Comparison of different processing approaches by SVM and RF on HS-MS eNose and NIR Spectrometry data for the discrimination of gasoline samples

Detection and quantification of cocoa powder adulteration using Vis-NIR spectroscopy with chemometrics approach

The rising demand for cocoa powder has resulted in an upsurge in market prices, leading to the emergence of adulteration practices aimed at achieving economic benefits. This study aimed to detect and quantify cocoa powder adulteration using visible and near-infrared spectroscopy (Vis-NIRS). The adulterants used in this study were powdered carob, cocoa shell, foxtail millet, soybean, and whole wheat. The NIRS data could not be resolved using Savitzky-Golay smoothing. Nevertheless, the application of a random forest and support vector machine successfully classified the samples with 100% accuracy. Quantification of adulteration using partial least squares (PLS), Lasso, Ridge, elastic Net, and RF regressions provided R2 higher than 0.96 and root mean square error <2.6. Coupling PLS with the Boruta algorithm produced the most reliable regression model (R2 = 1, RMSE = 0.0000). Finally, an online application was prepared to facilitate the determination of adulterants in the cocoa powder.

Development of a Novel HS-GC/MS Method Using the Total Ion Spectra Combined with Machine Learning for the Intelligent and Automatic Evaluation of Food-Grade Paraffin Wax Odor Level

The intensity of the odor in food-grade paraffin waxes is a pivotal quality characteristic, with odor panel ratings currently serving as the primary criterion for its assessment. This study presents an innovative method for assessing odor intensity in food-grade paraffin waxes, employing headspace gas chromatography with mass spectrometry (HS/GC-MS) and integrating total ion spectra with advanced machine learning (ML) algorithms for enhanced detection and quantification. Optimization was conducted using Box-Behnken design and response surface methodology, ensuring precision with coefficients of variance below 9%. Analytical techniques, including hierarchical cluster analysis (HCA) and principal component analysis (PCA), efficiently categorized samples by odor intensity. The Gaussian support vector machine (SVM), random forest, partial least squares regression, and support vector regression (SVR) algorithms were evaluated for their efficacy in odor grade classification and quantification. Gaussian SVM emerged as superior in classification tasks, achieving 100% accuracy, while Gaussian SVR excelled in quantifying odor levels, with a coefficient of determination (R2) of 0.9667 and a root mean square error (RMSE) of 6.789. This approach offers a fast, reliable, robust, objective, and reproducible alternative to the current ASTM sensory panel assessments, leveraging the analytical capabilities of HS-GC/MS and the predictive power of ML for quality control in the petrochemical sector's food-grade paraffin waxes.

A rapid method for identification of Lanxangia tsaoko origin and fruit shape: FT-NIR combined with chemometrics and image recognition

Lanxangia tsaoko's accurate classifications of different origins and fruit shapes are significant for research in L. tsaoko difference between origin and species as well as for variety breeding, cultivation, and market management. In this work, Fourier transform-near infrared (FT-NIR) spectroscopy was transformed into two-dimensional and three-dimensional correlation spectroscopies to further investigate the spectral characteristics of L. tsaoko. Before building the classification model, the raw FT-NIR spectra were preprocessed using multiplicative scatter correction and second derivative, whereas principal component analysis, successive projections algorithm, and competitive adaptive reweighted sampling were used for spectral feature variable extraction. Then combined with partial least squares-discriminant analysis (PLS-DA), support vector machine (SVM), decision tree, and residual network (ResNet) models for origin and fruit shape discriminated in L. tsaoko. The PLS-DA and SVM models can achieve 100% classification in origin classification, but what is difficult to avoid is the complex process of model optimization. The ResNet image recognition model classifies the origin and shape of L. tsaoko with 100% accuracy, and without the need for complex preprocessing and feature extraction, the model facilitates the realization of fast, accurate, and efficient identification.

