Detailed results of "Insights into imbalance-aware Multilabel Prototype Generation mechanisms for k-Nearest Neighbor classification in noisy scenarios"

Resumo

Detailed experimental results of the different Prototype Generation strategies for k-Nearest Neighbour classification in multilabel data attending to the particular issues of label-level imbalance and noise: 1. Noise-free scenarios - Study of the considered strategies for addressing label-level imbalance in PG scenarios without induced noise. - Individual results provided for each corpus. - Statistical tests (Friedman and Bonferroni-Dunn with significance level of p < 0.01) to assess the improvement compared to the base multilabel PG strategies - Corresponds to Section 5.1 in the manuscript. 2. Noisy scenarios - Study of the noise robustness capabilities of the proposed strategies. - Individual results provided for each corpus. - Statistical tests (Friedman and Bonferroni-Dunn with significance level of p < 0.01) to assess the improvement compared too the base multilabel PG strategies - Corresponds to Section 5.2 in the manuscript. 3. Results ignoring the Editing stage - Assessment of the relevance of the Editing stage in the general pipeline. - Individual results provided for each corpus. - Corresponds to Section 5.3 in the manuscript.