Rapid analysis the type of customs paper using Micro-NIR spectrometers and machine learning algorithms.

Quick identification of paper types for customs is extremely crucial. Although there are a variety of researches focus on the discrimination of paper, these techniques either require complex preprocessing or large-scale instruments, which are not suitable for customs environments. In this study, we predicted the type of customs paper by using a Micro-NIR spectrometer, and compared the results with Attenuated Total Reflection-Fourier Transform Infrared Spectroscopy (ATR-FTIR). Four different classification algorithms, including linear and non-linear classifiers: K-nearest neighbor (KNN), soft independent modeling of class analogy (SIMCA), partial least squares discriminant analysis (PLS-DA), and least squares-support vector machine (LS-SVM) were employed to classify the type of paper. 20 groups of datasets were selected by Monte Carlo sampling. For Micro-NIR data, the performances of KNN and LS-SVM were outstanding than SIMCA and PLS-DA, with the average accuracies 96.06% and 98.91%, respectively. The outcome of SIMCA and PLS-DA were similar, with the average accuracies 93.00% and 93.97%. Based on the standard derivation, the best stability of models was LS-SVM (1.06%), followed by PLS-DA (1.12%), KNN (1.22%) and SIMCA (3.07%). Compared with ATR-FTIR, the effects of Micro-NIR were better, which were embodies in the better KNN and SIMCA models, and the comparable LS-SVM model. The result demonstrated that the Micro-NIR combined with machine learning algorithms was an effective method to classify the type of customs paper efficiently and quickly, even better than ATR-FTIR.

Fourier transform infrared spectroscopy and chemometrics for the discrimination of animal fur types.

Rapid and reliable animal fur identification has remained a challenge for customs inspection. The accurate distinction between fur types has a significant meaning in implementing the correct tariff policy. A variety of analytical methods have been applied to work on distinguishing animal fur types, with tools of microscopy, molecular testing, mass spectrometry, Fourier transform infrared spectroscopy (FTIR), and Raman spectroscopy. In this research, the capability of attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) combined with pattern recognition methods was investigated for the discrimination of animal fur in six types. This work was to explore the non-destructive application of ATR-FTIR technique in discriminant analysis of animal fur. All spectra were collected by ATR-FTIR of the wavenumber ranging from 4000 to 650 cm-1. Data pretreatments included moving average smoothing and multiplicative scatter correction (MSC). Four supervised classification algorithms were chosen to categorize the types of fur: soft independent modeling of class analogy (SIMCA), principal component analysis linear discriminant analysis (PCA-LDA), partial least squares discriminant analysis (PLS-DA), least squares support vector machine (LS-SVM). PLS-DA and LS-SVM were both effective approaches, with a 100% classification accuracy rate. The accuracy of PCA-LDA and SIMCA was 98.33% and 99.44%, respectively. Furthermore, LS-SVM model obtained using Monte-Carlo sampling method also obtained 100% prediction accuracy, while all other methods produced misclassification. LS-SVM corrected the non-linearities for the animal fur FTIR data but also remarkably improved the prediction performance level. The results of this study revealed that the combination of ATR-FTIR and chemometrics has a huge potential for animal fur discrimination.

Feature selection of infrared spectra analysis with convolutional neural network.

Data-driven deep learning analysis, especially for convolution neural network (CNN), has been developed and successfully applied in many domains. CNN is regarded as a black box, and the main drawback is the lack of interpretation. In this study, an interpretable CNN model was presented for infrared data analysis. An ascending stepwise linear regression (ASLR)-based approach was leveraged to extract the informative neurons in the flatten layer from the trained model. The characteristic of CNN network was employed to visualize the active variables according to the extracted neurons. Partial least squares (PLS) model was presented for comparison on the performance of extracted features and model interpretation. The CNN models yielded accuracies with extracted features of 93.27%, 97.50% and 96.65% for Tablet, meat, and juice datasets on the test set, while the PLS-DA models obtained accuracies with latent variables (LVs) of 95.19%, 95.50% and 98.17%. Both the CNN and PLS models demonstrated the stable patterns on active variables. The repeatability of CNN model and proposed strategies were verified by conducting the Monte-Carlo cross-validation.

