Honey authentication based on physicochemical parameters and phenolic compounds

Computers and Electronics in Agriculture 138 (2017) 148–156

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag

Original papers

Honey authentication based on physicochemical parameters and phenolic compounds Mircea Oroian ⇑, Ropciuc Sorina Faculty of Food Engineering, Stefan cel Mare University of Suceava, Romania

a r t i c l e

i n f o

Article history: Received 5 October 2016 Received in revised form 24 April 2017 Accepted 25 April 2017 Available online 4 May 2017 Keywords: Honey Authentication Phenolics Physicochemical properties Chemometrics

a b s t r a c t The aim of this study is to assess the usefulness of physicochemical parameters (pH, water activity, free acidity, refraction index, Brix, moisture content and ash content), color parameters (L⁄, a⁄, b⁄, chroma, hue angle and yellow index) and phenolics (quercetin, apigenin, myricetin, isorhamnetin, kaempherol, caffeic acid, chrysin, galangin, luteolin, p-coumaric acid, gallic acid and pinocembrin) in view of classifying honeys according to their botanical origin (acacia, tilia, sunflower, honeydew and polyfloral). Thus, the classification of honeys has been made using the principal component analysis (PCA), linear discriminant analysis (LDA) and artificial neural networks (ANN). The multilayer perceptron network with 2 hidden layers classified correctly 94.8% of the cross validated samples. Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction Honey is an ancient food, which is largely consumed due to its nutritional, medicinal and cosmetic properties. The high value of honey is given by its nutritional value, macro and microelements and many other compounds it contains (Jasicka-Misiak et al., 2012). The honey composition (sugars, organic acids, enzymes, vitamins, proteins and phytochemicals) is influenced by the botanical and geographical origin and environmental climatic conditions (Baroni et al., 2015; Solayman et al., 2016). Glucose and fructose are the major sugars present in honey, but there have been reported smaller quantities of twenty-two other compounds (e.g. maltose, sucrose, maltulose, turanose, isomaltose, etc.) (Siddiqui et al., 2017). Honey contains different types of enzymes such as: oxidase, catalyse, acid phospatase, invertase and diastase, which make it unique in the sweeteners domain. Moisture content, reducing sugars, free acids, electrical conductivity, sucrose content and 5-HMF influence nutritional quality, granulation, flavor and texture parameters. In addition to the previously mentioned compounds, phyto-chemical compounds present in honey play a major role in determining the antioxidant activity, which can be correlated with the anti-inflammatory, anti-carcinogenic, antithrombotic, anti-atherogenic activity of honey (Piljac-Zˇegarac et al., 2009). Among the phyto-chemical compounds present in honey, the phenolic compounds play a major role in the antioxi⇑ Corresponding author. E-mail address: [email protected] (M. Oroian). http://dx.doi.org/10.1016/j.compag.2017.04.020 0168-1699/Ó 2017 Elsevier B.V. All rights reserved.

dant activity. The phenolic compounds found in honey are free phenols, phenolic acids, polyphenols (usually in the form of flavonoids), anthocyanins, procyanidins and pigments. Their total content depends on the species of plant from which bees collected the nectar and their amount varies from 5 to 1300 mg/kg (Mattonai et al., 2016; Mellen et al., 2015). The necessity for determining some parameters in terms of the botanical or geographical authentication of honeys derives from the increasing demand for mono-floral honeys on markets. Mono-floral honeys are more expensive than multi-floral ones. A honey can be mono-floral or polyfloral in origin depending on whether it is derived from one or several plant species. According to the international food standards, for a honey to be labelled with floral origin, it must originate wholly or predominantly from a particular floral source and display the corresponding organoleptic, physicochemical and microscopic properties (Codex Alimentarius, 2001). Adulteration of honey can be determined using the quality parameters, and these parameters can confirm the hygiene conditions for the manipulation and storage of honey (da Silva et al., 2016). The authentication of honey has started with the melissopalynological method, which can be used for botanical authentication (Karabagias et al., 2014). An alternative for honey authentication is the combination of melissopalynological method with physicochemical parameters (Oroian et al., 2015a; Karabagias et al., 2014; Escriche et al., 2014; Juan-Borrás et al., 2014). Over the last decades there have been implemented different methods for the authentication of honey such as: e-tongue and optical

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156

spectroscopy (Ulloa et al., 2013), potentiometric and voltammetric electronic tongue (Wei and Wang, 2014), headspace volatile profile (Oroian et al., 2015a), phenolic compounds, physicochemical parameters and chemometrics (Karabagias et al., 2014), mineral profile (Oroian et al., 2015b), NIR spectroscopy (Guelpa et al., 2017). The huge data resulted from the physicochemical properties, volatile fraction, e-tongue, mineral profile, NIR spectroscopy, etc., cannot be used for honey authentication without applying proper statistical methods. The statistical methods have been used to study the usefulness of different parameters in the authentication of honeys. Over the last decades there have been used different statistical methods for the honey authentication such as: principal component analysis (Wei and Wang, 2014; Oroian et al., 2015a, b, Ulloa et al., 2013), discriminant analysis (Wei and Wang, 2014), least square discriminant analysis (Guelpa et al., 2017), cluster analysis (Ulloa et al., 2013), artificial neural networks (Ramzi et al., 2015). There are only a few systematic studies on the classification of Romanian honeys according to the botanical origin using the mineral content (Oroian et al., 2015b), volatile compounds (Oroian et al., 2015a), stable isotope (Dinca et al., 2015), rheological parameters (Dobre et al., 2012) and physicochemical parameters (Marghitas et al., 2010). The purpose of this paper is to investigate the usefulness of phenolics and physicochemical parameters for the authentication of acacia, tilia, sunflower, polyfloral and honeydew from Romania. 2. Materials and methods 2.1. Materials 50 honey samples of five different botanical origins (acacia, tilia, sunflower, honeydew and polyfloral) have been purchased from local beekeepers from Suceava County, Romania. Quercetin, apigenin, myricetin, isorhamnetin, kaempherol, caffeic acid, chrysin, galangin, luteolin, p-coumaric acid, gallic acid and pinocembrin have been purchased from Plant MetaChem (Germany). Amberlite XAD-2 resin, methanol, HCl, diethyl ether, membrane filter 0.45 lm have been purchased form Sigma Aldrich (Germany). 2.2. Methods 2.2.1. Melissopalynological analysis The pollen analysis was made according to the method of Louveaux et al. (1970), using a non-acetolytic method. Ten grams of honey were mixed with about 40 ml of distilled water; then, centrifuged at 4500 rpm (3383g) for 15 min and the supernatant was carefully removed. The residue was re-dissolved again and centrifuged for other 15 min. The full sediment was used to prepare the slide. The pollen spectrum of each honey sample was determined by a light microscopy (Motic
40) by counting at least 800 pollen grains. For all pollen types the individual occurrence was expressed as percentage (Dobre et al., 2012). 2.2.2. Physicochemical properties Moisture content, pH, refraction index, Brix concentration, electrical conductivity and ash content have been determined using the Harmonised methods of the international honey commission (Bogdanov et al., 2002). Water activity was measured using a water activity meter AquaLab Lite (Decagon, USA). Colour has been determined using a Konica CR400 cromameter (Konica Minolta, Japan). The samples were placed in a 20 mm vat and they were measured to a white spectrum. The color intensity, hue angle and yellow index (YI) were computed as follows:

