Root Mean Square Error Of Cross-validation Research Articles

Prediction of formaldehyde and residual methanol concentration in formalin using near infrared spectroscopy

Amino resins are produced by two main processes: the strong acid process and the alkaline-acid process. Both use formaldehyde and a base (e.g. sodium hydroxide) in their formulation. In this work, Forward Interval Partial Least Squares methodology was applied to create prediction models for the determination of the concentration of formaldehyde and residual methanol (that is present in the formaldehyde solution) used in the production of amino resins. Near infrared (NIR) spectra were acquired at two different temperatures: 18 and 35°C. A Partial Least Squares calibration models were established with the measured values from reference methods: namely, sodium sulfite (formaldehyde) and gas chromatography (methanol). The performances of the best models were compared using the root mean square error of cross validation (RMSECV) and coefficient of determination for prediction (r2). The best results obtained a r2 above 0.994. The RMSECV values obtained were 0.063% (m/m) and 0.031% (m/m) for the formaldehyde and methanol concentration, respectively. External validation was performed using different formaldehyde solution samples. The NIR methodology presented in this work proved to be effective and enables a significant time reduction, when compared to the reference methods, in the measurement of formaldehyde and methanol concentrations.

Evaluation of Kraft pulp yield and syringyl/guaiacyl ratio from standing trees (Eucalyptus camaldulensis, E. urophylla, Leucaena leucocephala and Casuarina junghuhniana) using portable near infrared spectroscopy

Measurement of pulpwood traits from a standing tree has considerable advantage when screening large populations for tree selection. It reduces time and also eliminates requirements of transport, powdering, and storing the sample. This study describes estimation of Kraft pulp yield (KPY) in Eucalyptus camaldulensis, E. urophylla, Leucaena leucocephala, and Casuarina junghuhniana by portable NIR spectroscopy of standing trees. Calibration models were developed for KPY estimation using portable NIR spectroscopy for the four species, along with a calibration model for syringyl/guaiacyl (S/G) ratio in E. camaldulensis. The calibration models for KPY showed R2 values ranging from 0.93 ( E. camaldulensis) to 0.83 ( L. leucocephala), and 0.95 for S/G ratio. The developed calibration models for E. camaldulensis and L. leucocephala were compared with laboratory NIR models, and a variation of <±2.0% was found between both methods. The models were validated by both external and cross validation which showed <2.0% RMSEP (root mean square error of prediction) and <2.0% RMECV (root mean square error of cross validation) in external and cross validations, respectively.

Chemometrics-Assisted Spectrophotometric Method Development and Validation for Simultaneous Estimation of Emtricitabine, Tenofovir Alafenamide Fumarate, and Dolutegravir Sodium in Dosage Form

Aim: This study aims on the development of a chemometric-assisted spectroscopic method for the analysis of combined dosage form of emtricitabine (EMT), tenofovir alafenamide fumarate (TEN), and dolutegravir sodium (DOL). The use of a multivariate algorithm to analyse spectrophotometric data is a novel approach to estimating drug concentrations in formulations. Materials and Methods: The quantitative estimation of EMT, TEN, and DOL in tablets was carried out using four chemometric approaches: Classical least square (CLS), inverse least square, partial least square, and principal component regression. Thirty-two ternary mixtures of calibration sets and 16 mixtures of validation sets were prepared. The absorbance data matrix was attained by calculating absorbance at 25 different wavelengths in a range of 240–336 nm (Δλ = 4 nm). The chemometric calculations were performed using Matlab2018a and Minitab software. The developed methods were validated. Results: The great accuracy of the current study was justified by the near-perfect recovery values (100%) and low standard deviation. For chemometrics approaches, the root mean square error of calibration (RMSEC), root mean square error of prediction (RMSEP), and root mean square error of cross-validation (RMSECV) outcomes display decent accuracy and precision. Conclusion: The CLS approach yielded the lowest predicted residual error sum of squares, RMSEC, RMSEP, and RMSECV scores. As a result, CLS might be regarded as the best chemometric approach among all techniques utilized. The label claim determined is in excellent accordance with the mean recoveries for EMT, TEN, and DOL. So, it can be used in quality control laboratories.