Non-destructive analysis the dating of paper based on convolutional neural network.

A non-destructive method based on Fourier Transformed Infrared Spectroscopy (FT-IR) was proposed to estimate the date of paper from different years in this article. For the paper samples, dated from 1940 to 1980, naturally aged and conserved in library. Partial least squares-discriminate analysis (PLS-DA), Logistic regression and convolutional neural network (CNN), were employed to evaluate the date of paper, with the accuracy 60.74%, 95.31% and 98.77%, respectively. Based on the characteristics of CNN model and with the help of network localization, active variables could be recognized in the whole spectrum. Although the localization of active variables showed a discriminative pattern, the selected spectral regions were similar. Most important variables focused on the 1700-1400 cm-1, were corresponding to cellulose crystallinity, which was consisted with the ageing processing. The present work gave the potential of FT-IR combined with chemometric techniques could estimate the dating of unknown paper. Meanwhile, the analysis of active variables obtained further indicated the worthy of CNN model for document dating.

A fast determination of insecticide deltamethrin by spectral data fusion of UV-vis and NIR based on extreme learning machine.

Spectral data fusion strategies combined with the extreme learning machine (ELM) algorithm was applied to determine the active ingredient in deltamethrin formulation. Ultraviolet-visible spectroscopy (UV-vis) is a rapid and sensitive detection method for specific components that are sensitive to ultraviolet irradiation. Alternatively, near-infrared spectroscopy (NIR) technology can be applied over a broader range. To determine a feasible method with a higher sensitivity and broader application range, the active ingredient of deltamethrin formulation was comprehensively investigated by combining the spectral data fusion strategy with ELM by employing UV-vis, NIR and fusion strategies, individually. Consequently, the results demonstrated that the low-level fusion strategy exhibited better predictive ability (lower RMSEP of 0.0645% and higher R2 of 0.9978) than mid-level fusion and individual methods. ELM combined with data fusion is proved to be an efficient method for the rapid analysis of deltamethrin formulations. Furthermore, this study provides a potential approach for pesticide quality control as well as on-site monitoring.

High-level Fusion Coupled with Mahalanobis Distance Weighted (MDW) Method for Multivariate Calibration.

Near infrared spectra (NIR) technology is a widespread detection method with high signal to noise ratio (SNR) while has poor modeling interpretation due to the overlapped features. Alternatively, mid-infrared spectra (MIR) technology demonstrates more chemical features and gives a better explanation of the model. Yet, it has the defects of low SNR. With the purpose of developing a model with plenty of characteristics as well as with higher SNR, NIR and MIR technologies are combined to perform high-level fusion strategy for quantitative analysis. A novel chemometrical method named as Mahalanobis distance weighted (MDW) is proposed to integrate NIR and MIR techniques comprehensively. Mahalanobis distance (MD) based on the principle of spectral similarity is obtained to calculate the weight of each sample. Specifically, the weight is assigned to the inverse ratio of the corresponding MD. Besides, the proposed MDW method is applied to NIR and MIR spectra of active ingredients in deltamethrin and emamectin benzoate formulations for quantitative analysis. As a consequence, the overall results show that the MDW method is promising with noticeable improvement of predictive performance than individual methods when executing high-level fusion for quantitative analysis.

A new hybrid filter/wrapper algorithm for feature selection in classification.