A multi-objective stacked regression method for distance based colour measuring device

Identifying colour from a distance is challenging due to the external noise associated with the measurement process. The present study focuses on developing a colour measuring system and a novel Multi-target Regression (MTR) model for accurate colour measurement from distance. Herein, a novel MTR method, referred as Multi-Objective Stacked Regression (MOSR) is proposed. The core idea behind MOSR is based on stacking as an ensemble approach with multi-objective evolutionary learning using NSGA-II. A multi-objective optimization approach is used for selecting base learners that maximises prediction accuracy while minimising ensemble complexity, which is further compared with six state-of-the-art methods over the colour dataset. Classification and regression tree (CART), Random Forest (RF) and Support Vector Machine (SVM) were used as regressor algorithms. MOSR outperformed all compared methods with the highest coefficient of determination values for all three targets of the colour dataset. Rigorous comparison with state-of-the-art methods over 18 benchmarked datasets showed MOSR outperformed in 15 datasets when CART was used as a regressor algorithm and 11 datasets when RF and SVM were used as regressor algorithms. The MOSR method was statistically superior to compared methods and can be effectively used to measure accurate colour values in the distance-based colour measuring device.

Machine learning-based approaches to Vis-NIR data for the automated characterization of petroleum wax blends

Petroleum waxes are products derived from lubricating oils with a wide spectrum of industrial and consumer applications that depend on their composition. In addition, the intended applications of this product are also subject to the practice of blending petroleum waxes with different chemical characteristics (e.g., paraffin waxes and microwaxes) to achieve the appropriate physicochemical properties. This study introduces a novel method based on visible and near-infrared spectroscopy (Vis-NIR) combined with machine learning (ML) for the characterization of blends of the two types of commonly marketed petroleum waxes (paraffin waxes and microwaxes). With spectroscopic data, Partial Least Squared Regression (PLSR), Support Vector Regression (SVR), and Random Forest (RF) Regression-based regression ML models have been developed, obtaining satisfactory results for the characterization of the percentage of blending in petroleum waxes. Moreover, strategies using wrapper variable selection methods like the Boruta algorithm and Genetic Algorithm (GA) have been implemented to assess if fewer predictors enhance model performance. Particularly, the application of wrapper variable selection methods, specifically the Boruta algorithm, has led to an improvement in the performance of the models obtained. Results obtained by the Boruta-PLS model showed the best performance with an RMSE of 2.972 and an R2 of 0.9925 for the test set and an RMSE of 1.814 and an R2 of 0.9977 for the external validation set. Additionally, this model allowed for establishing the relative importance of the variables in the characterization of the waxes mixture, pointing out that the hydrocarbon content ratio is critical in the determination of this value. An interactive web application was developed using the best model developed for easy processing of the data by the users.

Gasoline discrimination using infrared spectroscopy and virtual samples based on measurement uncertainty

In a previous work, we proposed a methodology for pair-wise discrimination of gasoline samples by creating virtual samples based on physicochemical assays or distillation curves. Satisfactory results were achieved, although specialist and specific apparatus (not commonly available at police laboratories) were required. The present study goes a step further and for the first time investigates the possibility of infrared (IR) spectroscopy to enable a virtual samples-based methodology for comparison of gasoline samples in pairs. IR spectroscopy feasibility for in situ applications is attractive for forensic investigations. The performances of one handheld NIR device and one dual-range (FT-NIR and FT-IR) benchtop spectrometer were evaluated. The estimation of uncertainty in infrared spectral measurement (needed to generate virtual samples) is barely discussed in literature. So far, there are no literature reports describing quantification and comparison of measurement uncertainties for the spectral acquisitions evaluated here, especially regarding their use for generating virtual samples. A stepwise procedure to quantify uncertainties associated with IR spectral acquisition, at each wavenumber, is described. This method can be useful for understanding both the sources of variability in IR measurements and the system under investigation. Uncertainty estimation was based on experimental data and considered intermediate precision, repeatability and variations in sample temperature as sources of variability. Virtual samples were employed in a discrimination approach using SIMCA models. Results for portable NIR, FT-NIR and FT-IR data sets showed complete discrimination for 96.3%, 93.4% and 93.7% of the 1431 pairs of gasoline samples evaluated, respectively. These results were comparable and similar to those obtained for the physicochemical properties data set (95.7%), although slightly inferior to the result obtained for distillation curves (99.2%). Using IR non-destructive methods in this case could enable faster investigations and simpler analysis, especially for the low-cost handheld spectrometer. In a screening approach, atmospheric distillation assays can be employed only if infrared techniques are not capable of distinguishing the samples subject to comparison. In this work, a pair of samples was considered to be completely discriminated only when a null false positive error (FPR) was achieved, although a more flexible criterium may be acceptable in practice. Finally, the methodology could be extended to other applications where sample comparison is important.

