UV-Induced Fluorescence Spectroscopy for Non-Destructive Prediction of Avocado Oil Content
Rut Juniar Nainggolan*, Dimas Firmanda Al Riza, Yusuf Hendrawan

Department of Biosystem Engineering, Faculty of Agricultural Technology, Universitas Brawijaya, Malang, 65145, Indonesia
*Email: niar.1906.rut[at]gmail.com

Abstract

Avocado (Persea americana Mill.) oil content is a primary indicator of maturity and commercial quality, and rapid non-destructive assessment supports harvest scheduling, yield estimation, and quality control. This study evaluates ultraviolet induced fluorescence spectroscopy to predict oil content from peel measurements under laboratory conditions. Ninety-six fruits spanning unripe, ripe, and overripe stages were analysed- peel fluorescence spectra were recorded under 360 and 380 nm excitation, and reference oil content was determined by Soxhlet extraction with hexane. Three regressors, namely ridge regression with L2 regularisation, partial least squares regression (PLSR), and an artificial neural network (ANN), were tuned by five-fold cross-validation and tested on a held out set without spectral preprocessing. Performance varied with excitation and was consistently led by ridge regression. At 360 nm, ridge achieved test set R-squared 0.718 with root mean squared error 5.76 percent by weight, outperforming PLSR at 0.699 and 5.94 and the ANN at 0.484 and 7.78. At 380 NM, ridge again performed best with R-squared 0.720 and root mean squared error 5.73, with PLSR close behind at 0.711 and 5.81 and the ANN lower at 0.498 and 7.66. These patterns indicate predominantly linear relations favouring L2 regularised linear modelling- PLAR provided a stable, interpretable baseline, whereas the ANN did not surpass the linear methods. Overall, the results support ultraviolet induced fluorescence modelling as a reliable, non-destructive tool for rapid screening and maturity informed quality management, with potential for portable or in line implementation.

Keywords: Fluorescence spectroscopy- machine learning- non-destructive measurements- oil content

Topic: Agricultural and bioprocess engineering