Feature selection can greatly enhance the performance of a learning algorithm when dealing with a high dimensional data set. The filter method and the wrapper method are the two most commonly approaches. However, these approaches have limitations. The filter method uses independent evaluation to evaluate and select features, which is computationally efficient but less accurate than the wrapper method. The wrapper method uses a predetermined classifier to compute the evaluation, which can afford high accuracy for particular classifiers, but is computationally expensive. In this study, we introduce a new feature selection method that we refer to as the large margin hybrid algorithm for feature selection (LMFS). In this method, we first utilize a new distance-based evaluation function, in which ideally samples from the same class are close together, whereas samples from other classes are far apart, and a weighted bootstrapping search strategy to find a set of candidate feature subsets. Then, we use a specific classifier and cross-validation to select the final feature subset from the candidate feature subsets. Six vibrational spectroscopic data sets and three different classifiers, namely k-nearest neighbors, partial least squares discriminant analysis and least squares support vector machine were used to validate the performance of the LMFS method. The results revealed that LMFS can effectively overcome the over-fitting between the optimal feature subset and a given classifier. Compared with the filter and wrapper methods, the features selected by the LMFS method have better classification performance and model interpretation. Furthermore, LMFS can effectively overcomes the impact of classifier complexity on computational time, and distance-based classifiers were found to be more suitable for selecting the final subset in LMFS.

Bootstrapping search margin-based nearest neighbor method for qualitative spectroscopic analysis.

Qualitative spectroscopic analysis depends in one way or another on comparing spectra of the specimens to be identified with spectra of "known" or "standard" samples. The k-nearest neighbor (k-NN) method is one of the oldest and simplest techniques for performing such comparisons. In this study, we present a new k-NN algorithm for qualitative spectroscopic analysis, which we refer to as the bootstrapping search margin-based nearest neighbor (BSMNN) method. This method consists of two phases. In the first phase, we attempt to find a feature space in which samples with different labels produce large margins, such that the classification has high confidence, by maximizing a margin-based objective function using a weighted bootstrap sampling search strategy. In the subsequent classification phase, we compute local Euclidean distances between different samples under the feature space. A new instance x is classified by the label of its nearest neighbor. Six widely used vibrational spectroscopic data sets were used to validate the performance of the BSMNN method. The results showed that, despite its simplicity, BSMNN yields better results compared with commonly used k-NN algorithms including Relief, neighborhood components analysis, neighborhood component feature selection, and large margin nearest neighbor. Furthermore, BSMNN can be used to identify important spectral regions. It is worth mentioning that the margin-based objective function used in BSMNN is proposed for the first time for measuring the quality of features. Although in this study the margin-based objective function is focused on k-NN classification, it also can be used for other distance-based classifiers, such as soft independent modeling of class analogies and least squares support vector machine.

A new strategy of applying modeling indicator determined method to high-level fusion for quantitative analysis.

A novel method, named as modeling indicator determined (MID) method, based on two model evaluation parameters i.e., root mean square error of prediction (RMSEP) and ratio performance deviation (RPD), is proposed to employ high-level fusion for quantitative analysis. The two MID methods of root mean square error of prediction weighted (RMSEPW) method and ratio performance deviation weighted (RPDW) method are put forward on the basis of the model evaluation indicators from the individual models. Performance of RMSEPW method and RPDW method are evaluated in terms of the predictive ability of root mean square error of prediction for fusion (RMSEPf) through the fused models. The two MID methods are applied to UV-visible (UV-vis), near infrared (NIR) and mid-infrared (MIR) spectral data of active ingredient in pesticide, and gas chromatography-mass spectrometer (GC-MS) and NIR spectral data of n-heptane in chemical complex for high-level fusion. Moreover, the results are compared with the individual methods. As a result, the overall results show that the two MID methods are promising with significant improvement of predictive performance for high-level fusion when executing quantitative analysis.

Spectral interval combination optimization (ICO) on rapid quality assessment of Solanaceae plant: a validation study.