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 c ¼ a2 þ b

ð1Þ

  b a

ð2Þ

142:86  b L

ð3Þ

h ¼ tan1

YI ¼

149

2.2.3. Phenolics extraction The phenolics extraction was made using the method described by Baltrušaityte˙ et al. (2007) and Escriche et al. (2014). Sixty grams of Amberlite XAD-2 resin, pore size 9 nm, and particle size 0.3–1.2 mm were soaked in methanol for 10 min, then, the most of methanol was decanted and replaced by distilled water. The mixture was stirred, allowed to stand for 5–10 min and packed into a glass column, 25
2 cm. The honey samples (25 g) were thoroughly mixed with 250 mL of distilled water and adjusted to pH 2 by concentrated HCl. The solution was slowly filtered through the column packed as previously described. The column was washed with 250 mL of acidified water (pH 2 with HCl) and subsequently rinsed with 300 mL of neutral distilled water to remove all sugars and other polar compounds of honey. The flavonoids and phenolic compounds were eluted from the sorbent with 250 mL of methanol. The methanol extracts were concentrated under vacuum at 40 °C in a rotary evaporator. The residue was dissolved in 5 mL of distilled water and extracted three times with 5 mL of diethyl ether. The dried residue was then re-dissolved in 1 mL of methanol (HPLC grade) and filtered through a membrane filter with a 0.45 lm pore size. Three replicate extractions were performed for each sample. 2.2.4. HPLC analysis of phenolics The phenolic compounds were separated and quantified using the method described by Coneac et al. (2008). A High Performance Liquid Chromatography (HPLC) (Shimadzu, Kyoto, Japan) system equipped with a LC-20 AD liquid chromatograph, SIL-20A auto samples, CTO-20AC auto sampler and a SPD-M-20A diode array detector was used. The separation was carried out on a Zorbax SP-C18 column, with 150 mm length, 4.6 mm i.d., and 5 µmdiameter particle was used; the phenolics detection was set at 200 nm and 210 nm. The mobile phase was acetonitrile: water ratio 48:52, temperature was of 25 °C, with a flow of 0.3 ml/min, the injected sample volume was of 20 ll. The diluted standard solutions of quercetin, apigenin, myricetin, isorhamnetin, kaempherol, caffeic acid, chrysin, galangin, luteolin, p-coumaric acid, gallic acid and pinocembrin were analyzed under the same HPLC conditions and furthermore the calibration of the detector response was made. Data collection and subsequent processing were performed using the LC solution software 1.21 version (Shimadzu, Kyoto, Japan). The phenolics were identified by comparing the chromatographic retention times and UV spectra of each compound. The calibration curves were constructed via least-squares linear regression analyses of the ratio of the peak area of each representative compound versus the respective concentration. The regression analysis (n = 5) showed higher correlation coefficients (R2) higher than 0.99 for all the compounds. The quantitative results were expressed as mg of compound per 100 g honey. 2.3. Statistical analysis The statistical analysis was performed using Unscrambler X 10.1 software system (Camo, Norway). The data corresponding to each variable were analyzed by one-factor analysis of variance

150

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156

(ANOVA). Multiple comparisons were performed using the least significant difference test (LSD) and Fisher ratio (F), and statistical significance was set at a = 0.05. The principal component analysis (PCA) and linear discriminant analysis (LDA) were made using Unscrambler X 10.1 software system (Camo, Norway). The artificial neural networks –ANN– (multilayer perceptron (MLP), probabilistic neural network (PNN), recurrent neural network (RNN) and modular neural network (MNN)) was made using Neurosolution 7.0 trial version (IBM, USA). Multivariate statistical methods such as PCA, LDA and ANN can be used to extract specific and sensitive information from the physicochemical parameters and phenolics concentration. The Principal component analysis (PCA) has in general two outputs: one is the score plot which presents the locations of the samples in the principal component that is able to detect the samples’ patterns, by grouping similarities and differences, and the second one is the loading plot which interprets the relationships between variables (Kara, 2009). The loadings can be used to understand how much each variable contributes to the meaningful variation in the data, and to interpret variable relationship (Kara, 2009). Linear discriminant analysis (LDA) assumes that the various classes collecting similar objects (different samples of honey of different botanical origins) are described by multivariate normal distributions having the same covariance, but difference location of centroids within the variable domain. Separations between classes are hyper planes and the allocation of a given object within one of the classes is based on a maximum likelihood discriminant rule (D’Archivio et al., 2016). Artificial neural networks (ANN) are a set of mathematical methods, which attempt to mimic the functioning of the human brain (Binetti et al., 2017), and consist of sophisticated non-linear computational tools that are able of modelling extremely complex learning functions. The artificial neural networks are configured for pattern recognition or data classification (Trafialek et al., 2015). The applications of the three statistical analyses have as goal the botanical classification of the honeys based on the physicochemical parameters and phenolic compounds, in order to see the usefulness of these parameters in the authentication and by which statistical analysis is much suitable for predicting the membership of a sample to a group. 3. Results and discussion 3.1. Honey classification based on melissopalynological analysis and electrical conductivity The main pollen types present in each honey analyzed are shown in the Table 1. For the classification of acacia honey it requires that the honey should contain at least 45% of the pollen to belong to Robinia pseudoacacia, in the case of sunflower at least

