Genetic algorithm-based threshold voltage prediction of SOI JLT using multi-variable nonlinear regression

A machine learning approach of multivariable non-linear regression with six input variables is used for the first time to assess the process variability impact on the threshold voltage (V
_th
) of silicon-on-insulator (SOI) junctionless transistor (JLT). The impact of source/drain electrode spacing (L
_SD
) is considered as one of the input variables in this analysis. The genetic algorithm is implemented in MATLAB and used for the optimization of the hypothesis. A dataset of 2000 samples is used for training and testing. More than 75% of the test samples offered a prediction error in the range of ±0.0125V despite nonlinearity caused by six input parameter variations. Average prediction accuracy of over 96 percent is achieved with efficient algorithm training and well-organized dataset.