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

Published: 2 April 2024| Version 1 | DOI: 10.17632/p6ytjt5rfy.1
Contributors:
,
,
,
,

Description

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.

Files

Steps to reproduce

All results may be obtained by executing the Github code in the "Related links" section (instructions are provided in the repository).

Institutions

Universitat d'Alacant

Categories

Machine Learning, Pattern Recognition

Licence