Simultaneous Quantification of Amorphous and Crystalline Valsartan in Tablets Using Raman Spectroscopy and Chemometrics Tools

Valsartan is an antihypertensive active pharmaceutical ingredient (API), it is used in the amorphous state in the commercial products. As amorphous materials are metastable, amorphous valsartan can crystallize to valsartan E, promoting changes in the dissolution and bioavailability of the drug. Tablets containing metastable forms of APIs need special conditions for transport and storage in order to avoid crystallization (from amorphous state) or polymorphic transitions (from less stable crystalline structures). A multivariate calibration model based on interval partial least squares (iPLS) regression allied to net analyte signal (NAS) algorithm was built to simultaneously quantify amorphous (VAL-AM) and crystalline (VAL-E) valsartan. Mixtures of VAL-AM and VAL-E were used to produce tablets in order to simulate the crystallization of VAL-AM in a range from 0 to 100% of conversion. The calibration set included 11 samples and 5 samples were used as the external validation set. The following parameters of merit (POM) were obtained for both polymorphs in order to evaluate the calibration model quality: root mean square error (RMSE) for cross validation (RMSECV), validation (RMSEV) and calibration (RMSEC), sensitivity (SEN), selectivity (SEL), analytical sensitivity (γ), inverse analytical sensitivity (γ-1), limit of detection (LOD) and limit of quantification (LOQ).

Rapid determination of adulteration of clove essential oil with benzyl alcohol and ethyl acetate: Towards quality control analysis by FTIR with chemometrics

In this study, Fourier transform infrared (FTIR) spectroscopy in tandem with chemometrics was used for the discrimination and quantification of the adulterants, such as benzyl alcohol (B) and ethyl acetate (E) in clove essential oil (CO). Different multivariate models with various spectral derivatization methods were developed and their analysis abilities the adulterants were compared using statistical quality parameters. To discriminate the adulterations thanks to the FTIR data, 130 chemometric models were built by principal component analysis (PCA) algorithm. The statistical performances of the PCA models developed were evaluated by the number of samples outside of explained variance (95 %) and eigen value. To quantify the adulterant concentrations in CO samples, 117 partial least squares (PLS) regression models employing the FTIR data were developed. To find out the best PLS model, the root mean square error of prediction (RMSEP), and root mean square error of cross-validation (RMSECV) were mainly used. Root mean square error of calibration (RMSEC) and R-square were also evaluated. The best discrimination results were achieved using 1st derivative spectra in the region 3500-3100 cm−1 (SRB1) and 2nd derivative spectra in the region 1077-1008 cm-1 (SRE3) for the adulterants of B and E, respectively. The best PLS calibration results were obtained from the combinations of the normal spectra in the regions 3500-3100 cm−1, 1027-993 cm−1, and 756-569 cm−1 (SRBC) and SRE3 for the quantification of B and E, respectively. The results of the study indicated that FTIR with chemometrics could be used for simultaneously discrimination and quantification of the adulterants of B and E in COs without using any toxic chemicals or pretreatments.

Transfer of gas chromatographic retention data among poly(siloxane) columns by quantitative structure-retention relationships based on molecular descriptors of both solutes and stationary phases

In the present study, computational molecular descriptors of 90 saturated esters and seven poly(siloxane) stationary phases with different polarity (SE-30, OV-7, DC-710, OV-25, XE-60, OV-225 and Silar-5CP) were combined into quantitative structure-retention relationship (QSRR) models aimed at predicting the Kováts retention indices (RIs) of the solutes. The molecular descriptors (174) of the stationary phases included in the models were computed using Dragon software from poly(siloxane) oligomers made of 20 siloxane units reflecting the nominal composition of the stationary phase, whereas 439 molecular descriptors were adopted to represent the esters. Different QSRR models were generated by means of Partial Least Squares (PLS) regression to assess the accuracy of this approach in predicting the RIs of unexplored solutes both in known and external stationary phases. After calibration of each PLS model, the descriptors were selected/discarded according to their relevance, evaluated by Covariance Selection (CovSel), and the PLS models were re-built, which resulted in a noticeable improvement of their predictive ability. Firstly, all the available data were equally divided into a training and a test set; the model built on the calibration set was used to predict the RIs of the validation observations. Successively, seven diverse PLS models were created following a “leave-one-column-out” fashion procedure, each one finalized to the estimation of the RIs of the 90 esters associated with a single stationary phase, whereas the calibration model was calculated on the remaining data. All the estimated models provided successful results on the external stationary phase, and predictive performance further increased after variable selection based on CovSel analysis. The final models provided a Root Mean Square Error in Cross Validation (RMSECV) in the range 12–20, a Root Mean Square Error in Prediction (RMSEP) in the range 11–26, and Mean Absolute Percentage Errors in Prediction (MAMEPs) in the range 0.7–1.5, revealing accurate cross-column prediction. Eventually, to test the robustness of the proposed approach, the 90 solutes were equally partitioned into a calibration and a test set and two further QSSR strategies were applied. The first PLS model was calibrated on all the seven stationary phases and the RIs of the 45 external solutes in the same seven columns were simultaneously predicted. The last QSRR approach followed a “leave-one-column-out” scheme and RI of 45 test solutes on an external stationary phase was predicted by a PLS model calibrated with the data of the 45 remaining solutes and the six left stationary phases. After selection of the significant molecular descriptors, PLS regression provided RMSECV values in the range 6–19, RMSEPs in the range 10–14, and MAPEPs in the range 0.9–2.4, revealing the suitability of the approach to deduce the RI of unknown solutes in uncharted stationary phases.

Non-destructive detection of moisture content for Ginkgo biloba fruit with terahertz spectrum and image: A preliminary study

Ginkgo biloba fruit is a typical recalcitrant seed, which is apt to dehydrate and harden, thus losing its activity, edibility and medicinal value during storage. The non-destructive detection method was investigated for monitoring moisture content (MC) variation of Ginkgo biloba fruit based on the sensitivity of terahertz (THz) wave to liquid water. The THz spectra and images of intact Ginkgo biloba seeds and their slices with different MC were collected. The influence of MC on THz wave was analyzed through the time domain spectra and frequency signal of absorption coefficients. The seed with higher MC achieved lower amplitude of time domain spectrum and longer flight time because of strong absorption of MC to THz radiation. The exponential equations were developed between MC and absorbance coefficients at characteristic frequencies of 1.68, 2.66, 2.79 and 3.02 THz with determination coefficients (R2) of 0.99 and root mean square error of cross validation (RMSECV) of 0.3–0.4 %, respectively. In addition, the gap between seed kernel and shell appeared because the seed kernel shrank due to water loss with the decrease of MC, and the model between the gap ratio and MC was established by THz image with R2 of 0.95 and RMSECV of 4.06 %, respectively. The results indicated that THz spectroscopy and image provided a new method for monitoring MC of Ginkgo biloba seed.

Rapid prediction method of α-Glycosidase inhibitory activity of Coreopsis tinctoria extract from different habitats by near infrared spectroscopy

