I have a labeled dataset with data under 3 classes. Some of the instances in the dataset can be similar although they may belong to different classes. So I tried clustering the dataset without considering the labels. Once clustering was done, the already known labels were assigned to instances in each cluster.This way I will have K clusters and each cluster has instances belonging to different class. I wanted to know if it makes sense to apply supervised learning algorithms like LDA, QDA, SVM,kNN etc. on each of these clusters separately and identify the best performing classifier for each cluster?
Hi, there are studies that use space partitioning and apply classification for each partition separately, and yours also makes sense. However a standard ensemble method may also perform well without the initial step, so it is best to try both approaches.
http://kilyos.ee.bilkent.edu.tr/~kozat/papers/40_1.pdf
Article Local Supervised Learning through Space Partitioning
It is possible, but you won't have better performance, as you will add error and bias twice for your two-step methods.Also you will want your classifier to generalize well beyond a particular cluster to be reliable on your entire dataset, I will use ensemble in this case!
Best regards
I am getting a +6% higher accuracy when I employ this scheme rather than the general supervised model.
Leader algorithm with KNN can find efficient classifier
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence