Metalearning Applied to Multi-label Text Classification

Resumo

Data Mining and Machine Learning fields have many techniques that can support data analysts in the text classification task. However, finding the most adequate techniques require advanced technical knowledge, exhaustive computational experiments and, consequently, time. To address this issue, researchers have proposed different approaches for selecting such techniques to be employed in classification tasks and the dynamic selection of classifiers is one of them. Therefore, this work proposes an approach that uses metalearning to automate the process of selecting the best classifier for each instance of a given multi-label textual dataset. Experiments were performed with multi-label text datasets and showed that the proposed approach is promising.