|
I am working on an imbalanced dataset with only 20% as positive class, and the rest belong to the negative class. I have tried using typical techniques for dealing with skewed datasets: undersample, oversample, Ensemble learning algorithms such as Random Forests, boosting algorithms, so on and so forth. However, my results have plateaued with 0.8 AUC. I'm hoping for Feature Extraction methods or some sort of mapping the dataset which could in turn ameliorate the classification results, given that I only have the number of features ranging from 31 to 60, and samples ranging from 3000 to 10,000. So what do you suggest ? Thanks in advance..! |
