Skip to content
BY 4.0 license Open Access Published by De Gruyter Open Access January 7, 2023

Machine learning-based prediction of total phenolic and flavonoid in horticultural products

  • Kusumiyati Kusumiyati EMAIL logo and Yonathan Asikin
From the journal Open Agriculture

Abstract

The purpose of this study was to predict the total phenolic content (TPC) and total flavonoid content (TFC) in several horticultural commodities using near-infrared spectroscopy (NIRS) combined with machine learning. Although models are typically developed for a single product, expanding the coverage of the model can improve efficiency. In this study, 700 samples were used, including varieties of shallot, cayenne pepper, and red chili. The results showed that the TPC model developed yielded R 2cal, root mean squares error in the calibration set, R 2pred, root mean squares error in prediction set, and ratio of performance to deviation values of 0.79, 123.33, 0.78, 124.20, and 2.13, respectively. Meanwhile, the TFC model produced values of 0.71, 44.52, 0.72, 42.10, and 1.87, respectively. The wavelengths 912, 939, and 942 nm are closely related to phenolic compounds and flavonoids. The accuracy of the model in this study produced satisfactory results. Therefore, the application of NIRS and machine learning to horticultural products has a high potential of replacing conventional laboratory analysis TPC and TFC.

1 Introduction

Natural antioxidants are produced from plants with secondary metabolites such as phenols and flavonoids. Phenolic compounds are the largest group of compounds that act as natural antioxidants. Furthermore, polyphenols are the most common types of natural phenolic compounds from which ether, ester, or glycoside compounds, such as flavonoids, tannins, tocopherols, coumarins, lignins, cinnamic acid derivatives, and polyfunctional organic acids, are produced. These phenolic compounds influence the sensory properties of food, with tannins contributing significantly to food astringency. The ability of antioxidants to reduce free radicals increases as the total phenolic and flavonoid levels rise [1].

Chemical analysis in the laboratory is generally used to determine quality attributes of horticultural products, which requires a significant amount of time and money. It also generates chemical waste, which can harm the environment [2]. Therefore, a technique for measuring the quality attributes of horticultural commodities without the need for chemical analysis in the laboratory is required.

In recent decades, various techniques have been developed as alternatives to chemical analysis, including hyperspectral imaging [3], visible/near-infrared spectroscopy [4,5,6], acoustic vibration [7,8], nuclear magnetic resonance [9], and electronic nose [10]. Near-infrared spectroscopy (NIRS) has several advantages over other techniques, including the ability to predict organic samples in solid, liquid, and gaseous forms. Various studies have shown that the accuracy of NIRS, when used on agricultural products, is generally high. Nieto-Ortega et al. [11] used NIRS to predict the non-polysaccharide content in monogastric cereal feed ingredients and obtained an average accuracy value (R 2) of 0.90. According to Digman and Runge [12], NIRS performed excellently (R 2 = 0.86) in predicting the maturity of green peas. Reis [13] also concluded that NIRS showed satisfactory performance in predicting meat attributes. Furthermore, the NIRS tool is easy to use, allowing beginners to use it with ease.

The agricultural industry has begun implementing automation in areas such as harvesting, watering, fertilizing, and pesticide application. Machine learning is a branch of artificial intelligence that allows a machine to learn from data. It is also the application of computers and mathematical algorithms to generate future predictions through data-driven learning [14]. In this case, machine learning was combined with NIRS to predict quality attributes and classify agricultural commodities without damaging them. The application of NIRS and machine learning allows for automation in grading agricultural commodities based on prediction results of desired quality attributes.

Most NIRS studies only develop models for specific products, such as apples [15,16], peaches [17], pears [18], and lemons [19,20]. Developing a model that predicts the quality attributes of various agricultural commodities can increase its efficiency. Therefore, this study aimed to predict quality attributes, such as total phenolic and flavonoid contents using NIRS combined with machine learning on several horticultural commodities. This study also tested several spectra preprocessing methods for reducing noise in the spectral data.

