Geometric SMOTE-Based Approach to Improve the Prediction of Alzheimer's and Parkinson's Diseases for Highly Class-Imbalanced Data
Lokeswari Y. Venkataramana, Shomona Gracia Jacob, Venkatavara Prasad D, R. Athilakshmi, V Priyanka, K. Yeshwanthraa, S. Vigneswaran · Advances in electronic government, digital divide, and regional development book series · 2023
In many applications where classification is needed, class imbalance poses a serious problem. Class imbalance refers to having very few instances under one or more classes while the other classes contain sufficient amount of data. This makes the results of the classification to be biased towards the classes containing many numbers of samples comparatively. One approach to handle this problem is by generating synthetic instances from the minority classes. Geometric synthetic minority oversampling technique (G-SMOTE) is used to generate artificial samples. G-SMOTE generates synthetic samples in the geometric region of the input space, around each selected minority instance. The performance of the classifier is compared after oversampling using G-SMOTE, synthetic minority oversampling technique and without oversampling the minority classes. The work presents empirical results that show around 10% increase in the accuracy of the classifier model when G-SMOTE is used as an oversampling algorithm compared to SMOTE, and around 30% increase in performance over a class imbalanced data.