α-Glucosidase is one of the main enzymes causing elevated blood glucose, and Coreopsis tinctoria extract can be used as a natural inhibitor of α-Glucosidase. Therefore, a new method was proposed for predicting the inhibitory activity on α-Glucosidase of Coreopsis tinctoria extract based on near infrared spectroscopy. The absorbance of the inhibitory system was measured by ultraviolet spectroscopy, which was used to study the inhibitory activity on a-glucosidase of Coreopsis tinctoria extract. The near infrared spectra of the solid samples were collected. By selecting spectral preprocessing and optimizing spectral bands, a rapid prediction model of the inhibitory activity was established by partial least squares regression. The root mean square error of cross-validation (RMSECV), correlation coefficient (R) value and the ratio of prediction to deviation (RPD) value were used as indicators of the evaluation model. The near infrared spectrum model was established by combining the best spectral preprocessing of the continuous wavelet transform (CWT) and the best spectral band. The root mean square error of cross-validation (RMSECV) of this model was 0.815%, the correlation coefficient (R) value was 0.942, and the ratio of prediction to deviation (RPD) was 3.0. The root mean square error of prediction (RMSEP) of the model by prediction set was 0.819%, the correlation coefficient (R) value was 0.950, and the RPD was 3.2. The model shows that the fitting relationship between the predicted inhibition value and the reference inhibition value of the near infrared spectral model is good. The results showed that there was a good correlation between near infrared spectroscopy and the inhibitory activity of Coreopsis tinctoria extract. Thus, the established model was robust and effective and could be used for rapid quantification of α-Glucosidase inhibitory activity. The prediction method is simple and rapid, and can be extended to study the inhibition of other medicinal plants.

Estimation of Soil Nutrient Content Using Hyperspectral Data

Soil nutrients play a vital role in plant growth and thus the rapid acquisition of soil nutrient content is of great significance for agricultural sustainable development. Hyperspectral remote-sensing techniques allow for the quick monitoring of soil nutrients. However, at present, obtaining accurate estimates proves to be difficult due to the weak spectral features of soil nutrients and the low accuracy of soil nutrient estimation models. This study proposed a new method to improve soil nutrient estimation. Firstly, for obtaining characteristic variables, we employed partial least squares regression (PLSR) fit degree to select an optimal screening algorithm from three algorithms (Pearson correlation coefficient, PCC; least absolute shrinkage and selection operator, LASSO; and gradient boosting decision tree, GBDT). Secondly, linear (multi-linear regression, MLR; ridge regression, RR) and nonlinear (support vector machine, SVM; and back propagation neural network with genetic algorithm optimization, GABP) algorithms with 10-fold cross-validation were implemented to determine the most accurate model for estimating soil total nitrogen (TN), total phosphorus (TP), and total potassium (TK) contents. Finally, the new method was used to map the soil TK content at a regional scale using the soil component spectral variables retrieved by the fully constrained least squares (FCLS) method based on an image from the HuanJing-1A Hyperspectral Imager (HJ-1A HSI) of the Conghua District of Guangzhou, China. The results identified the GBDT-GABP was observed as the most accurate estimation method of soil TN ( of 0.69, the root mean square error of cross-validation (RMSECV) of 0.35 g kg−1 and ratio of performance to interquartile range (RPIQ) of 2.03) and TP ( of 0.73, RMSECV of 0.30 g kg−1 and RPIQ = 2.10), and the LASSO-GABP proved to be optimal for soil TK estimations ( of 0.82, RMSECV of 3.39 g kg−1 and RPIQ = 3.57). Additionally, the highly accurate LASSO-GABP-estimated soil TK (R2 = 0.79) reveals the feasibility of the LASSO-GABP method to retrieve soil TK content at the regional scale.

Near Infrared Spectroscopic Evaluation of Starch Properties of Diverse Sorghum Populations

Starch, mainly composed of amylose and amylopectin, is the major nutrient in grain sorghum. Amylose and amylopectin composition affects the starch properties of sorghum flour which in turn determine the suitability of sorghum grains for various end uses. Partial least squares regression models on near infrared (NIR) spectra were developed to estimate starch and amylose contents in intact grain sorghum samples. Sorghum starch calibration model with a coefficient of determination (R2) = 0.87, root mean square error of cross validation (RMSECV) = 1.57% and slope = 0.89 predicted the starch content of validation set with R2 = 0.76, root mean square error of prediction (RMSEP) = 2.13%, slope = 0.93 and bias = 0.20%. Amylose calibration model with R2 = 0.84, RMSECV = 2.96% and slope = 0.86 predicted the amylose content in validation samples with R2 = 0.76, RMSEP = 2.60%, slope = 0.98 and bias = −0.44%. Final starch and amylose cross validated calibration models were constructed combining respective calibration and validation sets and used to predict starch and amylose contents in 1337 grain samples from two diverse sorghum populations. Protein and moisture contents of the samples were determined using previously tested NIR spectroscopy models. The distribution of starch and protein contents in the samples of low amylose (<5%) and normal amylose (>15%) and the overall relationship between starch and protein contents of the sorghum populations were investigated. Percent starch and protein were negatively correlated, low amylose lines tended to have lower starch and higher protein contents than lines with high amylose. The results showed that NIR spectroscopy of whole grain can be used as a high throughput pre-screening method to identify sorghum germplasm with specific starch quality traits to develop hybrids for various end uses.