2 Materials and methods

2.1 Sample preparation

The samples used in this study were shallot (var. Batu Ijo, var. Bima, var. Trisula, and var. Sumenep), cayenne pepper (var. Domba, var. Manik, and var. Ratuni UNPAD), and red chili (var. UNPAD CB2, var. Lingga, var. Tanjung, and var. Tanjung 2). Planting took place at an altitude of 829 m above sea level (masl), with average daily temperatures ranging from 20 to 30°C. The harvested samples were in good condition and free of pests and diseases before transferring them to the Laboratory of Horticulture, Faculty of Agriculture, Universitas Padjadjaran, for further analysis. The sample was thinly sliced, dried in an oven at 60°C for 24 h, and ground with a mortar until smooth [21]. Furthermore, powdered samples were prepared for spectral data collection.

2.2 Spectra data acquisition

NirVana AG410 (Integrated Spectronics Pty, Ltd., North Ryde, Australia) with a wavelength of 702–1,065 nm was used for data collection of spectra samples. The sample was placed in a Petri dish with a black cardboard base. Each sample was then scanned four times before obtaining the spectra data from the mean value of the measurements. Spectral data in this study were collected in diffuse reflectance mode and then converted into absorbance values.

2.3 Phenolic and flavonoid analysis

Total phenolic content (TPC) and total flavonoid content (TFC) measurements were carried out using a UV–Vis spectrophotometer (Shimadzu, UV mini-1240, Tokyo, Japan). The Folin–Ciocalteu method, developed by Lim and Murtijaya, was used to measure TPC [22]. The extracted filtrate was mixed with 2.5 mL of Folin–Ciocalteu reagent, followed by 2 mL of sodium carbonate. After incubation for 1 h, the filtrate was measured at a wavelength of 765 nm. Sytar et al. developed a procedure for measuring TFC [23]. The extracted filtrate was mixed with 2 mL of methanol and 0.1 mL of AlCl3. Subsequently, 1 M sodium acetate and 2.3 mL of water were added, before incubating for 30 min and measuring at 432 nm.

2.4 Multivariate data analysis

Preprocessed of the sample absorbance spectra was performed using multiplicative scatter correction (MSC), standard normal variate (SNV), moving average (MA), baseline correction, de-trending, first derivative (dg1), and second derivative Savitzky–Golay (dg2). The spectra preprocessing was done to reduce the variety of spectra produced by light interference, temperature differences during data collection, and background information [24,25]. Furthermore, spectra preprocessing was expected to improve the predictive ability of the developed model [26]. Partial least squares (PLS) regression was used in model development for predicting phenolic (TPC), and flavonoid content (TFC), while data analysis and model development were carried out using the Unscrambler X 10.4 software (Camo AS, Oslo, Norway).

3 Results and discussion

3.1 Measured data analysis

Table 1 shows the TPC and TFC of various horticultural commodities. Several factors influence physiological properties in agricultural products, including genetics, growing environment, plant cultivation techniques, postharvest handling, and analytical methods. Cayenne pepper var. Ratuni UNPAD had values ranging from 1136.14 to 1982.53 mg/100 g dry weight (DW), with the highest mean TPC value being 1453.87 mg/100 g DW. Meanwhile, shallot var. Sumenep had the lowest mean TPC value of 455.79 mg/100 g DW, with values ranging from 276.86 to 734.64 mg/100 g DW. Shallot var. Trisula showed the highest mean TFC value of 310.33 mg/100 g DW, with values ranging from 57.02 to 587.80 mg/100 g DW, while shallot var. Batu Ijo had the lowest total flavonoid of 45.28 mg/100 g DW, with values ranging from 18.09 to 73.87 mg/100 g DW. The difference in TPC and TFC values for each commodity and variety is due to differences in chemical composition. The range of TPC and TFC for powdered samples in this study has values similar to that reported in previous studies [27,28]. However, the analytical method for determining TPC and TFC affects the measurement results. In this study, phenolic compounds are sensitive to high temperatures, but the samples used were powdered samples that had been oven-dried. Therefore, the findings in this study are lower than previous reports on TPC and TPF in fresh samples [29,30].

