Kernel Machine Learning Techniques Modeling for Enhanced Prediction of Oil Flux and Separation Efficiency in Oily Wastewater Treatment Using Ceramic Membrane

This study delves into the exploration of Kernel Support Vector Regression (KSVR) models for predicting oil flux and separation efficiency in wastewater treatment. The research underscores the KSVR-S.E model is notably superior, exhibiting an R2 value of 0.989739, elucidating nearly 99% of the variance in training data. A rigorous evaluation reveals the model's exceptional predictive precision and robustness, supported by a high NSE value of 0.988929 and a remarkable PCC of 0.995402. In contrast, the KSVR-Flux, while commendable, manifests lower efficacy with an
$\mathrm{R}^{2}$
value around 0.93. Transitioning to the testing phase, both models exhibit slight performance diminutions, yet maintain substantial predictive accuracies. Notably, KSVR-S.E showcases a resilient performance with persistent high values in
$\mathrm{R}^{2}$
and PCC, emphasizing its robust generalizability and predictive consistency across various datasets. Comprehensive evaluations, encompassing boxplots and probability distribution plots, elucidate the models' proficiency, indicating a heightened prediction accuracy and reduced variability, particularly in separation efficiency predictions by KSVR-S.E. From the quantitative findings, the study affirms the enhanced predictive capacity of the KSVR-S.E model in oil flux and separation efficiency, marking a significant contribution to the optimization of wastewater treatment processes. This research's findings serve as a pivotal reference, encouraging further explorations employing KSVR models in environmental applications for improved predictability and operational efficacy.