Hyperparameter Optimisation for Improving Classification under Class Imbalance

This is the source code used in the paper below:

Jiawen Kong, Wojtek Kowalczyk, Duc Anh Nguyen, Stefan Menzel and Thomas Bäck, “Hyperparameter Optimisation for Improving Classification under Class Imbalance”, in 2019 IEEE Symposium Series on Computational Intelligence (SSCI), Xiamen, China, 6-9 December 2019, doi: 10.1109/SSCI44817.2019.9002679

Although the class-imbalance classification problem has caught a huge amount 
of attention, hyperparameter optimisation has not been studied in detail in 
this field. Both classification algorithms and resampling techniques involve 
some hyperparameters that can be tuned. This paper sets up several 
experiments and draws the conclusion that, compared to using default 
hyperparameters, applying hyperparameter optimisation for both 
classification algorithms and resampling approaches can produce the best 
results for classifying the imbalanced datasets. Moreover, this paper shows 
that data complexity, especially the overlap between classes, has a big impact 
on the potential improvement that can be achieved through hyperparameter 
optimisation. Results of our experiments also indicate that using resampling 
techniques cannot improve the performance for some complex datasets, which 
further emphasizes the importance of analyzing data complexity before dealing 
with imbalanced datasets.