DISCRIMINANT ANALYSIS AS A VALIDATION CRITERIA FOR CLUSTERING IN TEXT MINING

Abstract

This article was chosen to choose, initially, three texts from websites that addressed the topic and the aggregation of them materialized the process of Discovery of Textual Knowledge. Under this guidance, pre-processing was carried out, a set of procedures to extract and retrieve textual data considered relevant, a step that allowed processing from the perspective of Quadratic Discriminant Analysis (QDA) using the equidistant approach proposed by covariance adjusted, in which 30 were provided for paragraphs dichotomized into 20 and 10. The accuracy of the model showed a discriminatory probability of 87 in 100, with the first group showing 17 hits and the second nine, recognizing appropriate keywords from the semantic perspective of Nutrigenomics and Nutrigenetics, results expected by experts who seek conglomerates textual cognitive.