Closed sachag678 closed 7 years ago
Implementing SMOTE instead.
Implementing undersampling of majority class since, SMOTE requires adding new synthetic data into a sequence which messes up the sequence. Need to find a way to make new synthetic sequence data.
Undersampling on the majority class performs better than the initial sampling algorithm that was implemented.
Implemented Random undersampling which performs better than just removing cases in succession.
Complete
A Novel Similarity Based Under-sampling of Imbalanced Datasets by Maira Anis