Mineração de Texto para identificação de termos relevantes em Nutrigenômica sob a ótica fuzzy

Abstract

This dissertation presents the mining of three texts under Nutrigenomics theme, whose objective was to compare the results related to the grouping created by fuzzy non-hierarchical methods in Text Mining. The use of text mining algorithms optmizes the search in accordance with keywords to conceive by them. First of all, the texts were collected from websites about the proposed theme and afterwards them passed for a process of cleaning and treatment a known as pre-processing. After this step, the fuzzy C-Means and fuzzy C-Medoids clustering algorithms, were applied where was observed that despite the number of groups have been defined previously through of the graphical visualization Sperman s Correlation Rank, the fuzzy C-Means constituted fewer clusters than the pre-fixed, while the fuzzy C-Medoids did the suggested. However, in both methods there were clustering overlapping, that is, instersection between two or more groups, fuzzy methods characteristics. In the results validation through Discriminant Analysis, despite of the rates have similar values, the fuzzy C-Medoids was more efficient in classifying groups because its accuracy rate was 94% cons 90% of fuzzy C-Means, so it was get more efficiently group terms that have common characteristics. The goal of the creating keywords that will contribute for the search of the information according the typology of Nutrigenomics and Nutrigenetics was attain what showed intuitive and conceitual differences, considered relevants aspects in systematic studies to define different categories.