A random-forest-assisted artificial-neural-network method for analysis of steel using laser-induced breakdown spectroscopy

A random-forest (RF)-assisted artificial-neural-network (ANN) method (RF-ANN method) was developed to improve the analytical performance of laser-induced breakdown spectroscopy (LIBS) in evaluating the composition of low-alloy steel. The RF algorithm was employed to identify the weight values of spectral variables in the quantitative analysis, and the weight values were used as the criteria to select the spectral variables. The selected variables were then used as inputs to train the ANN, which was then used as the analysis model. Compared to the calibration curve method, the RF-ANN method performs well in nonlinear fitting and utilizes an abundance of spectral information; moreover, it can accommodate large amounts of plasma data via filter spectral variables, thus overcoming the problem of the ANN training speed being too slow for practical use due to the redundancy in spectral variables. The results revealed an improvement in the analytical performance of LIBS using the RF-ANN method to predict the concentrations of chromium (Cr) and vanadium (V) in low-alloy steel samples. Specifically, the root-mean-square errors of cross-validation (RMSECV) for Cr and V were reduced from 0.060 and 0.042 wt% using the calibration curve method to 0.021 and 0.007 wt%, respectively, using the RF-ANN method.

Self-organizing maps-based generalized feature set selection for model adaption without reference data for batch process

When fourier transform infrared spectroscopy (FTIR) techniques combined with multivariate calibration are used to measure the key process features or analyte concentrations during batch process, model adaption is indispensable for maintaining the predictability of a primary calibration model in new secondary batches. Many model adaption methods conforming to the actual application scenario of batch process have been proposed. Here we report on a novel standard-free model adaption method without reference measurement called variable selection strategy with self-organizing maps (VSSOM). It uses self-organizing maps (SOM) to classify the whole spectral variables into multiple classes according to the spectra from primary batch and secondary batch, respectively; and the corresponding primary feature subsets and secondary feature subsets are formed firstly. Secondly, candidate feature subsets without empty elements are generated by operating intersection between any primary feature subsets and any secondary feature subsets. Thirdly, the candidate feature subset with minimum root mean square error of cross-validation (RMSECV) for the primary calibration set is selected as the optimal feature subset. In this manner, the optimal feature subset can be identified from the candidate feature subsets. In other words, VSSOM aims to create a stable and consistent feature subset across different batches provided that it selects better features within the intersection sets between primary feature subsets and any secondary feature subsets. Two batch process datasets (γ-polyglutamic acid fermentation and paeoniflorin extraction) are presented for comparing the VSSOM method with No transfer partial least squares (PLS), boxcar signal transfer (BST), successive projection algorithm (SPA), transfer component analysis (TCA) and domain-invariant iterative partial least squares (DIPALS). Experimental results show that VSSOM has superior performance and comparable prediction performance in all the scenarios.

