBSG, as long as statistical assumptions are met to guarantee applicability, and improvement results.
This technique has been consistently applied in Mexican industry to generate initial databases that serve organizations to cover the lack of data, which has allowed us to generate consistent and statistically significant estimation models.
Francisco is the founder of the Mexican software metrics association – Spingere and the president of the Common Software Measurement Consortium (COSMIC). He is a UNAM Researcher Specialist in Measurement, Estimation and Formal Evaluation of Software Projects.