Table 1

Wet chemistry data of total phenolic (TPC) and flavonoid content (TFC) derived from several horticultural products

Commodity Variety TPC (mg/100 g DW) TFC (mg/100 g DW)
Range Mean Range Mean
Shallot Batu Ijo 270.65–669.70 526.18 18.09–73.87 45.28
Bima 614.18–1612.53 944.64 65.78–475.59 212.66
Trisula 717.36–1531.78 1045.32 57.02–587.80 310.33
Sumenep 276.86–734.64 455.79 48.02–461.12 174.93
Cayenne pepper Domba 1129.29–1358.46 1221.30 162.69–210.31 187.03
Manik 1025.49–1279.43 1153.44 169.75–232.72 204.98
Ratuni UNPAD 1136.14–1982.53 1453.87 30.83–625.56 117.50
Red chili UNPAD CB2 784.16–1418.33 1078.20 76.88–324.29 224.60
Lingga 893.42–1242.37 1065.22 112.98–282.14 173.19
Tanjung 606.47–1571.25 945.63 112.34–406.96 243.87
Tanjung 2 968.61–1209.78 1082.45 167.46–419.49 262.03

3.2 Spectra characteristics

The absorbance data reflect the characteristic spectra displayed by the sample. Differences in commodity, variety, sample surface, and temperature affect the spectral data obtained. In this study, spectral data with a wavelength of 702–1,065 nm covering the near-infrared region for model development were used. These areas detect various quality attributes in agricultural commodities, including water content, starch, sugar, and other physicochemical properties [31,32,33]. Moreover, several other studies reported using NIRS to monitor adulteration in agricultural commodities, fruit pesticide residues, and early disease detection in potatoes [34,35,36,37]. One of the challenges in analyzing NIRS data is the amount of noise in the sample spectra. Therefore, special handling is sometimes required to reduce the noise. The sample spectra data were analyzed using various preprocessing methods before the model development stage. Based on the results of the spectral preprocessing shown in Figure 1, the peak was not visible in the 702–1,065 nm region in the original spectra, as well as in the MSC and MA spectra. Furthermore, the SNV and MSC methods are similar in that they normalize data by reducing the effect of multiplicative and light scattering on the original spectra. In several other studies, the application of MSC and SNV showed similar spectral patterns, but this was not the case in this study. MA works by averaging adjacent points and using the value as a new point. Although the detrending method is generally used to process grain samples [38], it was used in processing powder samples in this study. Several high peaks were seen in the baseline, SNV, detrending, dg1, and dg2 spectra at 840, 900, 940, and 1,025 nm. The peak was clearly visible, particularly in the detrending spectra, dg1 and dg2. The Savitzky–Golay derivative method (dg1 and dg2) is typically used to detect compounds in small concentrations. Therefore, the peaks in the spectral data are more clearly observed with this method. It is necessary to have the right strategy in determining the spectra preprocessing method since the Savitzky–Golay derivative method sometimes increases the error value in the resulting model. Despite preprocessing, this study was still unable to determine the best spectra preprocessing method for this dataset. This can be determined after modeling the results of each preprocessing spectra.

Figure 1 
                  Original and preprocessed spectra of samples. (a) Original, (b) MSC, (c) SNV, (d) MA, (e) baseline, (f) detrending, (g) dg1, and (h) dg2.
Figure 1

Original and preprocessed spectra of samples. (a) Original, (b) MSC, (c) SNV, (d) MA, (e) baseline, (f) detrending, (g) dg1, and (h) dg2.

