Efficiency Analysis of Colombian Higher Education Institutions in Scientific Production using Data Envelopment Analysis and Bayesian Networks

Data Envelopment Analysis (DEA) is a nonparametric tool based on linear programming elements trying to capture the relative eciency of a set of decision making units (DMU), it has been applied in various fields of knowledge, especially in the area of scientometrics it has been gaining recognition and acceptance. For this work, we apply two dierent extensions of the DEA calls: CCRO and Cross Eciency. However, the relative eciencies from the DEA only give a measure that reveals the multiple relationships between dierent variables of bibliographic production (Articles, Book Chapters, Conferences, Books, Software, Working Papers) within each of the areas of knowledge, that’s why it’s necessary to establish a connection between the results from the DEA and the specifics of production within each of them, as some previous studies tend to homogenize the behavior of all, but the dynamics and characteristics are dierent from each other. That’s why the DEA results are analyzed using Bayesian networks. A Bayesian network is a directed acyclic graph composed of a set of nodes, one of edges and a probability distribution or a family of them. The use of Bayesian networks allow to identify the particularities of each of the areas of knowledge in terms of scientific production