Machine learning-based classification for predicting the power flow of surface plasmon polaritons in nanoplasmonic coupler

Using a combination of the finite element method (FEM) applied in COMSOL Multiphysics and the machine learning (ML)-based classification models, a computational tool has been developed to predict the appropriate amount of power flow in a plasmonic structure. As a plasmonic coupler, a proposed structure formed of an annular configuration with teeth-shaped internal corrugations and a center nanowire is presented. The following representative data mining techniques: standalone J48 decision tree, support vector machine (SVM), Hoeffding tree, and Naïve Bayes are systematically used. First, a FEM is used to obtain power flow data by taking into consideration a geometrical dimensions, involving a nanowire radius, tooth profile, and nanoslit width. Then, we use them as inputs to learn about machine how to predicate the appropriate power flow without needing FEM of COMSOL, this will reduce financial consumption, time and effort. Therefore, we will determine the optimum approach for predicting the power flow of the proposed structure in this work based on the confusion matrix. It is envisaged that these predictions’ results will be important for future optoelectronic devices for extraordinary optical transmission (EOT).