Avaliação do uso da biclusterização na seleção de características em classificadores monorótulo

Abstract

Data mining is a widely used technique for exploring large datasets in search of patterns and relationships between variables. This study focuses on the effectiveness of biclustering algorithms in attribute selection, a crucial step in the data mining process. Biclustering differs from traditional clustering by grouping data simultaneously across both rows and columns, uncovering more complex patterns in two-dimensional data matrices. The study evaluates the applicability of the Biclustering method on different datasets, comparing its results with conventional techniques such as J48 and Support Vector Machine (SVM). The results show that despite a significant reduction in the number of attributes, biclustering maintains strong predictive power, emerging as a promising technique for attribute selection and dimensionality reduction.