60% of the pollen to belong to Helianthus annuus, and in the case of tilia at least 60% of the pollen to belong to Tilia europea (JuanBorrás et al., 2014). Accordingly to the main pollen, 41 samples were classified as follows: 10 samples of acacia, 8 samples of tilia, 11 samples of sunflower and 12 samples of polyfloral. Some other pollen grains were observed in honey samples, others than the major ones: in the case of acacia - Brassica napus, Prunus, Plantago, Trifolium and Rubus, in the case of tilia - Brassica napus, Helianthus annuus, Galium and Trifolium, in the case of sunflower – Tilia, Taraxacum officinale, Trifolium, Fragaria, Brassica napus and Robinia pseudoacacia, in the case of polyfloral honeys – Brassica napusßPrunus, Tilia, Robinia pseudoacacia, Plantago, Rubus, Taraxacum officinale, Fragaria, Trifoloium europea and Galium. In the case of honeydew honeys, the authentication is not made based on the pollen grains, but their pollen profile was analyzed and the following types of pollen were identified: Brassica napus, Quercus, Trifolium repens, Castanea sativa and Helianthus annuus. Honeydew samples can be authenticated by using the electrical conductivity parameter. If the electrical conductivity is higher than 800 lS/cm, then the honey is a honeydew one (Bogdanov et al., 2004). The honeys analyzed (9 samples) had higher electrical conductivity than 800 lS/cm. 3.2. Physicochemical parameters The physicochemical parameters studied in the case of the 50 samples of honeys are: moisture content, electrical conductivity, pH, aw, free acidity, refraction index, °Brix concentration, ash, color parameters (L⁄, a⁄, b⁄, chroma, hue angle, yellow index). The values of the physicochemical parameters are shown in the Table 2. The moisture content is one of the most important parameters of honey; the Codex Alimentarius (2001) established a limit of 20% in the case of honey, because at levels higher than 20% fermentation processes are accelerated during storage (Oroian, 2012). In our case, the moisture content of honey ranged between 14.4 – 19.9%, all the samples meeting the threshold established by the Codex Alimentarius (2001). The difference of moisture content according to their origin is a significant one (P < 0.05). The sunflower honeys registered the highest moisture content, while the honeydew samples the smallest levels of moisture content, respectively. Moisture content is influenced by the climatic conditions, degree of maturity and extraction and storage conditions (Fechner et al., 2016). The levels of moisture content of the analyzed honey are in agreement with those reported by Fechner et al. (2016) in the case of Argentinian honeys, Escriche et al. (2014) in the case of Spanish honeys and by Can et al. (2015) in the case of Turkish honeys (see Table 3). Honey is an acidic food produce, which contains organic acids and amino acids and the variation in acidity which depends on the botanical origin may be a result of these two chemical classes

Table 1 Main pollen types in honey samples. Honey type

Samples

Principal pollen type

Secondary pollen types

Other significant pollen types

Acacia

10

Robinia pseudoacacia (48.5%, 45.1–49.6%)

Plantago, Trifoloium and Rubus

Tilia

8

Tilia europea (64.5%, 62.1–69.8%)

Sunflower

11

Helianthus annuus (63.5%, 60.1–68.7%)

Brassica napus (12.2%, 8.8–13.4%) Prunus (5.1%, 2.9–6.8%) Brassica napus (10.2%, 8.8–11.5%) and Helianthus annuus (2.3%, 1.5–3.2%) Tilia (5.9%, 4.2–6.9%)

Polyfloral

12

Brassica napus (29.3%, 21.3–33.8%)

Prunus (7.6%, 6.5–8.3%) and Tilia (6.1%, 5.5–6.8%)

Honeydew

8

Brassica napus (17.2%, 15.1–17.9%)

Quercus (12.3%, 10.1–13.9%) Trifolium repens (11.2%, 8.3–13.5%) Castanea sativa (8.9%, 6.7–10.1%)

Galium and Trifolium Taraxacum officinale, Trifolium, Fragaria, Brassica napus and Robinia pseudoacacia Robinia pseudoacacia, Plantago, Rubus, Taraxacum officinale, Fragaria, Trifoloium europea, Galium Helianthus annuus

151

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156 Table 2 Physicochemical parameters of honey according to botanical origin. Physicochemical parameters

Acaciaa

Tiliaa

Polyflorala

Honeydewa

Sunflowera

F-ratio

Moisture content (g/100 g) pH aw Free acidity (meq acid/k) Refractive index

17.0(16.2–17.9)abc 4.6(4.0–4.8)ab 0.53(0.48–0.55)a 9.08(4.86–13.29)cd 1.4929(1.4870–1.4984) ab 81.2(78.2–84.1)a 156.5(109.9–244.7)d

17.8(16.9–18.8)ab 5.5(5.0–6.4)a 0.54(0.52–0.56)a 6.63(1.91–11.34)d 1.4920(1.4848– 1.4982)a 80.9(77.6–84.1)a 549.3(450.6–647.9)b

17.1(16.3–17.8)bc 4.4(3.9–5.4)b 0.54(0.52–0.55)a 20.83(16.98–24.68)a 1.4938(1.4869– 1.4981)a 78.9(76.3–81.7)a 431.4(350.9–511.9)bc

18.2(17.4–19.0)a 4.9(4.0–4.8)ab 0.55(0.53–0.60)a 13.02(9.00–17.04)bc 1.4907(1.4844– 1.4971)a 80.3(77.5–83.1)a 346.1(261.9–430.2)c

2.95a 2.18 ns 0.38 ns 7.31*** 2.09 ns

0.08(0.03–0.12)d 45.6(43.7–47.6)a 1.0(1.9 to 0.2)d 11.9 (10.7–13.2)c 12.0 (10.9–13.1)b 0.6(3.1 to 1.8)b 37.6(33.2–42.1)c

0.27(0.22–0.32)b 40.9(38.7–43.1)b 0.7 (0.2–1.7)c 14.9(13.5–16.3)ab 14.9(13.7–16.2)a 1.6(4.4 to 1.2)b 52.3(47.3–57.4)ab

0.21(0.17–0.25)bc 39.9(38.1–41.7)b 3.4(2.6–4.2)b 13.9(12.8–15.1)b 14.5(13.5–15.6)a 0.6(2.9 to 1.6)b 49.8(45.7–53.9)b

16.3(15.4–17.2)c 4.9(4.2–5.2)ab 0.54(0.52–0.56)a 16.08(11.63–20.52)ab 1.4947(1.4847– 1.5006)b 82.2(79.1–85.3)a 1007.9(914.9–1100.9) a 0.49(0.45–0.54)a 21.6(19.6–23.7)c 5.8(4.9–6.7)a 6.6(5.3–7.9)d 8.9(7.7–10.0)c 0.5(2.1–3.1)ab 43.0(38.3–47.8)c

0.17(0.13–0.21)c 39.5(37.6–41.4)b 1.2(0.9–2.6)c 15.7(14.5–16.8)a 15.8(14.7–16.9)a 3.5(1.1–5.8)a 57.0(52.8–61.3)a

48.61*** 80.36*** 34.30*** 32.88*** 24.42*** 2.65a 12.01***

°Brix concentration Electrical conductivity (µS/ cm) Ash (%) L* a* b* Chroma Hue angle Yellow index a ***

0.68 ns 49.77***

Average value (minimum – maximum). P < 0.001.

Table 3 Honey phenolics concentrations according to botanical origin.

* **

Phenolics (mg/100 g)

Acacia

Tilia

Polyfloral

Honeydew

Sunflower

F-ratio

Apigenin Caffeic acid Chrysin Galangin Gallic acid Isorhamnetin Kaempherol Luteolin Myricetin p-coumaric acid Pinocembrin Quercetin

0.22(0.06–0.39)a 0.21(0–0.51)bc 0.02(0–0.15)b 0.04(0–0.08)b 0.05(0.01–0.08) 0.01(0–0.03)a 0.16(0–0.35)a 0.02(0–0.04)a 0.11(0.02–0.20)a 0.36(0–1.40)b 0.87(0.09–1.65)b 0.39(0.02–0.77)

0.24(0.06–0.42)a 0.46(0.14–0.80)ab 0.29(0.15–0.43)a 0.02(0–0.07)b 0.06(0.02–0.10) 0.03(0–0.05)a 0.21(0–0.43)a 0.01(0–0.03)a 0.17(0.07–0.28)a 2.91 (1.75–4.08)a 2.31(1.44–3.17)a 0.15(0–0.57)

0.30(0.15–0.45)a 0.02(0–0.29)c 0.06 (0–1.76)b 0.06(0.02–0.10)ab 0.09(0.06–0.12) 0.003(0–0.02)a 0.25(0.07–0.42)a 0.02(0.01–0.04)a 0.14(0.05–0.22)a 0.84(0–1.88)b 1.08(0.37–1.79)b 0.45(0.11–0.79)

0.27(0.10–0.44)a 0.75(0.44–1.06)a 0.02(0–0.16)b 0.11(0.06–0.15)a 0.08(0.04–0.11) 0.03(0–0.05)a 0.21(0–0.41)a 0.03(0.01–0.05)a 0.12(0.025–0.22)a 0.91(0–2.00)b 1.18(0.36–2.00)ab 0.50(0.11–0.90)

0.20(0.05–0.36)a 0.14(0–0.40)bc 0.08(0–0.20)b 0.03(0–0.07)b 0.06(0.03–0.09) 0.03(0–0.05)a 0.31(0.16–0.50)a 0.02(0.01–0.04)a 0.17(0.04–0.08)a 1.06(0.07–2.05)b 1.23(0.49–1.97)ab 0.23(0–0.59)

0.26 ns 3.98** 2.75* 2.28 ns 1.26 ns 1.34 ns 0.35 ns 0.73 ns 0.38 ns 3.03* 1.81 ns 0.59 ns

P < 0.05. P < 0.01.

presented above (Oroian, 2012). In the case of the 50 honey samples analyzed, the pH ranged between 3.9 and 6.4. The pH values are in the same range with those reported for honeys from Algeria (Ouchemoukh et al., 2007), Argentina (Fechner et al., 2016) and South Africa (Chuttong et al., 2016). The highest levels of pH were observed in the case of honeydew and polyfloral honeys while acacia had the lowest levels of pH (table 2). A significant difference (P > 0.05) of pH in function of the botanical origin was not observed. Honey pH is influenced by the extraction and storage conditions and it is not influenced by the botanical origin (Oroian, 2012). Water activity is influenced by the molar concentration of the soluble species present in honey. The substances which have a high molecular mass or which are present in small quantities such as the compounds with nitrogen (proteins, enzymes, amino-acids), acids, vitamins, aroma compounds or minerals do not contribute to the magnitude of water activity (Chirife et al., 2006). Taking into account this fact, it can be concluded that honey water activity is influenced mainly by the fructose and glucose concentrations and to a little extent by the sucrose content (Chirife et al., 2006). The water activity ranged between 0.48–0.60. The values are not influenced by honey origin (P > 0.05). The values are in the same range with those reported in the case of honeys from Argentina (Chirife et al., 2006). Free acidity is an indicator of the deterioration of honey; this parameter is characterized by the presence of organic acids in

equilibrium with internal esters, lactone and some inorganic ions such as sulphates, chlorides and phosphates; in addition, higher values of free acidity may indicate that sugars are transformed into organic acids (da Silva et al., 2016). In terms of free acidity the Codex Alimentarius (2001) established a level of 40 meq acid/kg for polyfloral honeys and 50 meq acid/kg for honeydew honeys respectively. None of the 50 honey samples (table 2) has exceeded the maximum allowable level established by the Codex Alimentarius (2001). The tilia and acacia honeys had the lowest free acidity, while the highest free acidity was observed in the case of polyfloral and honeydew honeys (P < 0.05). Some authors (Fechner et al., 2016) suggest that free acidity can be used at the differentiation of nectar honeys from honeydew honeys, however in this study the polyfloral honeys registered higher values than honeydew ones. The values of free acidity were in the same range with those reported by the Fechner et al. (2016) and Chuttong et al. (2016). The refractive index of honey depends on moisture content, for this reason there is a correlation table between the refractive index of honey and the moisture content (Bogdanov 2002). The refractive index of the honey analyzed ranged between 1.1870–1.5006 (Table 2), in agreement with the literature (Oroian et al., 2013). The honey °Brix concentration, which represents the dissolved solid content mainly sugar compounds ranged between 76.3 and 85.3°Brix (Table 2), in agreement with the literature (Oroian et al., 2013).

152

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156

The values obtained for the electrical conductivity of honeys are shown in the Table 2. They were lower in the case of acacia honeys, while the honeydew samples had higher levels than 1000 lS/cm. The difference of electrical conductivity according to their origin is a significant one (P < 0.05) and can be used in the differentiation of the nectar honeys from honeydew (honey that honeybees produce from secretions from the living parts of plants or excretions of plant sucking insects on the living parts of plants). The honeydew honeys have a higher electrical conductivity because of the high concentrations of ions reflected by the ash content (Table 2). The electrical conductivity of honey is a parameter correlated to the ash content and acidity, revealing the presence of ions, proteins and organic acids (Yücel and Sultanog˘lu, 2013). Alqarni et al. (2014) observed that the electrical conductivity ranged between 0.21 to 3.13 mS/cm, being higher in the case of dark-colored honeys. These values are closed to those reported in our study, and in contrast to floral honeys, the honeydew honeys exhibited higher values which are derived from the high mineral contents (Table 2), being ascertained by studies from literature (Oroian, 2012). The electrical conductivity of the honey samples analyzed were in the same range with those reported by Fechner et al. (2016), Can et al. (2015), Chuttong et al. (2016) and Kaškoniene˙ and Venskutonis (2010). The Codex Alimentarius (2001) has not established a level for the ash content, this one can range between 0.02 and 1.03%. The ash content is expressed by the mineral content present in honey. The honey analyzed had levels from 0.03% to 0.54%; the acacia honeys had the lowest levels while honeydew the highest ones (Table 2). The difference of ash content according to their origin is a significant one (P < 0.05). The ash content is in the same range with those reported by Chuttong et al. (2016). The commercialization of honey is greatly influenced by color. This parameter varies significantly depending on the botanical origin, from light to dark amber or black shades. Exposure to heat and the storage time of honey may also affect honey’s color. Honey color is related to its flavor. Light colored honey is mild whereas darker types have stronger flavors. It is one of the parameters mostly taken into consideration by consumers in terms of quality appreciation and acceptability (da Silva et al., 2016). The color parameters of the honey analyzed are shown in the Table 2. Can et al. (2015) explained that honey color is mostly reliant on nectar sources and pollen contents which contain various color pigments, i.e. anthocyanins, phenolic acids, proanthocyanidins and flavonoids, and mineral constituents. Tilia and acacia had the highest luminosity level (L⁄); high values of L⁄ indicates clearness, while low values of L⁄ indicates darkness. The highest chroma (C⁄) value was observed in the case of sunflower, while the honeydew had the lowest values. The highest yellow index was observed in the case of sunflower. Regarding the color parameters there is a significant difference (P < 0.001) between honey samples. The differences in terms of color between the different honey types are due to the chemical composition and variety (Oroian, 2012). The acacia honey can be considered a light-colored honey, because of the high level of L⁄ in comparison with the other honeys analyzed; the acacia honeys do not have high concentrations of color pigments, minerals and pollens or a high electrical conductivity (Can et al., 2015). Honeydew honeys are dark colored honeys and they are reported to have high levels of pigments, pollen, phenolic compounds, minerals and Maillard reaction products (Tezcan et al., 2011). The differentiation of the honeydew honeys from the nectar honeys can be made using the CIEL⁄a⁄b⁄ parameters because they are placed in another range from the nectar honeys parameter values (Table 2). The color parameters are in agreement with those reported in other studies (Escriche et al., 2014; Oroian, 2012; Can et al., 2015).

4. Phenolics – validation and concentrations The method was validated and carried out according to the criteria established by the European Commission (2002). The parameters taken into account were: linearity, recovery and precision. The linearity of the method was established by using honey matrix spiked with 6 levels of phenolics and then extracted using the procedure described above. Linear calibration curves were constructed from the peak area ratios versus analyte concentrations for phenolics, and versus analyte/internal standard for volatile compounds. Recovery was studied for each flavonoid and phenolic acid fortifying the honey at three concentrations levels (0.02, 0.06 and 0.12 mg/100 g); 6 replicates were analyzed at each level. Recoveries were calculated on the basis of the difference between the total amount determined in the spiked samples and the amount determined in the non-spiked samples divided by the amount added. Likewise, recovery for the phenolics was calculated at three levels (0.12, 0.20 and 0.40 lg/100 g) taking into account the amount of internal standard added to each sample (Escriche et al., 2014). The repeatability was calculated by relative standard deviation (RSD) of six injections at the same level concentration as used before the same day. To evaluate inter-day precision, the analyses were repeated on three consecutive days (Escriche et al., 2014). Myricetin, p-coumaric acid, chrysin, caffeic acid, pinocembrin, luteolin, gallic acid and galangin were identified at 210 nm, while quercetin, apigenin, kaempherol and isorhamnetin were identified at 200 nm. Regarding the validation of parameters, this was good for all the phenolics analyzed, with regression coefficients (R2) higher than 0.994. The average recovery range varied from 97.0% for luteolin to 115.0% for quercetin. For the caffeic acid the average recovery was lower, at 78.0%. The repeatability for all compounds was less than 8% and reproducibility was always less than 12.0%. All the data obtained were in agreement with other authors (Escriche et al., 2014). Blossom honeys are produced from nectars of lower-bearing plants, while honeydew honeys are rather the product of the digestive by-products from phids collected by honeybees (Can et al., 2015). These insects meet their nutritional requirements from phloem, and their digestive by-products are collected by honeybees (Ülgentürk et al., 2013). Twelve phenolics were identified in all five honey types. It can be observed that the abundant phenolic was p-coumaric acid (4.08 mg/100 g) in the case of a tilia sample, followed by pinocembrin (3.17 mg/100 g) in the case of a tilia sample and quercetin (0.90 mg/100 g) in a honeydew sample. Only 3 phenolics (caffeic acid, chrysin and p-coumaric acid) have significant differences in concentrations depending on the botanical origin. The total content of phenolics in tilia honey is the highest from all the honeys analyzed (6.87 mg/100 g), followed by honeydew honeys (4.22 mg/100 g), sunflower honeys (3.88 mg / 100 g), polyfloral (3.42 mg/ 100 g), and acacia honeys (2.47 mg/100 g), respectively. The tilia honeys have 2.78 times more phenolics than acacia honeys. Attempts to apply single phenolics analytical data for authentication of honey type can lead to fallacious results, because quantities of some individual phenolic acids could be present in similar quantities and these are subject to annual variations. All the phenolics were present in all the five honey types analyzed. P-coumaric acid and pinocembrin were in the highest concentration. The comparative ratios of phenolics in honeys could be considered as potential indicators of botanical/floral origin of honey (Jasicka-Misiak et al., 2012). In agreement with the literature (Can et al., 2015), the acacia honeys show high concentration of apigenin, kaempferol and isorhamnetin, however none of these flavonoids is considered a chemical marker for acacia honeys. The polyfloral honeys regis-

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156

tered higher concentration of p-coumaric acid, being in agreement with the literature (Can et al., 2015). The p-coumaric acid concentrations in sunflower honey are in agreement with the literature (Nayik and Nanda, 2016). Kaempferol, pinocembrin and quercetin are reported in most honeys by many authors from different countries (Jasicka-Misiak et al., 2012; Nayik and Nanda, 2016). Caffeic acid, p-coumaric acid, apigenin, kaempferol, chrysin, pinocembrin and galangin concentrations of tilia, acacia and sunflower are in the same range with those reported in the case of Serbian honeys (Kecˇkeš et al., 2013).

