Modeling the electrical properties of heterojunctions using ANFIS, ANFIS-GA and ANFIS-PSO Models

The present research introduces a theoretical study that aims to utilize ANFIS in estimating and predicting the electrical behavior of heterojunctions. For this purpose, five different heterojunctions were chosen. The experimental datasets that represent the electrical behavior of the chosen heterojunctions were extracted and employed in ANFIS as targets. To enhance the ANFIS performance two hybrid heuristic algorithms, genetic algorithm (GAs) and particle swarm optimization (PSO) were combined with ANFIS. The major contribution of the current research is to predict the electric characteristics of heterojunctions using ANFIS and increase the modeling accuracy of ANFIS by optimizing the premise and consequent parameters using (GAs) and (PSO). Also, compare the proportion of enhancement produced by using ANFIS-GA and ANFIS-PSO to decide which of them is more powerful under the study conditions. However, to the author's knowledge, the presented goals have not been investigated before for heterojunctions. The mean squared error (MSE), the correlation coefficient (R2), and the standard deviation error (Std. error) were calculated for all trained models. The modeling errors of ANFIS-GA and ANFIS-PSO were compared to the error values produced by ANFIS. According to modeling results, simulation ANFIS outputs follow the experimental data patterns in excellent response. Predictions of electrical characteristics for heterojunctions using the trained models provide acceptable results where the MSE values obtained by training ANFIS-PSO are lower than their values obtained by ANFIS and ANFIS-GA models. The improvements in average percentages in ANFIS performance when combined with GA and PSO are equal to 2.2% and 3%, respectively. Consequently, the proposed ANFIS-PSO model is more accurate in predicting the electrical behavior of heterojunctions under the study conditions.