Correlation-Based Feature Grouping with Decision Tree for Classifying High-Dimentional Imbalanced Data

Abstract

Classifying high-dimensional imbalanced data is a big challenge in mining real-world big data. Existing algorithms are classifying the majority class instances and get the maximum classification accuracy and minority class instance is overpowered by getting misclassified. In real life applications minority class instances are more significant than the majority class. For classifying imbalanced data sets few techniques based on sampling (Under-sampling / over-sampling), cost sensitive learning methods and ensemble learning are used. In our research, A new technique has been introduced, \correlation-based feature grouping with decision tree for classifying high-dimensional imbalanced data". We have assessed the dispatch of the the proposed algorithm on few of the high dimensional imbalanced data sets with different imbalance correspondences. The results are tremendously better to work with high imbalanced data sets.