Rapid determination of diesel fuel properties by near-infrared spectroscopy

The properties of middle distillates, which form diesel fuels, must be monitored during production according to the specific requirements established by standard analytical test methods. Although these test methods are accurate and well accepted, they are time-consuming and expensive. Hence, the best solution is to apply a method that reduces investment, maintenance costs, and time.The aim of this study was to determine seven properties of middle distillates (kinematic viscosity, cold filter plugging point, pour point, contents of sulfur, mono-, di-, and poly-aromatics) using rapid and inexpensive method, near-infrared (NIR) spectroscopy combined with chemometrics multivariate calibrations and the partial least squares (PLS) algorithm. The root mean square error of calibration (RMSEC) and root mean square error of cross-validation (RMSECV) were applied as basic chemometric diagnostic tools to optimize the NIR model predictions. We confidently predicted satisfactory results: the calibration samples were equally distributed along the line of the NIR models and observed a significant correlation between the reference values obtained by the standard methods and the calculated values by the NIR model. Therefore, NIR spectroscopy is an efficient tool for simultaneously determining several diesel fuel properties, which makes it an excellent option for online monitoring during fuel production.

Seasonal climate signals preserved in biochemical varves: insights from novel high-resolution sediment scanning techniques

Abstract. Varved lake sediments are exceptional archives of paleoclimatic information due to their precise chronological control and annual resolution. However, quantitative paleoclimate reconstructions based on the biogeochemical composition of biochemical varves are extremely rare, mainly because the climate–proxy relationships are complex and obtaining biogeochemical proxy data at very high (annual) resolution is difficult. Recent developments in high-resolution hyperspectral imaging (HSI) of sedimentary pigment biomarkers combined with micro X-ray fluorescence (µXRF) elemental mapping make it possible to measure the structure and composition of varves at unprecedented resolution. This provides opportunities to explore seasonal climate signals preserved in biochemical varves and, thus, assess the potential for annual-resolution climate reconstruction from biochemical varves. Here, we present a geochemical dataset including HSI-inferred sedimentary pigments and µXRF-inferred elements at very high spatial resolution (60 µm, i.e. > 100 data points per varve year) in varved sediments of Lake Żabińskie, Poland, over the period 1966–2019 CE. We compare these data with local meteorological observations to explore and quantify how changing seasonal meteorological conditions influenced sediment composition and varve formation processes. Based on the dissimilarity of within-varve multivariate geochemical time series, we classified varves into four types. Multivariate analysis of variance shows that these four varve types were formed in years with significantly different seasonal meteorological conditions. Generalized additive models (GAMs) were used to infer seasonal climate conditions based on sedimentary variables. Spring and summer (MAMJJA) temperatures were predicted using Ti and total C (Radj2=0.55; cross-validated root mean square error (CV-RMSE) = 0.7 ∘C, 14.4 %). Windy days from March to December (mean daily wind speed > 7 m s−1) were predicted using mass accumulation rate (MAR) and Si (Radj2=0.48; CV-RMSE = 19.0 %). This study demonstrates that high-resolution scanning techniques are promising tools to improve our understanding of varve formation processes and climate–proxy relationships in biochemical varves. This knowledge is the basis for quantitative high-resolution paleoclimate reconstructions, and here we provide examples of calibration and validation of annual-resolution seasonal weather inference from varve biogeochemical data.

On-Line Real-Time Monitoring of a Rapid Enzymatic Oil Degumming Process: A Feasibility Study Using Free-Run Near-Infrared Spectroscopy.

Enzymatic degumming is a well established process in vegetable oil refinement, resulting in higher oil yield and a more stable downstream processing compared to traditional degumming methods using acid and water. During the reaction, phospholipids in the oil are hydrolyzed to free fatty acids and lyso-phospholipids. The process is typically monitored by off-line laboratory measurements of the free fatty acid content in the oil, and there is a demand for an automated on-line monitoring strategy to increase both yield and understanding of the process dynamics. This paper investigates the option of using Near-Infrared spectroscopy (NIRS) to monitor the enzymatic degumming reaction. A new method for balancing spectral noise and keeping the chemical information in the spectra obtained from a rapid changing chemical process is suggested. The effect of a varying measurement averaging window width (0 to 300 s), preprocessing method and variable selection algorithm is evaluated, aiming to obtain the most accurate and robust calibration model for prediction of the free fatty acid content (% (w/w)). The optimal Partial Least Squares (PLS) model includes eight wavelength variables, as found by rPLS (recursive PLS) calibration, and yields an RMSECV (Root Mean Square Error of Cross Validation) of 0.05% (w/w) free fatty acid using five latent variables.