A novel spectral variable selection method, named as interval combination optimization (ICO), was proposed in the previous study of us. In the present study, ICO coupled with near infrared (NIR) spectroscopy was applied to the rapid determination of four primary constituents including total sugar, reducing sugar, total nitrogen and nicotine in Nicotiana plant. Partial least squares regressions was performed after ICO algorithm. The full spectrum was divided into forty equal-width intervals, and the interval with lower root mean squared error of cross-validation was selected for further analysis. As a result, only 155 variables were retained from 1555 variables for each constituent. Particularly, as a variables selection method, ICO improved the prediction accuracy of calibration model and obtained a satisfactory result compared with full-spectrum data. Results revealed that NIR combined with ICO could be efficiently used for rapid analysis of quality associated constituents of Nicotiana plant. Moreover, this study provided a supplementary verification of the proposed variable selection method for the further applications.

Fourier transform infrared spectroscopy and chemometrics for the discrimination of paper relic types.

The paper relic identification is a pending issue to be resolved in the field of cultural heritage. As we all known, heritage paper has significant importance in archaeological research. Nowadays, there are a variety of research methodologies focuses on the analysis of inks for dating documents. While the paper analysis attained little attention. This work is to explore the non-destructive application of ATR-FTIR technique in discrimination of paper relics. 15 types of paper spectra were collected by ATR-FTIR, which wavenumber range were range from 4000 to 650 cm-1. And the moving average smoothing and normalization was used for pretreatment analysis. Five different classification algorithms, principal component analysis-linear discriminant analysis (PCA-LDA), partial least squares discriminant analysis (PLS-DA), soft independent modeling of class analogy (SIMCA), least squares-support vector machine (LS-SVM), partial least squares-linear discriminant analysis (PLS-LDA) were selected to classify the types of paper. PLS-LDA and LS-SVM are effective techniques with 100% classification accuracy. PCA-LDA, PLS-DA and SIMCA give accuracy of 98.67%, 97.33% and 95.56%, respectively. The present experiment suggested that ATR-FTIR combining with chemometrics will be highly useful in paper identification of cultural heritage.

Moving window smoothing on the ensemble of competitive adaptive reweighted sampling algorithm.

A novel chemometrical method, named as MWS-ECARS, which is based on using the moving window smoothing upon an ensemble of competitive adaptive reweighted sampling, is proposed as the spectral variable selection approach for multivariate calibration in this study. In terms of elimination of uninformative variables, an ensemble of CARS is carried out first and MWS is then performed to search for effective variables around the high frequency variables. The variable subset with the lowest standard error of cross-validation (SECV) is treated as the optimal threshold and the corresponding moving window width is regarded as the optimal window width. The method was applied to mid-infrared (MIR) spectra of active ingredient in pesticide, near-infrared (NIR) spectra of soil organic matter and NIR spectra of total nitrogen in Solanaceae plants for variable selection. Overall results show that MWS-ECARS is a promising selection method with an improved prediction performance over three variable selection methods of variable importance projection (VIP), uninformative variables elimination (UVE) and genetic algorithms (GA).

An ensemble variable selection method for vibrational spectroscopic data analysis.

Wavelength selection is a critical factor for pattern recognition of vibrational spectroscopic data. Not only does it alleviate the effect of dimensionality on an algorithm's generalization performance, but it also enhances the understanding and interpretability of multivariate classification models. In this study, a novel partial least squares discriminant analysis (PLSDA)-based wavelength selection algorithm, termed ensemble of bootstrapping space shrinkage (EBSS), has been devised for vibrational spectroscopic data analysis. In the algorithm, a set of subsets are generated from a data set using random sampling. For an individual subset, a feature space is determined by maximizing the expected 10-fold cross-validation accuracy with a weighted bootstrap sampling strategy. Then an ensemble strategy and a sequential forward selection method are applied to the feature spaces to select characteristic variables. Experimental results obtained from analysis of real vibrational spectroscopic data sets demonstrate that the ensemble wavelength selection algorithm can reserve stable and informative variables for the final modeling and improve predictive ability for multivariate classification models.

A modification of the bootstrapping soft shrinkage approach for spectral variable selection in the issue of over-fitting, model accuracy and variable selection credibility.

