Hi all,
I am doing a multi-label classification using machine learning algorithms. In order to evaluate the performance of the multi-label classification, I am using label-based and sample-based performance evaluation metrics listed in Article A Tutorial on Multi-Label Learning
.When I compared both types of approaches to performance evaluation, sample-based metrics are lower in value than the label-based metrics. I mean precision and recall of sample-based very low than the precision and recall of label-based method.
Just wanted to understand that is correct behavior or I am doing something wrong.
I think it is because, label based data have strong relational information of labeling, however., labeling is a bit time consuming for large databases, that is why some researchers avoiding it. See for example https://www.kdnuggets.com/2017/06/acquiring-quality-labeled-training-data.html
Visit the following link for more info
https://www.researchgate.net/publication/283807144_Evaluation_of_Example-Based_Measures_for_Multi-label_Classification_Performance
In addition, you could try out a more diverse process generating improved synthetic samples. Maybe this will help you out: Preprint Multi-Label Sampling based on Local Label Imbalance
- Badges
- Science topic