Comparative analysis of machine learning models for performance evaluation of defective plasmonic grating sensor
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Plasmonic sensors enable accurate refractive index measurements with remarkable sensitivity and efficiency through the use of the Surface Plasmon Resonance (SPR) phenomenon. Plasmonic grating is used to investigate the effect of filtering a broad stop band region, and thus, a localized mode is excited in the stop band by using a defective grating structure. The performance is analyzed using sensitivity to detect shifts in resonance conditions for per-unit changes in the refractive index of the defect region of the grating. The sensitivity of plasmonic grating structures can be precisely modulated by changing the length of the defect. Several machine learning techniques, such as linear regression, polynomial regression, support vector regression (SVR), and K-nearest neighbor (KNN) regression, are examined in order to describe this nonlinear relationship between defect length and sensitivity. Predictive accuracy was evaluated and improved using three dataset preparation techniques: Generative Adversarial Network (GAN)-based data augmentation, polynomial interpolation, and linear interpolation. Upon comparative performance analysis, the polynomial regression model is proved to be the optimal model for sensitivity prediction in defective plasmonic grating sensors. This model is trained using GAN-augmented data, which demonstrated excellent predictive accuracy with a minimum Mean Squared Error (MSE) of 0.2805 and an outstanding R² score of 0.9999.
Fields of Study & Subjects
Plasmonic and Surface Plasmon Research
Advanced Fiber Optic Sensors
Machine Learning in Materials Science
Publication details
- Journal / source
- International Conference on Optical and Wireless Technologies (OWT)
- Pages
- 129–129
- Publication date
- 30 March 2026
- Language
- EN