In this study, we proposed a new computational method stabilized bootstrapping soft shrinkage approach (SBOSS) for variable selection based on bootstrapping soft shrinkage approach (BOSS) which can enhance the analysis of chemical interest from the massive variables among the overlapped absorption bands. In SBOSS, variable is selected by the index of stability of regression coefficients instead of regression coefficients absolute value. In each loop, a weighted bootstrap sampling (WBS) is applied to generate sub-models, according to the weights update by conducting model population analysis (MPA) on the stability of regression coefficients (RC) of these sub-models. Finally, the subset with the lowest RMSECV is chosen to be the optimal variable set. The performance of the SBOSS was evaluated by one simulated dataset and three NIR datasets. The results show that SBOSS can select the fewer variables and supply the least RMSEP and latent variable number of the PLS model with the best stability comparing with methods of Monte Carlo uninformative variables elimination (MCUVE), genetic algorithm (GA), competitive reweighted sampling (CARS), stability of competitive adaptive reweighted sampling (SCARS) and BOSS.

Spectral interval optimization on rapid determination of prohibited addition in pesticide by ATR-FTIR.

BACKGROUND: Acetamiprid, as a low-toxicity pesticide, has already been extensively used to increase plant production and quality. Although fipronil has been prohibited, it is usually illicitly added to acetamiprid due to its particular insecticidal action and effect, so it is highly desirable to obtain a rapid and effective method to detect its concentration. Mid-infrared spectroscopy (MIR) combined with two variable selection methods, interval combination optimization (ICO) and interval partial least squares (iPLS), were used to determinate the prohibited addition of fipronil. RESULTS: The full spectra for both ICO and iPLS were divided into 40 equal-width intervals. Consequently, 45 and 135 characteristic variables were extracted from ICO and iPLS to establish the models. Compared with iPLS, the ICO model acquired a more suitable spectral region and as a result gained a higher prediction accuracy. Specifically, the ICO method selected the characteristic wavelengths ascribed to CF and CN (in five-membered heterocyclics), iPLS chose the intervals associated with CF and SO. CONCLUSION: Results revealed that MIR combined with ICO could be efficiently used for rapid identification of illegal addition and had great potential to provide on-site pesticide quality control. © 2018 Society of Chemical Industry.

Towards improvement in prediction of iodine value in edible oil system based on chemometric analysis of portable vibrational spectroscopic data.

Iodine value (IV) is a significant parameter to illustrate the quality of edible oil. In this study, three portable spectroscopy devices were employed to determine IV in mixed edible oil system, a new Micro-Electro-Mechanical-System (MEMS) Fourier Transform Infrared Spectrometer (MEMS-FTIR), a MicroNIRTM1700 and an i-Raman Plus-785S. Quantitative model was built by Partial least squares (PLS) regression model and four variable selection methods were applied before PLS model, which are Monte Carlo uninformative variables elimination (MCUVE), competitive reweighted sampling (CARS), bootstrapping soft shrinkage approach (BOSS) and variable combination population analysis (VCPA). The coefficient of determination (R2), and the root mean square error prediction (RMSEP) were used as indicators for the predictability of the PLS models. In MicroNIRTM1700 dataset, MCUVE gave the lowest RMSEP (2.3440), in MEMS-FTIR dataset, CARS showed the best performance with RMSEP (2.2185), in i-Raman Plus-785S dataset, BOSS gave the lowest RMSEP (2.5058). They all had great improvements than full spectrum PLS model. Four variable selection methods take a smaller number of variables and perform significant superiority in prediction accuracy. It was demonstrated that three new portable instruments would be suitable for the on-site determination of edible oil quality in infrared and Raman field.

Rapid qualitative and quantitative analysis of methamphetamine, ketamine, heroin, and cocaine by near-infrared spectroscopy.