5. Multivariate analysis The physicochemical parameters and phenolic concentrations have been submitted to different statistical analyses (Principal component analysis, Linear discriminant analysis and Artificial neural networks) is order to see if the parameters studied can be used for botanical authentication of honey and which of the multivariate analysis applied is much suitable for predicting the membership of a sample to a group.

153

5.2. Linear discriminant analysis The physicochemical parameters and phenolic compound concentrations were submitted to linear discriminant analysis to establish the usefulness of these parameters for the authentication of honeys. The authentications were made by using the phenolic compounds (1st method) and the physicochemical parameters and phenolic compounds (2nd method).

5.2.1. Linear discriminant analysis of honeys based on phenolics The classification of honey samples accordingly to their botanical origin is shown in the Table 4. The classification of honey according to the botanical origin by using phenolic compounds reached a good classification of 58.0% of the samples. Function 1 explains 55.7% of the total variance, while function 2 explains 30.8%. Many samples were classified as polyfloral even if they were of a different botanical origin. 83.3% of the polyfloral samples were correctly classified, while only 27.3% of the sunflower samples were correctly classified, respectively. The LDA projection of honey based on phenolics is shown in Fig. 3; there cannot be observed the grouping of the samples according to their botanical origin, but it can be observed that the samples are mixed.

5.1. Principal component analysis (PCA) A PCA was conducted to evaluate the total effect of the honey type on the physicochemical and phenolic parameters (Figs. 1 and 2). The Fig. 1 shows that five different groups can be observed, each one corresponding to a botanical group (acacia, sunflower, polyfloral, tilia and honeydew). The groups of acacia, honeydew and tilia are very well defined due to their physicochemical and phenolic parameters. Sunflower honeys and polyfloral groups are very closed, because polyfloral honeys have a great variability of pollen types, this type of honey is not well defined. Regarding the correlation loadings presented in Fig. 2, the parameters which are in the outer ellipse contribute more than the parameters which are in the inner ellipse. Phenolic compounds do not influence the projection of the loadings being in the inner ellipse. Regarding the physicochemical parameters, it seems that the a⁄, ash content and electrical conductivity are in opposition with all the rest of the parameters.

5.2.2. Linear discriminant analysis of honeys based on physicochemical parameters and phenolics The classification of honey accordingly to their botanical origin using phenolics and physicochemical parameters is shown in the Table 4. The classification of honey according to the botanical origin by using physicochemical parameters and phenolics reached a good classification of 92.0% of the samples. Accordingly to the data presented in the Table 4, tilia and honeydew were perfectly classified, while the sunflower honeys have been correctly classified in 81.82% of the samples. The linear discriminant analysis applied to all the physicochemical and textural parameters resulted in two canonical functions (Wilks’s Lambda = 0.001, Chisquare = 253.46, df = 72, p < 0.01 for the first function, and Wilks’s Lambda = 0.026, Chi-square = 136.91, df = 51, p < 0.01 for the second function, respectively) with the eigenvalues of 21.379 and 4.728, respectively. Function 1 explains 65.7% of the total variance,

Fig. 1. Principal component analysis – scores: a- acacia, t – tilia, p – polyfloral, s – sunflower, h – honeydew honeys.

154

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156

Fig. 2. Principal component analysis – loadings: M-moisture content, F-fructose content, EC-electrical conductivity, YI – yellow index, G – glucose, S – sucrose, RI – refractive index, F + G – fructose and glucose content, C* – chroma, H – hue angle, My – myricetin, p-C – p – coumaric acid, C – Chrysin, Ca – caffeic acid, P – Pinocembrin, Q – quercetin, A – Apigenin, K – Kaempherol, I – isorhamentin, L – Luteolin, Ga – Gallic acid, Gal – galangin.

Table 4 Classification of honey based on phenolics and physicochemical parameters. Validation – cross validation

Original group

Phenolics

Physicochemical parameters and phenolics

Corect,%

Acacia Tilia Polyfloral Honeydew Sunflower Acacia Tilia Polyfloral Honeydew Sunflower

Acacia

Tilia

Polyfloral

Honydew

Sunflower

5 2 1 0 2 9 0 0 0 1

0 5 0 0 0 0 8 0 0 0

2 0 10 1 5 0 0 11 0 1

2 0 0 6 1 0 0 0 9 0

1 1 1 2 3 1 0 1 0 9

4.0

5.0

3.0

4.0

50.00% 62.50% 83.33% 66.67% 27.27% 90.00% 100% 91.67% 100% 81.82%

3.0

F2 (18.6%)

F2 (32.5%)

2.0 1.0 0.0 -1.0

2.0 1.0 0.0 -1.0 -2.0

-2.0 -3.0 -5.0

-3.0 -4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

-4.0 -6.0

-4.0

-2.0

0.0

Fig. 3. Linear discriminant score plot: acacia, tilia, polyfloral, sunflower honeys projection based on phenolics concentrations.

2.0

4.0

6.0

8.0

10.0

F1 (65.7%)

F1 (53.9%) honeydew,

while function 2 explains 18.6%. Only three samples of the fifty honey samples were not correctly classified. The first discriminate functions were dominated by conductivity (F1 = 1.502, F2 = 0.209) followed by b⁄ (F1 = 1.093, F2 = 2.809) and pH (F1 = 0.644, F2 = 0.374), respectively. The second discriminate functions were dominated by chroma (F1 = 0.978, F2 = 1.562), p-coumaric acid content (F1 = 0.944, F2 = 0.887) and yellow index (F1 = 0.451, F2 = 0.679). The projection of the linear discriminate analysis based on phenolics and physicochemical properties is shown in the Fig. 4. It can

Fig. 4. Linear discriminant score plot: acacia, tilia, polyfloral, honeydew, sunflower honeys projection based on phenolics and physicochemical parameters.

be observed that the groups of acacia, tilia and honeydew are well defined, while the groups of sunflower and polyfloral honeys are intertwined. 5.3. Artificial neural networks The classification of honey based on their physicochemical parameters and phenolics has been made up using artificial neural networks, too. Of all the types of artificial neural networks, the

155

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156 Table 5 Artificial neural networks statistical parameters for the honey classification.

No.

