Attribute selection and imbalanced data for detection of chronic kidney disease
Keywords:
Imbalanced data, medical data analysis, traditional classifiers, CKD, feature selection, SMOTE algorithm, BFO-WD, KSVM) classifier.Abstract
In medical data analysis, imbalanced data is considered as a significant issue, because its class labels don’t have a balanced
medical datasets. This behaviour affects the conventional classifiers. SO, they focus on optimizing the overall accuracy
without considering the relative distribution of every class. With the help of various techniques, a database with several
attributes of healthy subjects and subjects with CKD were examined. If the samples become imbalanced and it contains
unimportant attributes, then the accuracy of the classifier will be minimized. A new feature selection and imbalanced
dataset handling algorithm is suggested to rectify this issue and to maximize the accuracy or disease prediction results of
the classifier. At first, enhanced SMOTE algorithm helps to rectify the imbalanced dataset. Through Bacterial Foraging
Optimization with Weighted Difference (BFO-WD), the significant features in the dataset were chosen. Kernel Support
Vector Machine (KSVM) classifier helps to perform the classification process, in order to recognize the dominant attributes
in the early detection of CKD.