3.3 Model development

Regression modeling involves spectral and wet chemistry data. As shown in Table 2, 525 samples were assigned to develop a regression model using PLS, while 175 samples were used to test the reliability of the developed model. The regression model was then validated using K-fold cross-validation, and the samples were divided into 20 segments, each segment containing 26–27 samples. This cross-validation aimed to determine the optimum principal components and avoid overfitting, a condition in which the model fails to predict an unknown sample. A good model is expected to have a high R 2 and a low error in the calibration and prediction set [39]. Therefore, to improve the accuracy of the developed model, various spectra preprocessing methods were tested. The determination of the best model was based on the coefficient of determination in the calibration set (R 2cal), root mean squares error in the calibration set (RMSEC), coefficient of determination in the prediction set (R 2pred), root mean squares error in prediction set (RMSEP), and the ratio of performance to deviation (RPD). However, when compared to other results, the original spectra (without preprocessing) presented the best model in this study. R 2cal, RMSEC, R 2pred, RMSEP, and RPD values for the TPC model were 0.79, 123.33, 0.78, 124.20, and 2.13, while the TFC model had values of 0.71, 44.52, 0.72, 42.10, and 1.87, respectively. This phenomenon sometimes occurs in NIRS data analysis, where preprocessed spectra do not improve model accuracy. It is possibly due to the high variability of the dataset in this study. Spectra preprocessing reduces the variability or diversity in data spectra samples. Therefore, the original spectra had the best performance on both quality attributes. Similar results were obtained in a study conducted by Zhao et al. [40], who found that the original spectra outperformed the MSC and SNV preprocessing results. Rubini et al. [41] reported that the model developed from the SNV spectra was less accurate than the original spectra in predicting levopimaric acid and turpentine in maritime pine.

Table 2

PLS regression results for prediction of total phenolic (TPC) and flavonoid content (TFC) in horticultural products

Trait Preprocessing Calibration (n = 525) Prediction (n = 175) RPD
R 2cal RMSEC R 2pred RMSEP
TPC Original 0.79 123.33 0.78 124.20 2.13
MSC 0.53 185.16 0.50 187.87 1.41
SNV 0.75 136.12 0.70 143.87 1.84
MA 0.77 128.61 0.75 131.78 2.01
Baseline 0.77 128.07 0.75 130.89 2.02
Detrending 0.77 128.88 0.75 130.64 2.02
dg1 0.75 135.10 0.70 143.53 1.84
dg2 0.74 138.00 0.70 145.04 1.82
TFC Original 0.71 44.52 0.72 42.10 1.87
MSC 0.49 58.88 0.57 52.70 1.53
SNV 0.66 47.85 0.69 44.95 1.80
MA 0.70 44.88 0.72 43.05 1.88
Baseline 0.70 45.37 0.67 46.22 1.75
De-trending 0.65 48.65 0.63 48.68 1.66
dg1 0.67 47.26 0.67 45.84 1.76
dg2 0.65 48.66 0.67 46.02 1.75

Figure 2 shows scatter plots for TPC and TFC obtained from the best model, which is the original spectra. The red data distribution is the result of the analysis of the calibration set, while the yellow one is the prediction set. The closer the regression line data distribution, the better the resulting model. PLS is the most commonly used regression method for NIRS data analysis. The advantage of PLS is that it can reduce correlated independent variables (spectral data) and convert them into new, uncorrelated variables. PLS also involves the dependent variable (wet chemistry data) in forming these new variables. Therefore, the new variable contains information from the dependent variable, affecting good model accuracy. Kusumiyati et al. [42,43] conducted a study using NIRS combined with PLS regression to predict SSC in sapodilla and mango. The results revealed that NIRS combined with PLS is effective for SSC prediction. Furthermore, accuracy in model development is influenced by various factors, including data variability, modeling techniques, and instruments [44].

Figure 2 
                  Scatter plot for calibration and prediction of total phenolic (TPC) and flavonoid content (TFC) in horticultural products.
Figure 2

Scatter plot for calibration and prediction of total phenolic (TPC) and flavonoid content (TFC) in horticultural products.

Agricultural commodities, particularly horticulture, have a complex chemical composition that comprises a large set of overtones and combination bands. In this research, PLS regression employed NIRS spectra and chemical composition, specifically TPC and TFC. The regression coefficient is utilized for understanding the calibration model resulting from PLS regression. As shown in Figure 3, regression coefficients indicate which wavelength contributes the most to the modeling of each quality attribute, and these important wavelengths are characterized by peaks and valleys. The regression coefficients were analyzed from the best model on both quality attributes, namely the model developed using the original spectra. Furthermore, peaks at wavelengths 912, 939, and 942 nm are associated with the third CH overtone and the second overtone of OH vibrations. These are also associated with the detection of antioxidants and phenol groups [45].

