DETERMINATION OF QUALITY AND RIPENING STAGES OF ‘PACOVAN’ BANANAS USING VIS-NIR SPECTROSCOPY AND MACHINE LEARNING

Ferreira, Iara J. S.; Almeida, Sarah L. F. de O.; Figueiredo Neto, Acácio; Costa, Daniel dos Santos

doi:10.1590/1809-4430-Eng.Agric.v42nepe20210160/2022

Acessibilidade / Reportar erro

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

Scientific Paper, Special Issue: Artificial Intelligence • Eng. agríc. (Online) 42 (spe) • 2022 • https://doi.org/10.1590/1809-4430-Eng.Agric.v42nepe20210160/2022 copy

DETERMINATION OF QUALITY AND RIPENING STAGES OF ‘PACOVAN’ BANANAS USING VIS-NIR SPECTROSCOPY AND MACHINE LEARNING

Authorship SCIMAGO INSTITUTIONS RANKINGS

ABSTRACT

This paper aimed to develop predictive models to determine total soluble solids, firmness, and ripening stages of ‘Pacovan’ bananas, using Vis-NIR spectroscopy and machine learning algorithms. A total of 384 bananas were divided into different days of storage (0, 3, 6, 9, 12, 15, 18, and 21 days) at two temperatures (25°C and 20°C). Bananas were subjected to spectral analysis using a spectrometer operating in spectral range of 350 – 2500 nm. Physicochemical parameters of quality, total soluble solids, and firmness were determined by reference analyses. Different machine learning algorithms were used to develop regression models and supervised classification. The best model for total soluble solids was the Random Forest with variable selection, showing an R²_cv of 0.90 and RMSECV of 2.31. The best model for firmness was the Support Vector Machine with variable selection, showing an R²_cv of 0.84 and RMSECV of 7.98. The best classification model for different ripening stages was the Multilayer Perceptron with variable selection, which achieved the precision of 74.22%. Therefore, Vis-NIR spectroscopy associated with machine learning algorithms is a promising tool for monitoring the quality and ripening stages of ‘Pacovan’ bananas.

KEYWORDS
quality attributes; non-destructive method; Musa spp.

Model
Qualit y parameter	Selection filter	RF			SVM			KNN			MLP
Qualit y parameter	Selection filter	R²_cv	RMSECV	MAECV	R²_cv	RMSECV	MAECV	R²_cv	RMSECV	MAECV	R²_cv	RMSECV	MAECV
TSS	no selection	0.79	3.37	2.41	0.85	2.85	2.13	0.79	3.40	2.35	0.61	4.78	3.24
	Wrapper	0.90	2.31	1.71	0.85	2.93	2.32	0.88	2.48	1.85	0.83	3.26	2.46
	CFS	0.86	2.71	2.00	0.69	4.13	3.30	0.86	2.63	1.96	0.81	3.38	2.52
Firmness	no selection	0.69	11.00	8.65	0.74	10.27	7.56	0.71	10.68	8.31	0.62	12.72	9.58
	Wrapper	0.82	8.52	6.50	0.84	7.98	6.12	0.82	8.57	6.64	0.74	10.44	8.33
	CFS	0.75	9.86	7.55	0.70	10.88	8.54	0.78	9.39	7.39	0.66	11.70	9.24

Model	Selection filter	precision (%)	Kappa	sensitivity (%)	selectivity (%)	FPR (%)
J48	no selection	62.50	0.42	60.52	76.86	23.14
	Wrapper	71.09	0.57	77.01	87.96	12.04
	CFS	71.87	0.59	76.43	88.25	11.75
RF	no selection	65.36	0.49	63.49	84.46	15.54
	Wrapper	68.75	0.51	81.02	88.04	11.96
	CFS	72.39	0.59	77.12	88.44	11.56
SVM	no selection	72.13	0.59	76.20	88.27	11.73
	Wrapper	73.18	0.60	78.53	89.25	10.75
	CFS	63.54	0.44	70.47	86.09	13.91
KNN	no selection	64.58	0.47	59.15	84.18	15.82
	Wrapper	73.96	0.62	76.22	89.02	10.98
	CFS	73.44	0.61	77.65	88.94	11.06
MLP	no selection	44.01	0.20	42.98	70.63	29.37
	Wrapper	74.22	0.62	78.26	89.31	10.69
	CFS	70.83	0.57	76.86	87.82	12.18

MLP model	Green	Nearly ripe	Ripe	Rotting
Precision	100.00	73.50	70.70	68.90
ROC area	100.00	89.90	82.80	83.00

Associação Brasileira de Engenharia Agrícola SBEA - Associação Brasileira de Engenharia Agrícola, Departamento de Engenharia e Ciências Exatas FCAV/UNESP, Prof. Paulo Donato Castellane, km 5, 14884.900 | Jaboticabal - SP, Tel./Fax: +55 16 3209 7619 - Jaboticabal - SP - Brazil
E-mail: revistasbea@sbea.org.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] *Corresponding author. College of Agricultural and Environmental Engineering, Federal University of the São Francisco Valley/ Juazeiro - BA, Brazil. E-mail: daniel.costa@univasf.edu.br