Development of a Portable Tomato Ripeness Measuring Instrument Based on Spectroscopy and Machine Learning Models Dimas Firmanda Al Riza[1]*, Rifqi Fadhlurrohman[2], Setiyaki Aruma Nandi[1], Marchella Alifian Amin Lestiawan[3], Rifda Alfia Safina[3], Ahmad Gibran M.[3], Nadya Agustin Nur Faizah[3]
1 Department of Biosystems Engineering, Faculty of Agricultural Technology, Universitas Brawijaya, Malang, Indonesia 65145
2 Department of Agroindustrial Technology, Faculty of Agricultural Technology, Universitas Brawijaya, Malang, Indonesia 65145
3 Department of Food Science and Biotechnology, Faculty of Agricultural Technology, Universitas Brawijaya, Malang, Indonesia 65145
*Corresponding author: dimasfirmanda[at]ub.ac.id
Abstract
Determining fruit ripeness is an important factor that plays a role in the distribution of fruit commodities because it affects the quality and selling value of the product. In tomato plants, fruit ripeness only relies on the external appearance of the fruit seen by the eye, this method produces subjective results and requires a lot of resources and time to evaluate tomatoes on a large scale. Spectroscopy technology integrated with machine learning can be used to predict fruit ripeness based on factors that change during the ripening process such as brix content, acidity, and firmness. This study took data from these 3 factors non-destructively and then processed them using several machine learning models. The Random Forest Regression model showed the best results where the brix model had an R square value of 0.147 with a Mean Square Error (MSE) value of 20,192. Then, the firmness model had an R square value of 0.678 and an MSE of 1,748. Then, the acidity model had an R square value of 0.817 and 0.004 for the MSE value.