Figure 3 
                  The regression coefficient of a machine learning model for predicting total phenolic (TPC) and flavonoid content (TFC).
Figure 3

The regression coefficient of a machine learning model for predicting total phenolic (TPC) and flavonoid content (TFC).

4 Conclusion

Based on the result of this study, NIRS can be combined with machine learning to predict TPC and TFC in horticultural products. The results show that the developed model has a fairly good performance with R 2 > 0.7 and that spectra preprocessing did not increase model accuracy. It could be because of the huge variance of the input data. Spectra preprocessing generally operates by minimizing the variability of data. Hence, the models generated from the preprocessed spectra performed no better than the original spectra. Furthermore, the wavelength of 900–950 nm is essential in developing TPC and TFC models. Therefore, the application of NIRS and machine learning is reliable enough to replace conventional measurement methods for TPC and TFC.

Acknowledgments

The authors are grateful to Yuda Hadiwijaya, Ine Elisa Putri, and Yusuf Eka Maulana as assistants who have provided direction, advice, and support for this study.

  1. Funding information: The authors state no funding involved.

  2. Conflict of interest: The authors state no conflict of interest.

  3. Data availability statement: The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

References

[1] Troszyńska A, Narolewska O, Robredo S, Estrella I, Hernández T, Lamparski G, et al. The effect of polysaccharides on the astringency induced by phenolic compounds. Food Qual Prefer. 2010;21(5):463–9.10.1016/j.foodqual.2009.12.005Search in Google Scholar

[2] Parrini S, Acciaioli A, Franci O, Pugliese C, Bozzi R. Near infrared spectroscopy technology for prediction of chemical composition of natural fresh pastures. J Appl Anim Res. 2019;47(1):514–20. 10.1080/09712119.2019.1675669.Search in Google Scholar

[3] Çetin N, Karaman K, Kavuncuoğlu E, Yıldırım B, Jahanbakhshi A. Using hyperspectral imaging technology and machine learning algorithms for assessing internal quality parameters of apple fruits. Chemom Intell Lab Syst. 2022 Nov 15;230:104650.10.1016/j.chemolab.2022.104650Search in Google Scholar

[4] Kusumiyati, Hadiwijaya Y, Putri IE, Mubarok S, Hamdani JS. Rapid and non-destructive prediction of total soluble solids of guava fruits at various storage periods using handheld near-infrared instrument. IOP Conference Series: Earth and Environmental Science. Yogyakarta, Indonesia: 2020. p. 1–7.10.1088/1755-1315/458/1/012022Search in Google Scholar

[5] Suhandy D, Yulia M, Kusumiyati. The authentication of peaberry and civet ground roasted robusta coffee using UV-visible spectroscopy and PLS-DA method with two different particle sizes. IOP Conference Series: Earth and Environmental Science. Vol. 258. Issue 1; 2019. p. 012043.10.1088/1755-1315/258/1/012043Search in Google Scholar

[6] Kusumiyati, Mubarok S, Sutari W, Farida, Hamdani JS, Hadiwijaya Y, et al. Non-destructive method for predicting sapodilla fruit quality using near infrared spectroscopy. IOP Conference Series: Earth and Environmental Science. Vol. 334. Issue 1; 2019. p. 012045.10.1088/1755-1315/334/1/012045Search in Google Scholar

[7] Tian S, Wang J, Xu H. Firmness measurement of kiwifruit using a self-designed device based on acoustic vibration technology. Postharvest Biol Technol. 2022;187:111851.10.1016/j.postharvbio.2022.111851Search in Google Scholar

[8] Zhang W, Lv Z, Shi B, Xu Z, Zhang L. Evaluation of quality changes and elasticity index of kiwifruit in shelf life by a nondestructive acoustic vibration method. Postharvest Biol Technol. 2021;173:111398.10.1016/j.postharvbio.2020.111398Search in Google Scholar

