On Some Aspects of Variable Selection for Partial Least Squares Regression Models

Development of novel univariate and multivariate validated chemometric methods for the analysis of dasatinib, sorafenib, and vandetanib in pure form, dosage forms and biological fluids

Discrimination of oils and fuels using a portable NIR spectrometer

Comparison of wavelength selected methods for improving of prediction performance of PLS model to determine aflatoxin B1 (AFB1) in wheat samples during storage

A nondestructive method for evaluation of tomato quality was proposed to recognize the different ripening stages of tomatoes and to predict their hardness. Computer vision and electronic nose were used, and data fusion technology was used to establish the decision model of multi‐attribute information fusion. Fisher discriminant analysis (FDA) model and support vector classifier (SVC) model for tomato grading were established. Partial least squares (PLS) model and support vector regression (SVR) model were used for predicting the hardness. Comparing with the single detection technique, the combined system which fused computer vision and electronic nose, achieved better results. In the combined system: with respect to the SVC model, classification accuracy was 96.38% for the training set, 94.20% for the prediction set; while to the FDA model, classification accuracy was 93.48% for the training set, 85.51% for the prediction set; to the PLS model, the correlation coefficient of prediction (Rp) was 94.09% and the root mean square error of prediction (RMSEP) was 2.33 N; to the SVR model, the Rp was 95.14% and RMSEP was 0.03 N. It was observed that more robust and better prediction performance on tomato detection were achieved by using fusion information.Practical applicationsComputer vision, electronic nose, data fusion, and pattern recognition were used for the evaluation of tomato ripeness and hardness. Computer vision captured the color information of tomatoes and electronic nose responded to the smell of tomatoes. After data fusion, information became more abundant and the discrimination models also got better performance. It proved the possibility of nondestructive method for the evaluation of tomatoes quality during storage. So it is feasible to apply this method to detect tomato quality in factory.

Two chemometric methods were performed for the determination of acetic acid of fruit vinegars using near infrared (NIR) spectroscopy. Three varieties of fruit vinegars were prepared and 135 samples (45 samples for each variety) were selected for the calibration set, whereas 45 samples (15 samples for each variety) for the validation set. Partial least squares (PLS) analysis was the calibration method as well as extraction method for latent variables (LVs). The first eight LVs were employed as the inputs of least squares-support vector machine (LS-SVM) model. Then LS-SVM model with radial basis function (RBF) kernel was applied to build the regression model compared with PLS model. The correlation coefficient (r), root mean square error of prediction (RMSEP) and bias for validation set were 0.994, 0.814 and -0.091 by PLS, whereas 0.997, 0.651 and 0.011 by LS-SVM, respectively. LS-SVM model outperformed PLS model, but both models achieved an excellent prediction precision. The results indicated that NIR spectroscopy combined with chemometrics could be utilized as a high precision and fast way for the determination of acetic acid of fruit vinegars.

A method of rapidly determining the total polar compounds (TPCs) in frying oils using attenuated total reflectance‐Fourier transform infrared spectroscopy combined with partial least squares (PLS) regression is developed. Oils of various types and geographic origins are used to ensure that the proposed model is robust. The first derivative spectrum is selected as the spectral processing method. The interval PLS, forward interval PLS, and backward interval PLS algorithms are compared in terms of their performance. A correlation coefficient (R2) of 0.9942, a root mean square error of calibration (RMSEC) of 1.1, a root mean square error of prediction (RMSEP) of 2.30, a residual predictive deviation (RPD) of 4.1, and a limit of detection (LOD) of 1.65% are obtained by the fiPLS33 model with fewer latent variables and a lower spectral interval number. In addition, sub‐models using a single type of oil showed higher performance (R2 0.9957–0.9998, RMSEC 0.12–0.92, RMSEP 0.79–1.58, RPD 4.79–9.64, LOD 0.66–1.26%) than the general model. The TPC models developed are accurate, stable, and adaptable, and they can be used to analyze general frying oil samples quickly, regardless of the oil type, and to analyze samples of specific oil types accurately.Practical applications: The content of TPCs is an important indicator of whether the oil has been overused and whether it will be harmful during the frying process. However, traditional chemical methods are time‐consuming, and they have not been used to determine large‐sized samples. In addition, due to a lack of regional optimization, most studies on determining TPCs with FTIR give unsatisfactory model performance. A general TPC model that incorporates several oil types and regional optimization is expected to improve prediction performance. Therefore, the proposed method represents a rapid and accurate tool for measuring TPCs in edible fats and oils.

