Multi-parameter Regression of Photovoltaic Systems using Selection of Variables with the Method: Recursive Feature Elimination for Ridge, Lasso and Bayes

Cruz J.; Mamani W.; Romero C.; Pineda F. · Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (2020) · DOI: 10.1007/978-3-030-64580-9_16

Resumen

The research focuses on the application of regularization techniques in a multiparameter linear regression model to predict the DC voltage levels of a photovoltaic system from 14 variables. Two predictions were made, in the first prediction, all the variables were taken, 14 independent variable and one dependent variable; Shrinkage Regularization types were applied, as a variable selection method. In the second prediction we propose the use of semiautomatic methods, we used Recursive Feature Elimination (RFE) as a variable selection method and to obtained results. We applied the following Shrinkage regularization methods: Lasso, Ridge and Bayesian Ridge. The results were validated demonstrating: linearity, normality of error terms, non-self-correlation and homoscedasticity. In all cases the precision obtained is greater than 91.99%.

Instituciones: Universidad Nacional del Altiplano