[9] Shicheng Q, Youwen T, Qinghu W, Shiyuan S, Ping S. Nondestructive detection of decayed blueberry based on information fusion of hyperspectral imaging (HSI) and low-Field nuclear magnetic resonance (LF-NMR). Comput Electron Agric. 2021 May 1;184:106100.10.1016/j.compag.2021.106100Search in Google Scholar

[10] Asikin Y, Kusumiyati, Shikanai T, Wada K. Volatile aroma components and MS-based electronic nose profiles of dogfruit (Pithecellobium jiringa) and stink bean (Parkia speciosa). J Adv Res. 2018;9:79–85.10.1016/j.jare.2017.11.003Search in Google Scholar PubMed PubMed Central

[11] Nieto-Ortega B, Arroyo JJ, Walk C, Castañares N, Canet E, Smith A. Near infrared reflectance spectroscopy as a tool to predict non-starch polysaccharide composition and starch digestibility profiles in common monogastric cereal feed ingredients. Anim Feed Sci Technol. 2022 Mar 1;285:115214.10.1016/j.anifeedsci.2022.115214Search in Google Scholar

[12] Digman MF, Runge WM. The utility of a near-infrared spectrometer to predict the maturity of green peas (Pisum sativum). Comput Electron Agric. 2022 Feb 1;193:106643.10.1016/j.compag.2021.106643Search in Google Scholar

[13] Reis MM. Predicting meat attributes from intact muscle using near-infrared spectroscopy. Ref Modul Food Sci; 2022 Jan 1.10.1016/B978-0-323-85125-1.00096-XSearch in Google Scholar

[14] Liakos KG, Busato P, Moshou D, Pearson S, Bochtis D. Machine learning in agriculture: A review. Sensors. 2018 Aug 14;18(8):2674. https://www.mdpi.com/1424-8220/18/8/2674/htm.10.3390/s18082674Search in Google Scholar PubMed PubMed Central

[15] Lan W, Bureau S, Chen S, Leca A, Renard CMGC, Jaillais B. Visible, near- and mid-infrared spectroscopy coupled with an innovative chemometric strategy to control apple puree quality. Food Control. 2021;120.10.1016/j.foodcont.2020.107546Search in Google Scholar

[16] Pourdarbani R, Sabzi S, Arribas JI. Nondestructive estimation of three apple fruit properties at various ripening levels with optimal Vis-NIR spectral wavelength regression data. Heliyon. 2021 Sep 1;7(9):e07942.10.1016/j.heliyon.2021.e07942Search in Google Scholar PubMed PubMed Central

[17] Minas IS, Blanco-Cipollone F, Sterle D. Accurate non-destructive prediction of peach fruit internal quality and physiological maturity with a single scan using near infrared spectroscopy. Food Chem. 2021 Jan 15;335:127626.10.1016/j.foodchem.2020.127626Search in Google Scholar PubMed

[18] Zhao F, Du G, Huang Y. Exploring the use of Near-infrared spectroscopy as a tool to predict quality attributes in prickly pear (Rosa roxburghii Tratt) with chemometrics variable strategy. J Food Compos Anal. 2022 Jan 1;105:104225.10.1016/j.jfca.2021.104225Search in Google Scholar

[19] Ruggiero L, Amalfitano C, Di Vaio C, Adamo P. Use of near-infrared spectroscopy combined with chemometrics for authentication and traceability of intact lemon fruits. Food Chem. 2022 May 1;375:131822.10.1016/j.foodchem.2021.131822Search in Google Scholar PubMed

[20] Moomkesh S, Mireei SA, Sadeghi M, Nazeri M. Early detection of freezing damage in sweet lemons using Vis/SWNIR spectroscopy. Biosyst Eng. 2017 Dec 1;164:157–70.10.1016/j.biosystemseng.2017.10.009Search in Google Scholar

[21] Kusumiyati K, Putri IE, Munawar AA, Suhandy D. A data fusion model to merge the spectra data of intact and powdered cayenne pepper for the fast inspection of antioxidant properties. Sustainability. 2022 Dec 25;14(1):201.10.3390/su14010201Search in Google Scholar