Determination of starch and amylose contents in various cereals using common model of near-infrared reflectance spectroscopy

Near-infrared reflectance spectroscopy (NIRS) was used to determine the total starch and amylose contents in various kinds of cereals namely wheat, waxy rice, non-waxy rice, millet, sorghum, waxy maize, buckwheat, barley, and hulless oat. The partial least-squares (PLS) analysis and principal component regression (PCR) were used to establish the calibration models. PLS model achieved a better effect than PCR at 1100 - 2500 nm, and the coefficient of determination (R2) of the calibration and prediction sets were both higher than 0.9 after the best pre-treatment method, first derivative plus Savitzky-Golay. Additionally, the root mean square error (RMSE) was lower than 2.50, and the root mean square error of cross-validation (RMSECV) was less than 3.50 for starch. By comparing PLS models at different waveband regions, the optimal determination results for starch and amylose were obtained at 1923 - 1961 and 1724 - 1818 nm, respectively. NIRS was found to be a successful method to determine of the starch and amylose contents in various cereals.

Characterization of Synthetic Non-Metallic Inclusions Consisting of TiN, Ti2O3, and Oxides of Al2O3 and MgO·Al2O3 Spinel Using Raman Spectroscopy

Steel quality and properties can be affected by the formation of complex inclusions, including Ti-based inclusions such as TiN and Ti2O3 and oxides like Al2O3 and MgO·Al2O3 (MA). This study assessed the prospective use of Raman spectroscopy to characterize synthetic binary inclusion samples of TiN–Al2O3, TiN–MA, Ti2O3–MA, and Ti2O3–Al2O3 with varying phase fractions. The relative intensities of the Raman peaks were used for qualitative evaluation and linear regression calibration models were used for the quantitative prediction of individual phases. The model performance was evaluated with root mean square error of cross-validation (RMSECV) and root mean square error of prediction (RMSEP). For the raw Raman spectra data, R2 values were between 0.48–0.98, the RMSECV values varied between 3.26–14.60 wt%, and the RMSEP ranged between 2.98–15.01 wt% for estimating the phases. The SNV Raman spectra data had estimated R2 values within 0.94–0.99 and RMSECV and RMSEP values ranged between 2.50–3.26 wt% and 2.80–9.01 wt%, respectively, showing improved model performance. The study shows that the specific phases of TiN, Al2O3, MA, and Ti2O3 in synthetic inclusion mixtures of TiN–(Al2O3 or MA) and Ti2O3–(Al2O3 or MA) could be characterized by the Raman spectroscopy.

Avocado dehydration negatively affects the performance of visible and near-infrared spectroscopy models for dry matter prediction

This study aims to test the hypothesis that skin dehydration can cause the development of cork-like layers in the avocado fruit skin which may negatively affect Vis-NIR spectroscopy. To test this, dehydration treatment was applied on avocado fruit by storing them at low relative humidity (RH) during ripening treatment. Furthermore, to demonstrate that the hypothesis was not only valid for a single instrument and in general valid for any type of Vis-NIR instrument the avocados were also measured with two different spectrometers i.e., lab-based, and hand-held. Since the two instruments have two different measurement geometries i.e., diffuse reflection and interaction, the study also tests which geometry was best for the measurement of DMC in dehydrated avocados. The results showed that the dehydration of avocado fruit negatively affects the performance of Vis-NIR calibrations compared to the non-dehydrated fruit. The root mean squared error of cross-validation (RMSEcv) on internal test set for dehydrated and non-dehydrated fruit were up to 1.49 % dw/fw and 1.02 % dw/fw, respectively. The hypothesis was true for both lab-based and hand-held instruments, and the root mean squared error of prediction on internal test set were up to 28 % higher for dehydrated fruits. The performance of interaction measurement mode was better (RMSEcv = 0.98 % dw/fw) than the diffuse reflection mode (RMSEcv = 1.21 % dw/fw) for non-dehydrated fruit, however, both modes achieved similar performance (RMSEcv = ∼1.42 % dw/fw) for dehydrated fruit. The poorer performance of Vis-NIR models on dehydrated avocado fruit can be accepted as a limitation of Vis-NIR spectroscopy for avocado fruit analysis.

