Knowledge Discovery in Data Bases: a Case Study in a Private Institution of Higher Education
Carla Regina Mazia Rosa, M.T.A. Steiner, Pedro José Steiner Neto · IEEE Latin America Transactions · 2018
This paper aims to present a methodology to discover knowledge in databases (Knowledge Discovery in Databases; KDD) which can be used in several areas in order to classify new instances. This methodology detects what are the most important variables (attributes, information) and realized the classification of new instances in an automatized way, maximizing its accuracy. Its application is shown in courses of a private institution of higher education in order to verify the students' satisfaction related to the quality and to the services offered. Based on the KDD process, it was initially realized a data exploratory analysis and, after that, it was applied three Data Mining techniques: Logistic Regression Binary (LRB), Generation of Surface that minimizes errors through a Linear Programming mathematical model (GSME-PL) and Fisher Discriminant Linear Function (FDLF). It was analyzed 885 instances, with 12 variables and an output (satisfaction). Through the results obtained, it can be concluded that the attributes relating to "teachers" are the most important, and for the case addressed, RLB was the technique with the highest accuracy rate (92.2%).