[22] Lim YY, Murtijaya J. Antioxidant properties of Phyllanthus amarus extracts as affected by different drying methods. LWT - Food Sci Technol. 2007;40(2007):1664–9.10.1016/j.lwt.2006.12.013Search in Google Scholar

[23] Sytar O, Hemmerich I, Zivcak M, Rauh C, Brestic M. Comparative analysis of bioactive phenolic compounds composition from 26 medicinal plants. Saudi J Biol Sci. 2018 May;25(4):631–41. https://pubmed.ncbi.nlm.nih.gov/29740227.10.1016/j.sjbs.2016.01.036Search in Google Scholar PubMed PubMed Central

[24] Kusumiyati K, Hadiwijaya Y, Suhandy D, Munawar AA. Prediction of water content and soluble solids content of ‘manalagi’ apples using near infrared spectroscopy. IOP Conference Series: Earth and Environmental Science. Vol. 922. Issue 1; 2021. p. 012062.10.1088/1755-1315/922/1/012062Search in Google Scholar

[25] Kusumiyati, Mubarok S, Sutari W, Hadiwijaya Y. Application of spectra pre-treatments on firmness assessment of intact sapodilla using vis-nir spectroscopy. IOP Conference Series: Earth and Environmental Science. Vol. 644. Issue 1; 2021. p. 012001.10.1088/1755-1315/644/1/012001Search in Google Scholar

[26] Kusumiyati K, Hadiwijaya Y, Putri IE, Munawar AA. Enhanced visible/near-infrared spectroscopic data for prediction of quality attributes in Cucurbitaceae commodities. Data Br. 2021;39:107458.10.1016/j.dib.2021.107458Search in Google Scholar PubMed PubMed Central

[27] Lu X, Wang J, Al-Qadiri HM, Ross CF, Powers JR, Tang J, et al. Determination of total phenolic content and antioxidant capacity of onion (Allium cepa) and shallot (Allium oschaninii) using infrared spectroscopy. Food Chem. 2011;129(2):637–44.10.1016/j.foodchem.2011.04.105Search in Google Scholar PubMed

[28] Woldemariam HW, Admassu Emire S, Getachew Teshome P, Toepfl S, Aganovic K. Physicochemical, functional, oxidative stability and rheological properties of red pepper (Capsicum annuum L.) powder and paste. Int J Food Prop. 2022 Oct 24;24(1):1416–37. 10.1080/10942912.2021.1969945 Search in Google Scholar

[29] Tundis R, Menichini F, Bonesi M, Conforti F, Statti G, Menichini F, et al. Antioxidant and hypoglycaemic activities and their relationship to phytochemicals in Capsicum annuum cultivars during fruit development. LWT - Food Sci Technol. 2013 Sep 1;53(1):370–7.10.1016/j.lwt.2013.02.013Search in Google Scholar

[30] Zhuang Y, Chen L, Sun L, Cao J. Bioactive characteristics and antioxidant activities of nine peppers. J Funct Foods. 2012 Jan 1;4(1):331–8.10.1016/j.jff.2012.01.001Search in Google Scholar

[31] Kusumiyati K, Hadiwijaya Y, Sutari W, Munawar AA. Global model for in-field monitoring of sugar content and color of melon pulp with comparative regression approach. AIMS Agric Food. 2022;7(2):312–25. 10.3934/agrfood.2022020.Search in Google Scholar

[32] Subedi PP, Walsh KB. Assessment of sugar and starch in intact banana and mango fruit by SWNIR spectroscopy. Postharvest Biol Technol. 2011 Dec 1;62(3):238–45.10.1016/j.postharvbio.2011.06.014Search in Google Scholar

[33] Acharya UK, Subedi PP, Walsh KB. Robustness of tomato quality evaluation using a portable vis-SWNIRS for dry matter and colour. Int J Anal Chem. 2017.10.1155/2017/2863454Search in Google Scholar PubMed PubMed Central

[34] Suhandy D, Yulia M, Kusumiyati. Chemometric quantification of peaberry coffee in blends using UV-visible spectroscopy and partial least squares regression. 2018.10.1063/1.5062774Search in Google Scholar