Rapid and nondestructive near infrared spectroscopy (NIR) methods have been developed for simultaneous qualitative and quantitative analysis of methamphetamine, ketamine, heroin, and cocaine in seized samples. This is the first systematic report regarding a qualitative and quantitative procedure of applying NIR for drug analysis. A total of 282 calibration samples and 836 prediction samples were used for the building and validating of qualitative and quantitative models. Two qualitative analysis modeling methods for soft independent modeling by class analogy (SIMCA) and supporting vector machine (SVM) were compared. From its excellent performance in rejecting false positive results, SIMCA was chosen. The drug concentrations in the calibration and validation sample sets were analyzed using high-performance liquid chromatography. Based on the use of first-order derivative spectral data after standard normal variate (SNV) transformation correction, in the wavelength range from 10,000 to 4000cm-1, four partial least squares quantitative-analysis models were built. The coefficients of determination for all calibration models were >99.3, and the RMSEC, RMSECV, and RMSEP were all less than 1.6, 2.9, and 3.6%, respectively. The results obtained here indicated that NIR with chemometric methods was accurate for qualitative and quantitative analysis of drug samples. This methodology provided a potentially useful alternative to time-consuming gas chromatography-mass spectroscopy and high-performance liquid chromatography methods.

Geometric search: A new approach for fitting PARAFAC2 models on GC-MS data.

PARAFAC2 is a powerful decomposition method which is ideally suited for modeling gas chromatography-mass spectrometry (GC-MS) data. However, the most widely used fitting algorithms (alternating least squares, ALS) are very slow which hinders use of the model. In this paper, an iterative method called geometric search is proposed to fit the PARAFAC2 model. This method models the PARAFAC2 loading parameters as geometric sequences with offsets during the ALS iterations. It extrapolates the optimal parameters from prior iterations to accelerate ALS convergence process. The performance of this method was evaluated by simulated datasets and two GC-MS datasets of wine and tobacco samples. This geometric search method proved an efficient way to fit PARAFAC2 models, compared with a standard ALS algorithm and two widely used line search algorithms in terms of convergence speed and fitting quality.

Quantitative determination of additive Chlorantraniliprole in Abamectin preparation: Investigation of bootstrapping soft shrinkage approach by mid-infrared spectroscopy.

A novel method, mid-infrared (MIR) spectroscopy, which enables the determination of Chlorantraniliprole in Abamectin within minutes, is proposed. We further evaluate the prediction ability of four wavelength selection methods, including bootstrapping soft shrinkage approach (BOSS), Monte Carlo uninformative variable elimination (MCUVE), genetic algorithm partial least squares (GA-PLS) and competitive adaptive reweighted sampling (CARS) respectively. The results showed that BOSS method obtained the lowest root mean squared error of cross validation (RMSECV) (0.0245) and root mean squared error of prediction (RMSEP) (0.0271), as well as the highest coefficient of determination of cross-validation (Qcv2) (0.9998) and the coefficient of determination of test set (Q2test) (0.9989), which demonstrated that the mid infrared spectroscopy can be used to detect Chlorantraniliprole in Abamectin conveniently. Meanwhile, a suitable wavelength selection method (BOSS) is essential to conducting a component spectral analysis.

A novel algorithm for spectral interval combination optimization.

In this study, a new wavelength interval selection algorithm named as interval combination optimization (ICO) was proposed under the framework of model population analysis (MPA). In this method, the full spectra are divided into a fixed number of equal-width intervals firstly. Then the optimal interval combination is searched iteratively under the guide of MPA in a soft shrinkage manner, among which weighted bootstrap sampling (WBS) is employed as random sampling method. Finally, local search is conducted to optimize the widths of selected intervals. Three NIR datasets were used to validate the performance of ICO algorithm. Results show that ICO can select fewer wavelengths with better prediction performance when compared with other four wavelength selection methods, including VISSA, VISSA-iPLS, iVISSA and GA-iPLS. In addition, the computational intensity of ICO is also economical, benefit from fewer tune parameters and faster convergence speed.