A Data-Augmented CGO-GAM Prediction Method for Plasma Hydrophilic Modification

Polymer materials often require surface modification to address inherent structural limitations and enhance performance. This study presents a data-augmented chaotic game optimization- generalized additive model (CGO-GAM) designed to predict plasma hydrophilic modification, achieving high accuracy even with a limited number of data samples. The data augmentation method utilizing Generative Adversarial Networks (GAN) enhances the balance of modified data, particularly by supplementing hydrophilicity data in instances where treatment effects are suboptimal and data distribution is limited. The adaptive GAM algorithm effectively manages the nonlinear relationships among discharge parameters, gas gaps, gas compositions, and the effects of nanosecond-pulse dielectric barrier discharge (DBD) surface modifications. CGO-generated random perturbations, refined through mutation and crossover operations, enhance population diversity, effectively mitigating issues of early convergence and overfitting while improving the balance between global and local search strategies. The proposed model achieves an accuracy of 90% in predicting hydrophilic modification performance, demonstrating superiority over traditional methods such as Support Vector Machines (SVM) and General Regression Neural Networks (GRNN).