Rapid Classification of Petroleum Waxes: A Vis-NIR Spectroscopy and Machine Learning Approach

Petroleum-derived waxes are used in the food industry as additives to provide texture and as coatings for foodstuffs such as fruits and cheeses. Therefore, food waxes are subject to strict quality controls to comply with regulations. In this research, a combination of visible and near-infrared (Vis-NIR) spectroscopy with machine learning was employed to effectively characterize two commonly marketed petroleum waxes of food interest: macrocrystalline and microcrystalline. The present study employed unsupervised machine learning algorithms like hierarchical cluster analysis (HCA) and principal component analysis (PCA) to differentiate the wax samples based on their chemical composition. Furthermore, nonparametric supervised machine learning algorithms, such as support vector machines (SVMs) and random forest (RF), were applied to the spectroscopic data for precise classification. Results from the HCA and PCA demonstrated a clear trend of grouping the wax samples according to their chemical composition. In combination with five-fold cross-validation (CV), the SVM models accurately classified all samples as either macrocrystalline or microcrystalline wax during the test phase. Similar high-performance outcomes were observed with RF models along with five-fold CV, enabling the identification of specific wavelengths that facilitate discrimination between the wax types, which also made it possible to select the wavelengths that allow discrimination of the samples to build the characteristic spectralprint of each type of petroleum wax. This research underscores the effectiveness of the proposed analytical method in providing fast, environmentally friendly, and cost-effective quality control for waxes. The approach offers a promising alternative to existing techniques, making it a viable option for automated quality assessment of waxes in food industrial applications.

A Rapid and Accurate Quantitative Analysis of Cellulose in the Rice Bran Layer Based on Near-Infrared Spectroscopy

Cultivating rice varieties with lower cellulose content in the bran layer has the potential to enhance both the nutritional value and texture of brown rice. This study aims to establish a rapid and accurate method to quantify cellulose content in the bran layer utilizing near-infrared spectroscopy (NIRS), thereby providing a technical foundation for the selection, screening, and breeding of rice germplasm cultivars characterized by a low cellulose content in the bran layer. To ensure the accuracy of the NIR spectroscopic analysis, the potassium dichromate oxidation (PDO) method was improved and then used as a reference method. Using 141 samples of rice bran layer (rice bran without germ), near-infrared diffuse reflectance (NIRdr) spectra, near-infrared diffuse transmittance (NIRdt) spectra, and fusion spectra of NIRdr and NIRdt were used to establish cellulose quantitative analysis models, followed by a comparative evaluation of these models' predictive performance. Results indicate that the optimized PDO method demonstrates superior precision compared to the original PDO method. Upon examining the established models, their predictive capabilities were ranked in the following order: the fusion model outperforms the NIRdt model, which in turn surpasses the NIRdr model. Of all the fusion models developed, the model exhibiting the highest predictive accuracy utilized fusion spectra (NIRdr-NIRdt (1st der)) derived from preprocessed (first derivative) diffuse reflectance and transmittance spectra. This model achieved an external predictive R2p of 0.903 and an RMSEP of 0.213%. Using this specific model, the rice mutant O2 was successfully identified, which displayed a cellulose content in the bran layer of 3.28%, representing a 0.86% decrease compared to the wild type (W7). The utilization of NIRS enables quantitative analysis of the cellulose content within the rice bran layer, thereby providing essential technical support for the selection of rice varieties characterized by lower cellulose content in the bran layer.

Determination of benzo(a)pyrene in peanut oil based on Raman spectroscopy and machine learning methods

Benzo(a)pyrene (BaP) generated in the production process of oil is harmful to human severely as a kind of carcinogenic substance. In this study, the qualitative and quantitative detection of BaP concentration in peanut oil was investigated based on Raman spectroscopy combined with machine learning methods. The glass substrates and magnetron sputtered gold substrates for the Raman spectra were compared and the data preprocessing methods of principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE) were used to process Raman signal. Back propagation neural network (BPNN), partial least squares regression (PLSR), support vector machine (SVM) and random forest (RF) algorithms were developed to obtain the qualitative and quantitative detection model of BaP concentration in peanut oil. The results showed that the Raman spectra with the glass substrate was more suitable for the BaP detection than magnetron sputtered gold substrates. RF combined with t-SNE could achieve an accuracy of 97.5% in the qualitative detection of BaP concentration levels in model validation experiment, and the correlation coefficient of the prediction set (R_p) in the quantitative detection was 0.9932, the root mean square error (RMSEP) was 0.8323 μg/kg and the bias was 0.1316 μg/kg. It can be concluded that Raman spectroscopy combined with machine learning methods could provide an effective method for the rapid determination of BaP concentration in peanut oil.

Comparison of various chemometric methods on visible and near-infrared spectral analysis for wood density prediction among different tree species and geographical origins

Visible and near-infrared (Vis-NIR) spectroscopy has been widely applied in many fields for the qualitative and quantitative analysis. Chemometric techniques including pre-processing, variable selection, and multivariate calibration models play an important role to better extract useful information from spectral data. In this study, a new de-noising method (lifting wavelet transform, LWT), four variable selection methods, as well as two non-linear machine learning models were simultaneously analyzed to compare the impact of chemometric approaches on wood density determination among various tree species and geographical locations. In addition, fruit fly optimization algorithm (FOA) and response surface methodology (RSM) were employed to optimize the parameters of generalized regression neural network (GRNN) and particle swarm optimization-support vector machine (PSO-SVM), respectively. As for various chemometric methods, the optimal chemometric method was different for the same tree species collected from different locations. FOA-GRNN model combined with LWT and CARS deliver the best performance for Chinese white poplar of Heilongjiang province. In contrast, PLS model showed a good performance for Chinese white poplar collected from Jilin province based on raw spectra. However, for other tree species, RSM-PSO-SVM models can improve the performance of wood density prediction compared to traditional linear and FOA-GRNN models. Especially for Acer mono Maxim, when compared to linear models, the coefficient of determination of prediction set ( R p 2 ) and relative prediction deviation (RPD) were increased by 47.70% and 44.48%, respectively. And the dimensionality of Vis-NIR spectral data was decreased from 2048 to 20. Therefore, the appropriate chemometric technique should be selected before building calibration models.

Rapid determination of viscosity and viscosity index of lube base oil based on near-infrared spectroscopy and new transformation formula

Viscosity and viscosity index are the key product properties in lubricating oil production process. Rapid and even online analysis of viscosity and viscosity index through near-infrared (NIR) spectroscopy combined with chemometrics is helpful to optimize the production process. However, due to the nonlinear effect, the commonly used linear multivariate correction method is not effective. In this work, the feasibility of four existing viscosity linear transformation formulas for establishing NIR models was studied, and a new viscosity linear transformation formula was developed based on the viscosity-gravity constant. The experimental results showed that three of the four existing viscosity linear transformation formulas made some improvement on the viscosity prediction of base oil, but not as good as the newly established viscosity linear transformation formula. For viscosity index, the accuracy of modeling with reference viscosity index directly was much better than calculating by prediction viscosity value. Both of the viscosity and viscosity index prediction results of NIR analysis were in good agreement with the results of reference method, indicating that the determination can meet the needs of rapid and on-line analysis in industrial field.

