This paper presents a comparative evaluation of various machine learning (ML) models for efficient Metal-Oxide-Semiconductor Field-Effect Transistor (MOSFET) device simulation. We investigated four popular regression models, including Extreme Gradient Boosting (XGBoost), Artificial Neural Networks (ANN), Polynomial Regression, and Random Forest Regressor (RFR) to predict nonlinear I-V behavior of MOSFET device. The data-driven nature of these approaches enables them to act as efficient alternative for traditional Technology Computer-Aided Design (TCAD) based modeling, dramatically lowering simulation time and computational requirements. After training with a large amount of I-V data generated through TCAD simulations, each of the models were evaluated using the coefficient of determination (R2) and root-mean-square error (RMSE). XGBoost achieved the highest predictive accuracy (R2 = 0.9999), demonstrated smooth convergence throughout the training period. The smooth training assisted XGBoost generalizing better. Thus, XGBoost outperformed ANN (R2= 0.9880), Polynomial regressor model (R2=0.9773), and RFR(R2 =0.9320). Therefore, XGBoost captures the overall trend of the data better among these four models. Based on these findings, the XGBoost model proves to be more efficient in predicting the I-V characteristics of MOSFET device.