Model Name

Hidden layers

Training

MSE

R2

Cross Validation

MAE

MSE

R2

Testing

MAE

R2

MSE

MAE

1

MLP

1

0.0287

0.992

0.112

0.522

0.888

0.554

0.512

0.854

0.587

2

MLP

2

0.010

0.998

0.055

0.239

0.948

0.386

0.496

0.877

0.535

3

MLP

3

0.045

0.994

0.135

0.441

0.906

0.507

0.486

0.825

0.556

4

PNN

1

0.584

0.869

0.702

1.094

0.732

0.864

0.896

0.675

0.830

5

PNN

2

0.505

0.896

0.638

1.032

0.765

0.841

1.08

0.614

0.864

6

PNN

3

0.543

0.903

0.671

1.250

0.686

0.870

1.042

0.613

0.875

7

RNN RNN

1

0.542

0.867

0.521

0.584

0.886

0.608

0.916

0.736

0.833

2 1.202

0.616

0.736

1.110

0.754

0.883

1.397

0.556

0.971

RNN

3 1.163

0.585

0.848

2.042

0.355

1.143

1.533

0.404

1.007

MNN

1

0.268

0.921

0.379

0.348

0.938

0.497

0.540

0.854

0.579

0.004

0.999

0.036

0.324

0.931

0.414

0.847

0.793

0.679

0.690

0.885

0.651

1.722

0.579

1.055

1.023

0.739

0.791

8 9 10 11 12

MNN MNN

2 3

following methods were chosen: multilayer perceptron (MLP), probabilistic neural network (PNN), recurrent neural network (RNN) and modular neural network (MNN). The data were divided into three categories: training, cross-validation and testing ones. Each category was represented by 17 samples of different origins. The methods were made in three ways: 1 hidden layer, 2 hidden layers and 3 hidden layers respectively. The mean squared error (MSE), mean absolute error (MAE) and regression coefficients (R2) results for each model are shown in the Table 5. Based on the three statistical parameters presented above, the MLP with 2 hidden layers is the suitable method for the classification of honey using artificial networks, based on the physicochemical parameters and phenolics, according to the data presented in the Table 5.

6. Conclusions The physico-chemical parameters and phenolic compounds obtained in the honey analyzed can be used in the classification of acacia, sunflower, tilia, polyfloral and honeydew. Regarding the phenolic compounds, no compound that can be used as a chemical marker has been identified. The application of chemometrics to the physicochemical parameters and phenolic compounds has led to a good classification of the samples according to their botanical origin. In the case of the linear discriminant analysis 92.0% of the samples have been correctly classified using the physico-chemical parameters and phenolic concentration. The MLP with two hidden layers have classified (cross validated) 94.8% of the honey samples based on their botanical origin. The MLP method is a suitable method for the classification of honey samples according to their botanical origin by using the physicochemical parameters and phenolic compounds.

Acknowledgement This work was supported by a grant of the Romanian National Authority for Scientific Research and Innovation, CNCS – UEFISCDI, project number PN-II-RU-TE-2014-4-0110.

References Alqarni, A.S., Owayss, A.A., Mahmoud, A.A., Hannan, M.A., 2014. Mineral content and physical properties of local and imported honeys in Saudi Arabia. J. Saudi Chem. Soc. 18 (5), 618–625.

ˇ eksteryte˙, V., 2007. Radical scavenging activity Baltrušaityte˙, V., Venskutonis, P.R., C of different floral origin honey and beebread phenolic extracts. Food Chem. 101 (2), 502–514. Baroni, M.V., Podio, N.S., Badini, R.G., Inga, M., Ostera, H.A., Cagnoni, M., Wunderlin, D.A., 2015. Linking soil, water, and honey composition to assess the geographical origin of Argentinean honey by multi-elemental and isotopic analyses. J. Agri. Food Chem. 63 (18), 4638–4645. Binetti, G., Del Coco, L., Ragone, R., Zelasco, S., Perri, E., Montemurro, C., Schena, F.P., 2017. Cultivar classification of Apulian olive oils: use of artificial neural networks for comparing NMR, NIR and merceological data. Food Chem. 219, 131–138. Bogdanov, S., Martin, P., Lullmann, C., 2002. Harmonised Methods of the International Honey Commission. Swiss Bee Research Centre, FAM, Liebefeld. Bogdanov, S., Ruoff, K., Oddo, L., 2004. Physico-chemical methods for the characterisation of unifloral honeys: a review. Apidologie 35 (Suppl. 1), S4–S17. Can, Z., Yildiz, O., Sahin, H., Turumtay, E.A., Silici, S., Kolayli, S., 2015. An investigation of Turkish honeys: their physico-chemical properties, antioxidant capacities and phenolic profiles. Food Chem. 180, 133–141. Chirife, J., Zamora, M.C., Motto, A., 2006. The correlation between water activity and % moisture in honey: fundamental aspects and application to Argentine honeys. J. Food Eng. 72 (3), 287–292. Chuttong, B., Chanbang, Y., Sringarm, K., Burgett, M., 2016. Physicochemical profiles of stingless bee (Apidae: Meliponini) honey from South East Asia (Thailand). Food Chem. 192, 149–155. Codex Alimentarius (2001) revised codex standards for honey. Codex standard 12– 1981, Rev. 2. Coneac, G., Gafitßanu, E., Ha˘da˘ruga˘, D.I., Ha˘da˘ruga˘, N.G., Pînzaru, I.A., Bandur, G., Gruia, A., 2008. Flavonoid contents of propolis from the west side of Romania and correlation with the antioxidant activity. Chem. Bull Politehnica Univ. Timisoara 53 (67), 56–60. D’Archivio, A.A., Giannitto, A., Maggi, M.A., Ruggieri, F., 2016. Geographical classification of Italian saffron (Crocus sativus L.) based on chemical constituents determined by high-performance liquid-chromatography and by using linear discriminant analysis. Food Chem. 212, 110–116. da Silva, P.M., Gauche, C., Gonzaga, L.V., Costa, A.C.O., Fett, R., 2016. Honey: chemical composition, stability and authenticity. Food Chem. 196, 309–323. Dinca, O.R., Ionete, R.E., Popescu, R., Costinel, D., Radu, G.L., 2015. Geographical and botanical origin discrimination of Romanian honey using complex stable isotope data and chemometrics. Food Anal. Meth. 8 (2), 401–412. Dobre, I., Georgescu, L.A., Alexe, P., Escuredo, O., Seijo, M.C., 2012. Rheological behavior of different honey types from Romania. Food Res. Int. 49 (1), 126–132. Escriche, I., Kadar, M., Juan-Borrás, M., Domenech, E., 2014. Suitability of antioxidant capacity, flavonoids and phenolic acids for floral authentication of honey. Impact of industrial thermal treatment. Food Chem. 142, 135–143. Commission, European., 2002. Commission decision of 12 August 2002 implementing Council Directive 96/23/EC concerning the performance of analytical methods and the interpretation of results (2002/657/EC). Official J. Eur. Commun. 221, 8. Fechner, D.C., Moresi, A.L., Díaz, J.D.R., Pellerano, R.G., Vazquez, F.A., 2016. Multivariate classification of honeys from Corrientes (Argentina) according to geographical origin based on physicochemical properties. Food Biosci. 15, 49– 54. Guelpa, A., Marini, F., du Plessis, A., Slabbert, R., Manley, M., 2017. Verification of authenticity and fraud detection in South African honey using NIR spectroscopy. Food Control 73, 1388–1396. Jasicka-Misiak, I., Poliwoda, A., Deren´, M., Kafarski, P., 2012. Phenolic compounds and abscisic acid as potential markers for the floral origin of two Polish unifloral honeys. Food Chem. 131 (4), 1149–1156.