[35] Yang K, An C, Zhu J, Guo W, Lu C, Zhu X. Comparison of near-infrared and dielectric spectra for quantitative identification of bovine colostrum adulterated with mature milk. J Dairy Sci. 2022 Nov 1;105(11):8638–49.10.3168/jds.2022-21969Search in Google Scholar PubMed

[36] Liang PS, Haff RP, Hua SST, Munyaneza JE, Mustafa T, Sarreal SBL. Nondestructive detection of zebra chip disease in potatoes using near-infrared spectroscopy. Biosyst Eng. 2018 Feb 1;166:161–9.10.1016/j.biosystemseng.2017.11.019Search in Google Scholar

[37] Xie L, Zhu J, Wang Y, Wang N, Liu F, Chen Z, et al. Rapid and accurate determination of prohibited components in pesticides based on near infrared spectroscopy. Infrared Phys Technol. 2022 Mar 1;121:104038.10.1016/j.infrared.2022.104038Search in Google Scholar

[38] Rinnan Å, Berg F, van den Engelsen SB. Review of the most common pre-processing techniques for near-infrared spectra. Trends Anal Chem. 2009 Nov 1;28(10):1201–22. https://www.sciencedirect.com/science/article/abs/pii/S0165993609001629.10.1016/j.trac.2009.07.007Search in Google Scholar

[39] Kusumiyati, Hadiwijaya Y, Putri IE, Mubarok S. Water content prediction of “crystal” guava using visible-near infrared spectroscopy and chemometrics approach. IOP Conference Series: Earth and Environmental Science. Vol. 393. Issue 1; 2019. p. 012099.10.1088/1755-1315/393/1/012099Search in Google Scholar

[40] Zhao P, Xing J, Hu C, Guo W, Wang L, He X, et al. Feasibility of near-infrared spectroscopy for rapid detection of available nitrogen in vermiculite substrates in desert facility agriculture. Agriculture. 2022 Mar 15;12(3):411. https://www.mdpi.com/2077-0472/12/3/411/htm.10.3390/agriculture12030411Search in Google Scholar

[41] Rubini M, Feuillerat L, Cabaret T, Leroyer L, Leneveu L, Charrier B. Comparison of the performances of handheld and benchtop near infrared spectrometers: Application on the quantification of chemical components in maritime pine (Pinus Pinaster) resin. Talanta. 2021 Jan 1;221:121454.10.1016/j.talanta.2020.121454Search in Google Scholar PubMed

[42] Kusumiyati, Hadiwijaya Y, Putri IE. Determination of water content of intact sapodilla using near infrared spectroscopy. IOP Conference Series: Earth and Environmental Science. Vol. 207. Issue 1; 2018. p. 012047.10.1088/1755-1315/207/1/012047Search in Google Scholar

[43] Kusumiyati, Munawar AA, Suhandy D. Fast and contactless assessment of intact mango fruit quality attributes using near infrared spectroscopy (NIRS). IOP Conference Series: Earth and Environmental Science. Vol. 644. Issue 1; 2021. p. 012028.10.1088/1755-1315/644/1/012028Search in Google Scholar

[44] Kusumiyati K, Putri IE, Hamdani JS, Suhandy D. Real-time detection of the nutritional compounds in green ‘Ratuni UNPAD’ cayenne pepper. Horticulturae. 2022 Jun 20;8(6):554, https://www.mdpi.com/2311-7524/8/6/554.10.3390/horticulturae8060554Search in Google Scholar

[45] Kljusurić JG, Mihalev K, Bečić I, Polović I, Georgieva M, Djaković S, et al. Near-infrared spectroscopic analysis of total phenolic content and antioxidant activity of berry fruits. Food Technol Biotechnol. 2016;54(2):236–42.10.17113/ftb.54.02.16.4095Search in Google Scholar PubMed PubMed Central

Received: 2022-11-06
Revised: 2022-12-03
Accepted: 2022-12-07
Published Online: 2023-01-07

© 2023 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Downloaded on 28.9.2023 from https://www.degruyter.com/document/doi/10.1515/opag-2022-0163/html
Scroll to top button