Implementasi bayessian regularization neural network pada model prediksi beban listrik jangka pendek di kota Langsa

Abstract

The accuracy level of the artificial neural network method in electricity load prediction models remains a focus for research. One area for improvement in artificial neural network electricity load prediction models is an occurrence potentially of overfitting. Overfitting occurs when the artificial neural network topology fails to produce the desired output values. Overfitting in the neural network topology affects the accuracy of the electricity load prediction model. This research aims to develop a Bayesian regularization algorithm for neural network topology to create an electricity load prediction model for the city of Langsa. The research results indicated that the Bayesian regularization algorithm is effective in minimizing the potential for overfitting in the neural network topology. Bayessian regularization algorithm had MAPE value is 2.45 %.