Color, phenolic content, and chemical age values of red wines made from Cretan grape varieties (Kotsifali, Mandilari) were evaluated over nine months of maturation in different containers for two vintages. The wines differed greatly on their anthocyanin profiles. Mid-IR spectra were also recorded with the use of a Fourier Transform Infrared Spectrophotometer in ZnSe disk mode. Analysis of Variance was used to explore the parameter's dependency on time. Determination models were developed for the chemical age indexes using Partial Least Squares (PLS) (TQ Analyst software) considering the spectral region 1830–1500 cm−1. The correlation coefficients (r) for chemical age index i were 0.86 for Kotsifali (Root Mean Square Error of Calibration (RMSEC) = 0.067, Root Mean Square Error of Prediction (RMSEP) = 0,115, and Root Mean Square Error of Validation (RMSECV) = 0.164) and 0.90 for Mandilari (RMSEC = 0.050, RMSEP = 0.040, and RMSECV = 0.089). For chemical age index ii the correlation coefficients (r) were 0.86 and 0.97 for Kotsifali (RMSEC 0.044, RMSEP = 0.087, and RMSECV = 0.214) and Mandilari (RMSEC = 0.024, RMSEP = 0.033, and RMSECV = 0.078), respectively. The proposed method is simpler, less time consuming, and more economical and does not require chemical reagents.

Dioxin screening in fish product by pattern recognition of biomarkers

This article presents a multivariate method of rapidly determining chlopyrifos residue in white radish, based on near-infrared spectroscopy and partial least squares (PLS) regression. Interval PLS (iPLS) was utilized to select the optimum wave number range. The number of PLS components and the number of intervals were optimized according to root mean square error of prediction (RMSEP) and correlation coefficient (R) in prediction set. The result showed that the iPLS model was more reliable than the full model and that near-infrared spectroscopy with iPLS algorithm could be used successfully to analyze chlorpyrifos residue in white radish.

High-precision maps of soil organic carbon (SOC) are beneficial for managing soil fertility and understanding the global carbon cycle. Digital soil mapping plays an important role in efficiently obtaining the spatial distribution of SOC, which contributes to precision agriculture. However, traditional soil-forming factors (i.e., terrain or climatic factors) have weak variability in low-relief areas, such as plains, and cannot reflect the spatial variation of soil attributes. Meanwhile, vegetation cover hinders the acquisition of the direct information of farmland soil. Thus, useful environmental variables should be utilized for SOC prediction and the digital mapping of such areas. SOC has an important effect on crop growth status, and remote sensing data can record the apparent spectral characteristics of crops. The normalized difference vegetation index (NDVI) is an important index reflecting crop growth and biomass. This study used NDVI time series data rather than traditional soil-forming factors to map SOC. Honghu City, located in the middle of the Jianghan Plain, was selected as the study region, and the NDVI time series data extracted from Landsat 8 were used as the auxiliary variables. SOC maps were estimated through stepwise linear regression (SLR), partial least squares regression (PLSR), support vector machine (SVM), and artificial neural network (ANN). Ordinary kriging (OK) was used as the reference model, while root mean square error of prediction (RMSEP) and coefficient of determination of prediction (R2P) were used to evaluate the model performance. Results showed that SOC had a significant positive correlation in July and August (0.17, 0.29) and a significant negative correlation in January, April, and December (−0.23, −0.27, and −0.23) with NDVI time series data. The best model for SOC prediction was generated by ANN, with the lowest RMSEP of 3.718 and highest R2P of 0.391, followed by SVM (RMSEP = 3.753, R2P = 0.361) and PLSR (RMSEP = 4.087, R2P = 0.283). The SLR model was the worst model, with the lowest R2P of 0.281 and highest RMSEP of 3.930. ANN and SVM were better than OK (RMSEP = 3.727, R2P = 0.372), whereas PLSR and SLR were worse than OK. Moreover, the prediction results using single-data NDVI or short time series NDVI showed low accuracy. The effect of the terrain factor on SOC prediction represented unsatisfactory results. All these results indicated that the NDVI time series data can be used for SOC mapping in plain areas and that the ANN model can maximally extract additional associated information between NDVI time series data and SOC. This study presented an effective method to overcome the selection of auxiliary variables for digital soil mapping in plain areas when the soil was covered with vegetation. This finding indicated that the time series characteristics of NDVI were conducive for predicting SOC in plains.

Data from a paperboard machine were used to compare the performance of linear partial least squares (PLS) and nonlinear feed-forward neural network (FFNN) modeling of a continuous process. Fifteen selected variables were used as input parameters to the models, while the quality class of the manufactured product was the output response. The models were validated with external data different to those used in the design of the models. Evaluation with root mean square error of prediction (RMSEP) showed that the FFNN models were better for prediction than the PLS models. For monitoring, however, the PLS models detected deviations from normal settings in the paperboard machine more sensitively than the FFNN models. It is suggested that these findings have general relevance to other continuous processes in manufacturing industries too.

Rapid discrimination and quantification of kudzu root with its adulterant part using FT-NIR and a machine learning algorithm

Biomass energy represents a huge supplement for meeting current energy demands. A hyperspectral imaging system covering the spectral range of 874–1734 nm was used to determine the pH value of anaerobic digestion liquid produced by water hyacinth and rice straw mixtures used for methane production. Wavelet transform (WT) was used to reduce noises of the spectral data. Successive projections algorithm (SPA), random frog (RF) and variable importance in projection (VIP) were used to select 8, 15 and 20 optimal wavelengths for the pH value prediction, respectively. Partial least squares (PLS) and a back propagation neural network (BPNN) were used to build the calibration models on the full spectra and the optimal wavelengths. As a result, BPNN models performed better than the corresponding PLS models, and SPA-BPNN model gave the best performance with a correlation coefficient of prediction (rp) of 0.911 and root mean square error of prediction (RMSEP) of 0.0516. The results indicated the feasibility of using hyperspectral imaging to determine pH values during anaerobic digestion. Furthermore, a distribution map of the pH values was achieved by applying the SPA-BPNN model. The results in this study would help to develop an on-line monitoring system for biomass energy producing process by hyperspectral imaging.

The color change resulting from anthocyanin and iron co-pigmentation has been a significant challenge for the food industry in the development of many iron-fortified foods. This present study aims to establish a quantitative model to predict the degree of color stability in the presence of dissolved iron using surface-enhanced Raman spectroscopic (SERS) spectra. The SERS spectra of anthocyanin extracts from seven different plant sources were measured and analyzed by principal component analysis (PCA). Discrimination among different sources of anthocyanin was observed in the PCA plot. Different stability indexes, obtained by measuring both the color intensity stability and color hue stability of each sample, were established based on UV-vis analysis of anthocyanin at pH 3 and 6 with and without ferric sulfate. Partial least square (PLS) regression models were applied to establish the correlation between SERS spectra and stability indexes. The best PLS model was built based on the stability index calculated from the bathochromic shift (UV-vis spectral range: 380-750 nm) in pH3 buffer and the SERS spectra, achieving a root mean square error of prediction (RMSEP) of 2.16 nm and a correlation coefficient value (R2) of 0.98. In conclusion, the present study developed a feasible approach to predict the stability of anthocyanin colorants against iron co-pigmentation. The developed method and models can be used for fast screenings of raw ingredients in iron-fortified food products.

Mobile devices were used to quantify ethanol content—via colorimetry—in cachaca samples to great analytical precision. Ethanol from the spirits reacted with ceric ammonium nitrate (CAN) and nitric acid (NA) to form a red-colored complex. Images from the reaction were taken over time using a smartphone device and processed using software PhotoMetrix Pro® via partial least squares (PLS). A priori, Doehlert matrix was used for optimization of reagents concentration in the complexation reaction, with hierarchical cluster analysis (HCA) as estimation method and a flatbed scanner as measuring tool. When compared with the UV-VIS reference method, results from mobile device imaging presented a root mean square error of prediction (RMSEP) of 0.0677% (v/v). A t test showed there is no significant difference between methods (P < 0.05).