Near-infrared spectroscopy and machine learning-based classification and calibration methods in detection and measurement of anionic surfactant in milk

The monitoring of milk and milk-based products are crucial in the dairy industry because of their global consumption, fraudulent practice applied to mask the quality parameters for gaining more profit. Non-destructive near-infrared (NIR) spectroscopy and machine learning (ML) methods like interval partial least squares (iPLS), principal components analysis (PCA), and soft independent modeling of class analogy (SIMCA) were used to classify and discriminate pure milk samples from adulterated with an anionic surfactant. The surfactant’s model was sodium dodecylbenzene sulfate (SDBS) in different concentrations (1.94–19.4 gkg−1). ML models were assessed by sensitivity, specificity and efficiency rates as classification figures of merit. As a result, iPLS regression method based on standard normal variate (SNV) pre-processed spectra provided better performance compared to partial least squares (PLS) regression with a full-spectrum dataset. The root mean square error of cross-validation (RMSECV) was 0.123 and 0.135 for the iPLS and full-spectrum models, respectively. PCA as a qualitative grouping analysis demonstrated separate clusters of samples in terms of their concentration level of SDBS. For the iPLS dataset in SIMCA model, all classification figures of merit were recorded 100 % for pure milk samples. In contrast, the average sensitivity, specificity and efficiency rates were 97.6 %, 93 %, and 94.9 % for adulterated samples.

A National-Scale 1-km Resolution PM2.5 Estimation Model over Japan Using MAIAC AOD and a Two-Stage Random Forest Model

Satellite-based models for estimating concentrations of particulate matter with an aerodynamic diameter less than 2.5 μm (PM2.5) have seldom been developed in islands with complex topography over the monsoon area, where the transport of PM2.5 is influenced by both the synoptic-scale winds and local-scale circulations compared with the continental regions. We validated Multi-Angle Implementation of Atmospheric Correction (MAIAC) aerosol optical depth (AOD) with ground observations in Japan and developed a 1-km-resolution national-scale model between 2011 and 2016 to estimate daily PM2.5 concentrations. A two-stage random forest model integrating MAIAC AOD with meteorological variables and land use data was applied to develop the model. The first-stage random forest model was used to impute the missing AOD values. The second-stage random forest model was then utilised to estimate ground PM2.5 concentrations. Ten-fold cross-validation was performed to evaluate the model performance. There was good consistency between MAIAC AOD and ground truth in Japan (correlation coefficient = 0.82 and 74.62% of data falling within the expected error). For model training, the model showed a training coefficient of determination (R2) of 0.98 and a root mean square error (RMSE) of 1.22 μg/m3. For the 10-fold cross-validation, the cross-validation R2 and RMSE of the model were 0.86 and 3.02 μg/m3, respectively. A subsite validation was used to validate the model at the grids overlapping with the AERONET sites, and the model performance was excellent at these sites with a validation R2 (RMSE) of 0.94 (1.78 μg/m3). Additionally, the model performance increased as increased AOD coverage. The top-ten important predictors for estimating ground PM2.5 concentrations were day of the year, temperature, AOD, relative humidity, 10-m-height zonal wind, 10-m-height meridional wind, boundary layer height, precipitation, surface pressure, and population density. MAIAC AOD showed high retrieval accuracy in Japan. The performance of the satellite-based model was excellent, which showed that PM2.5 estimates derived from the model were reliable and accurate. These estimates can be used to assess both the short-term and long-term effects of PM2.5 on health outcomes in epidemiological studies.