Predicción del índice de madurez de musa acuminata (banano orito) mediante la aplicación de la inteligencia artificial

Abstract

The production of orito banana (Musa acuminata AA) is a very significant activity that is part of the country's economy, being used as a fruit or as an input in food processing. In this sense, the automation of the fruit classification process contributes to saving time and reducing errors in tasks that involve people. The present research aims to predict the maturity index of the orito banana Musa acuminata AA using artificial intelligence, thus reducing errors that can be caused by human vision to classify this fruit. To achieve the proposed objective, Artificial Intelligence (AI) algorithms were used, in particular Deep Learning such as: tensorflow, keras, Neural Networks, Backpropagation, numpy and the Matplolib pyplob and Scklearn libraries, which were fed with images and results of laboratory physicochemical analyses such as: titratable acidity, soluble solids, texture and pH. A color chart was created with three stages of ripening: immature (IM), mature (MA) and overripe (SM). After processing the information by the algorithms, a test accuracy of 0.94 was obtained; this result was confirmed in the confusion matrices of each algorithm analyzed according to the subject results. It can be deduced that AI algorithms can be used to predict the maturity index of orito banana with high precision.