Modelo predictivo para la producción de azúcar usando algoritmos de Machine learning.

Abstract

Abstract: The current paper presents the development and implementation of a predictive model based on artificial intelligence to estimate the daily sugar production of Ingenio Azucarero Monterrey C.A. Historical daily sugar production data were used, followed by statistical techniques for data cleaning and preparation. The CRISP-DM methodology was applied where 4 models were trained and evaluated: Linear Regression, Generalized Additive Models (GAM), Random Forest and Artificial Neural Networks. For the evaluations, metrics such as: mean square error (MSE), root mean square error (RMSE) and coefficient of determination (R²) were used. Following the results, GAM gave a MSE error equal to a very low RMSE and an excellent accuracy value. An interactive web application was implemented on the Streamlit platform, which allows users to enter data and obtain sugar production predictions. The study has demonstrated the usefulness of applying artificial intelligence techniques in this industrial sector of our country.