156

M. Oroian, S. Ropciuc / Computers and Electronics in Agriculture 138 (2017) 148–156

Juan-Borrás, M., Domenech, E., Hellebrandova, M., Escriche, I., 2014. Effect of country origin on physicochemical, sugar and volatile composition of acacia, sunflower and tilia honeys. Food Res. Int. 60, 86–94. Kara, D., 2009. Evaluation of trace metal concentrations in some herbs and herbal teas by principal component analysis. Food Chem. 114 (1), 347–354. Karabagias, I.K., Vavoura, M.V., Nikolaou, C., Badeka, A.V., Kontakos, S., Kontominas, M.G., 2014. Floral authentication of Greek unifloral honeys based on the combination of phenolic compounds, physicochemical parameters and chemometrics. Food Res. Int. 62, 753–760. Kaškoniene˙, V., Venskutonis, P.R., 2010. Floral markers in honey of various botanical and geographic origins: a review. Comp. Rev. Food Sci. Food 9 (6), 620–634. Kecˇkeš, S., Gašic´, U., Velicˇkovic´, T.C´., Milojkovic´-Opsenica, D., Natic´, M., Tešic´, Zˇ., 2013. The determination of phenolic profiles of Serbian unifloral honeys using ultra-high-performance liquid chromatography/high resolution accurate mass spectrometry. Food Chem. 138 (1), 32–40. Louveaux, J., Maurizio, A., Vorwohl, G., 1970. Commision internationale de botanique apicole de L’U.I.S.B. les methodes of melisso-palynologie. Apidologie 1 (2), 211–227. Marghitas, L.A., Dezmirean, D.S., Pocol, C.B., Marioara, I.L.E.A., Bobis, O., Gergen, I., 2010. The development of a biochemical profile of acacia honey by identifying biochemical determinants of its quality. Not Bot Horti Agrobo 38 (2), 84. Mattonai, M., Parri, E., Querci, D., Degano, I., Ribechini, E., 2016. Development and validation of an HPLC-DAD and HPLC/ESI-MS 2 method for the determination of polyphenols in monofloral honeys from Tuscany (Italy). Microchem. J. 126, 220– 229. Mellen, M., Fikselová, M., Mendelová, A., Hašcˇík, P., 2015. Antioxidant effect of natural honeys affected by their source and origin. Pol. J. Food Nutr. Sci. 65 (2), 81–85. Nayik, G.A., Nanda, V., 2016. A chemometric approach to evaluate the phenolic compounds, antioxidant activity and mineral content of different unifloral honey types from Kashmir India. LWT Food Sci. Technol. http://dx.doi.org/ 10.1016/j.lwt.2016.08.016. Oroian, M., 2012. Physico-chemical and rheological properties of Romanian honeys. Food Biophy. 7 (4), 296–307. Oroian, M., Amariei, S., Escriche, I., Gutt, G., 2013. Rheological aspects of Spanish honeys. Food Bio Tech. 6 (1), 228–241. Oroian, M., Amariei, S., Rosu, A., Gutt, G., 2015a. Classification of unifloral honeys using multivariate analysis. J. Essent. Oil Res. 27 (6), 533–544.

Oroian, M., Amariei, S., Leahu, A., Gutt, G., 2015b. Multi-element composition of honey as a suitable tool for its authenticity analysis. Pol. J. Food Nutr. Sci. 65 (2), 93–100. Ouchemoukh, S., Louaileche, H., Schweitzer, P., 2007. Physico-chemical characteristics and pollen spectrum of some Algerian honeys. Food Control 18 (1), 52–58. Piljac-Zˇegarac, J., Stipcˇevic´, T., Belšcˇak, A., 2009. Antioxidant properties and phenolic content of different floral origin honeys. J. ApiProd. ApiMed. Sci. 1 (2), 43–50. Ramzi, M., Kashaninejad, M., Salehi, F., Mahoonak, A.R.S., Razavi, S.M.A., 2015. Modeling of rheological behavior of honey using genetic algorithm–artificial neural network and adaptive neuro-fuzzy inference system. Food Biosci. 9, 60– 67. Siddiqui, A.J., Musharraf, S.G., Choudhary, M.I., 2017. Application of analytical methods in authentication and adulteration of honey. Food Chem. 217, 687– 698. Solayman, M., Islam, M., Paul, S., Ali, Y., Khalil, M., Alam, N., Gan, S.H., 2016. Physicochemical properties, minerals, trace elements, and heavy metals in honey of different origins: a comprehensive review. Comp. Rev. Food Sci. Food 15 (1), 219–233. Tezcan, F., Kolayli, S., Ulusoy, H., Erim, F.B., 2011. Evaluation of organic acid, saccharide composition and antioxidant properties of some authentic Turkish honeys. J. Food Nutr. Res. 50, 33–40. Trafialek, J., Laskowski, W., Kolanowski, W., 2015. The use of Kohonen’s artificial neural networks for analyzing the results of HACCP system declarative survey. Food Control 51, 263–269. Ülgentürk, S., Szentkirályi, F., Uygun, N., Fent, M., Gaimari, S.D., Civelek, H., Ayhan, B., 2013. Predators of Marchalina hellenica (Hemiptera: Marchalinidae) on pine forests in Turkey. Phytoparasitica 41 (5), 529–537. Ulloa, P.A., Guerra, R., Cavaco, A.M., da Costa, A.M.R., Figueira, A.C., Brigas, A.F., 2013. Determination of the botanical origin of honey by sensor fusion of impedance etongue and optical spectroscopy. Comput. Electron Agr. 94, 1–11. Wei, Z., Wang, J., 2014. Tracing floral and geographical origins of honeys by potentiometric and voltammetric electronic tongue. Comput. Electron Agr. 108, 112–122. Yücel, Y., Sultanog˘lu, P., 2013. Characterization of Hatay honeys according to their multi-element analysis using ICP-OES combined with chemometrics. Food Chem. 140 (1), 231–237.