An Improved RUS-Based Sampling for Classifying Imbalance Water Quality Data

Machine learning models, when trained on imbalanced data, face challenges in effectively identifying minority data points. To mitigate this problem, an approach involves resampling techniques is widely used to rebalance the class distribution. This study investigates the impact of proposed RACOG-RUS resampling technique on the performance of four classifiers: Support Vector Machine, Artificial Neural Networks, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$k$</tex>-Nearest Neighbors and Naïve Bayes in classifying imbalanced water quality data. The RACOG-RUS technique was compared with other resampling methods of Random Under Sampling (RUS), Random Oversampling (ROS), Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic Sampling (ADASYN), Rapidly Converging Gibbs Sampler (RACOG), and three RUS-based hybrid sampling such as ROS-RUS, RUS-SMOTE, RSYN. Results revealed that all classifiers showed the best performance using RACOG-RUS technique except for naïve bayes classifier which achieved the best performance using ROS-RUS method. Support vector machine exhibited superior performance (accuracy <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=0.9337, B A=0.9026$</tex>, sensitivity <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=0.8500$</tex>, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$f$</tex> measure <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=0.8395$</tex>) compared to the other classifiers when utilizing the RACOG-RUS resampling method in improving imbalance class. The findings presented within this paper provide valuable perspectives that can prove beneficial for forthcoming research aimed at tackling class imbalance challenges and predictive modeling in diverse domains.