This repository shows an example of comparisons of harmonization and oversampling methods via different machine learning classificers. In the imbalanced heterogenetic multicentric context, 4 harmonization and 5 oversampling methods can be compared via nested k-flod cross validation using this framework.Especially, the harmozation process was imporved to investigate and evaluate its generalizeability with the pypothesis that the batch effect in the testing data is same with that of training data.
ACKNOWLEDGEMENTS
We highly appreciated the software packages developed by Fortin et al. https://github.com/Jfortin1/ComBatHarmonization (ComBat), Renard et al. http://sites.uclouvain.be/absil/2017.01 (normFact), Inoue et al. https://github.com/minoue-xx/Oversampling-Imbalanced-Data (Oversampling) and Gu et al. https://onlinelibrary.wiley.com/doi/full/10.1002/int.22230.
For any queries about the codes, please contact Dongyang Du and Lijun Lu (ljlubme@gmail.com).
Programmed by Dongyang Du (dongyang330@gmail.com)