Performance Analysis of Ensemble Technique for Classifying Imbalanced Dataset Using SMOTE-TOMEK Links Sampling

Abstract

This century has seen a substantial increase in the value of data. Many decisions are made based on data. More the data, more possibility to get accurate result. But in many cases, available datasets are imbalanced i.e. one class have more data and other class have few data which leads to inaccurate classification while using machine learning. So, this paper focuses to minimize this problem. In this paper, first the imbalanced data are sampled using SMOTE-TOMEK Links sampling which balances the imbalance dataset. Then, Ensemble Technique is used for data classification. In ensemble technique, XGBoost and Random Forest algorithm are used as base classifier and Logistic Regression algorithm is used as meta classifier using Stacking Classifier for final prediction. The performance of this technique is measured using different evaluation metrics, precision, recall, f1-score and ROC (Receiver Operating Characteristics) Curve.