Feasibility of an NIR spectral calibration transfer algorithm based on optimized feature variables to predict tobacco samples in different states

The prediction accuracy of calibration models for near-infrared (NIR) spectroscopy typically relies on the morphology and homogeneity of the samples. To achieve non-homogeneous tobacco samples for non-destructive and rapid analysis, a method that can predict tobacco filament samples using reliable models based on the corresponding tobacco powder is proposed here. First, as it is necessary to establish a simple and robust calibrated model with excellent performance, based on full-wavelength PLSR (Full-PLSR), the key feature variables were screened by three methods, namely competitive adaptive reweighted sampling (CARS), variable combination population analysis-iteratively retaining informative variables (VCPA-IRIV), and variable combination population analysis-genetic algorithm (VCPA-GA). The partial least squares regression (PLSR) models for predicting the total sugar content in tobacco were established based on three optimal wavelength sets and named CARS-PLSR, VCPA-IRIV-PLSR and VCPA-GA-PLSR, respectively. Subsequently, they were combined with different calibration transfer algorithms, including calibration transfer based on canonical correlation analysis (CTCCA), slope/bias correction (S/B) and non-supervised parameter-free framework for calibration enhancement (NS-PFCE), to evaluate the best prediction model for the tobacco filament samples. Compared with the previous two transfer algorithms, NS-PFCE performed the best under various wavelength conditions. The prediction results indicated that the most successful approach for predicting the tobacco filament samples was achieved by VCPA-IRIV-PLSR when coupled with the NS-PFCE method, which obtained the highest determination coefficient (R_p² = 0.9340) and the lowest root mean square error of the prediction set (RMSEP = 0.8425). VCPA-IRIV simplifies the calibration model and improves the efficiency of model transfer (31 variables). Furthermore, it pledges the prediction accuracy of the tobacco filament samples when combined with NS-PFCE. In summary, calibration transfer based on optimized feature variables can eliminate prediction errors caused by sample morphological differences and proves to be a more beneficial method for online application in the tobacco industry.

Rapid and accurate determination of diesel multiple properties through NIR data analysis assisted by machine learning

The rapid and accurate detection of diesel multiple properties is an important research topic in petrochemical industry that is conducive to diesel quality assessment and environmental pollution mitigation. To that end, this paper developed a new machine learning model for near infrared (NIR) spectroscopy capable of simultaneously determining diesel density, viscosity, freezing point, boiling point, cetane number and total aromatics. The model combined improved XY co-occurrence distance (ISPXY) and differential evolution-gray wolf optimization support vector machine (DEGWO-SVM) to attain the goal of rapidity and accuracy. Experimental results indicated that the average recovery, mean square error, mean absolute percentage error and determination coefficient of the presented method outperformed those of the existing machine learning methods. The proposed hybrid model provides superior solution to the problem of low efficiency and high cost of diesel quality detection, and has the potential to be utilized as a promising tool for diesel routine monitoring.

Improvement of NIR prediction ability by dual model optimization in fusion of NSIA and SA methods

Feature selection and sample partitioning are both important to establish a quantitative analytical model for near-infrared (NIR) spectroscopy. The classical interval partial least squares (iPLS) model for waveband selection can be improved in combination of the simulated annealing (SA) algorithm. The sample set partitioning based on a joint x-y distance (SPXY) method for sample partitioning is based on the distances of both the x- and y- dimensions; it is expected to be optimized using the non-dominant sorting strategies (NS) combined with the immune algorithm (IA). In this study, we investigated the dual model optimization mode for simultaneous selection of feature waveband and sample partitioning, and proposed a novel method defined as SA-iPLS & SPXY-NSIA. The method explores a population evolution process, and takes the candidate individual as the link for the fusion optimization of SA-iPLS and SPXY-NSIA. The method screens feature wavebands and observes a good partition of the modeling samples, to construct a combined optimization strategy for fusion optimization of the target waveband and suitable sets of sample partitioning. The performance of the SA-iPLS & SPXY-NSIA method was tested using a soil sample dataset. To prove model enhancement, the proposed method was compared to the two traditional methods of Kennard-Stone (KS) and SPXY in combination with SA-iPLS. Experimental results show that the fusion model established by SA-iPLS & SPXY-NSIA performed better than the KS-SA-iPLS and SPXY-SA-iPLS models. The best testing results of the fusion model is with RMSET, RPD_T and R_T observed as 0.0107, 1.7233 and 0.9097, respectively. The proposed method is prospectively able to effectively improve the predictive ability of the NIR analytical model.