Predictive Modeling of Nanoscale Junctionless FinFET Using XGBoost

In recent years, computing performance has been increased and new Machine Learning (ML) algorithms have been developed for fast and accurate prediction. With increasing demand for quick estimation of semiconductor device parameters during the early stages of development, prediction using ML is essential for new technology. Tree boosting type of algorithms are highly efficient among the regression algorithms where dataset is comparatively small and nature of the dataset does not support linear regression. Dataset generated through TCAD simulation is highly time consuming and internal device physics supports non-linear nature in the dataset produced. In this paper, machine learning model using Extreme Gradient Boosting (XGBoost) algorithm has been trained with a dataset of semiconductor device, generated from TCAD simulation and is used for device performance prediction. This fast and accurate prediction method will certainly be a real alternative to TCAD in the semiconductor industries.