Data-Driven Methods to Create Knowledge Maps for Decision Making in Academic Contexts

Abstract

Knowledge is the primary asset of today’s organisations; thus, knowledge management has been focused on discovery, representation, modification, transformation, and creation of knowledge within an enterprise. A knowledge map is a knowledge management tool that makes organisational processes more visible, feasible, and practicable. It is a graphical representation of decision-related information. What happens, how various events can be managed, and why they happened: all can be demonstrated very precisely by a well-designed knowledge map. There are diverse knowledge-related roles; for example, each university dean’s office — as an instance of a knowledge-based organisation — usually relies upon their institutional memory to make daily decisions. However, utilising a knowledge map greatly facilitates any individual’s or group’s decision-making process, by proposing or establishing key required information. In this study, two important managerial roles — Associate Deans of Research and Education — were selected; then we reviewed their key managerial decisions and proposed three different techniques for supporting their decisions. The chief superiority of the approach offered here was in the creation of role-based knowledge maps, including an expertness map and a collaboration map for the Associate Dean of Research, which were formed using clustering, taxonomy formation, and information retrieval methods. A third map was created for the Associate Dean of Education, including a Bayesian reasoning map based on an Improved PC (IPC) algorithm, which learned the structure and the parameters of a Bayesian network to describe decision-making in the domain of education. To evaluate the proposed approaches, structural and functional evaluation measures and standard datasets (in the available cases) were chosen. The results found that the approaches were comparable to the selected benchmarks within the real data; even after considering the challenging nature of the real data, which included problems such as incomplete and unclean data extracted from the University of Tehran’s education and research management information systems.