Performance Improvement of Classification Model with Imbalanced Dataset Classification models based on machine learning for the application of real life carry out classification tasks using real life dataset. Classification models have class imbalance pro

Abstract

Classification models based on machine learning for the application of real life carry out classification tasks using real life dataset. Classification models have class imbalance problems when the dataset is imbalanced in nature. Classification models show biases for performing the classification towards the majority class due to class imbalance issues. The purpose of the study is to examine and control the class imbalance problem using the cluster centroid undersampling technique with the motive of improving the performance of machine learning classification. In order to accomplish the goal, this study performs experimental analysis for examining and controlling the class imbalance problem before and after applying the cluster centroid undersampling technique on imbalanced dataset. The experimental study is performed by using five different imbalanced datasets, cluster centroid undersampling technique and decision tree classification model. The results of this study are promising that supports this study and confirm that the class imbalance problem can be handled using undersampling techniques very effectively with performance improvement of a classifier from 11% to 67%. This study highlights the influences of class imbalance problems on machine learning classification models and experimental results with analysis provide an appropriate conclusion with an improvement in the performance of machine learning based classification models.