Application of Supervised Machine Learning Regression Algorithm to Prediction of Dielectric Properties of PPy/Kufeki Stone Composites for Energy Implementations

Authors : Önder EYECİOGLU, Yaşar KARABUL, Mehmet KILIÇ, Zeynep GÜVEN ÖZDEMİR

Pages : 235-254

Doi:10.35378/gujs.810948

View : 22 | Download : 6

Publication Date : 2022-03-01

Article Type : Research Paper

Abstract :The present study deals with the application of the supervised machine learning regression algorithms known as Linear Regression insert ignore into journalissuearticles values(LR);, Support Vector Machine insert ignore into journalissuearticles values(SVM);, and Gaussian process regression insert ignore into journalissuearticles values(GPR); to the frequency and temperature-dependent dielectric parameters of polymer/inorganic film composites. The frequency and temperature-dependent experimental data set of the dielectric parameters insert ignore into journalissuearticles values(ε^` and ε^``); of Polypyrrole/Kufeki Stone insert ignore into journalissuearticles values(PPy/KS); has been utilized. ML models were compared based on their model performance and the most suitable was chosen. After choosing the most suitable ML model, at first, the predictions of the same dielectric parameters of the same samples for different temperatures have been made. Then, the predictions of temperature and frequency-dependent ε^` and ε^`` have been performed for the new PPy based composites consisting of different KS additives that were not produced experimentally. As a result of machine learning, the saturation for KS reinforcing material weight % for dielectric parameters has been determined for capacitor applications. In the light of experimental data and the estimations made by the GPR algorithm, some specific KS additive percentage, working temperature, and frequency ranges have been suggested for the capacitor applications of PPy. 
Keywords : Machine learning, Supervised regression algorithms, Gaussian process regression, Dielectric parameters