Eureka! Institutional Repository

This thesis focuses on the study and comparison of algorithms and techniques for multi-label classification, aiming to understand their performance metrics. The concept of multi-label classification is analyzed, highlighting its importance in processing and analyzing complex data sets. Various approaches are examined, such as problem transformation techniques and specific algorithms for solving such problems. Additionally, the Scikit-multilearn library is presented as a tool for the practical implementation of these techniques. The work includes an experimental study where various algorithms are compared based on real data sets, evaluating their performance on different parameters. Finally, the thesis concludes with findings regarding the applicability of the various techniques and suggests future directions for research in the field of multi-label classification.