Is a Data-Driven Approach still Better than Random Choice with Naive Bayes classifiers?
This work addresses the problem of label space partitioning for multi-label classification, providing insights for researchers and practitioners, but it is incremental as it builds on prior comparisons with tree classifiers.
The study compared data-driven, a priori, and random approaches for label space partitioning in multi-label classification using Gaussian Naive Bayes, finding that data-driven methods significantly outperform random baselines on average across 12 benchmark datasets, with advantages in worst-case scenarios for F1 scores and Subset Accuracy.
We study the performance of data-driven, a priori and random approaches to label space partitioning for multi-label classification with a Gaussian Naive Bayes classifier. Experiments were performed on 12 benchmark data sets and evaluated on 5 established measures of classification quality: micro and macro averaged F1 score, Subset Accuracy and Hamming loss. Data-driven methods are significantly better than an average run of the random baseline. In case of F1 scores and Subset Accuracy - data driven approaches were more likely to perform better than random approaches than otherwise in the worst case. There always exists a method that performs better than a priori methods in the worst case. The advantage of data-driven methods against a priori methods with a weak classifier is lesser than when tree classifiers are used.