Cuantificación inteligente de adulteraciones en cacao en polvo mediante espectroscopía NIR
- Pérez Calle, José Luis
- Nur Millatina, N. R. 2
- Barea-Sepulveda, M. 1
- Ferreiro-González, M. 1
- Setyaningsih, W. 2
- Palma, M. 1
-
1
Universidad de Cádiz
info
-
2
Gadjah Mada University
info
- Fernández-Daza Álvarez, Carmen (coord.)
Publisher: Centro Universitario Santa Ana
ISBN: 84-7930-113-9
Year of publication: 2024
Pages: 222-237
Congress: Congreso Agroalimentario de Extremadura (5. 2023. Almendralejo)
Type: Conference paper
Abstract
The adulteration of cocoa powder is a common problemin the food industry, which has been aggravated due to theincreasing and high demand for this product. Therefore, thepresent study aims to develop a methodology using nearinfrared spectroscopy (NIRs) coupled with machine learning(ML) algorithms to quantify adulterations in cocoa powder ina reliable and automated way. For this purpose, pure cocoasamples were adulterated at different percentages in the rangeof 0.5-40% using various substituents. In the exploratoryanalysis, it was observed that the samples tended to clusterbased on both the type and percentage of adulterant used. Thesupervised analysis allowed the creation of ML models withvery satisfactory performances for both quantification andadulterant detention. The best results were obtained for therandom forest (RF) model and the support vector machines(SVM) which correctly identified 100% of the adulterants. In thequantification, the best result was obtained for support vectorregression (SVR) model with a coefficient of determination(R2) higher than 0.99 and a root mean square error (RMSE)lower than 1. The results indicate that the models are highlyaccurate. As a result, a web application has been developedto make these models accessible to users, which facilitates thecharacterization of the samples.