JNIRS RSS Feed

Exploring spectroscopic regression modelling using Eureqa: black-box chemometrics or a useful ...

“Eureqa (pronounced ‘eureka‘) is a software tool for detecting equations and hidden mathematical relationships in your data” made available by the Cornell Creative Machines Lab. (http://creativemachines.cornell.edu/eureqa). Based on the use of evolutionary genetic programming, the program is capable of testing a wide variety of mathematical functions, and combinations thereof, to find relationships between data. Functions include the standard functions used in multi-linear regression (MLR; +, ?, /, *, k) plus others, such as exponentials, roots, trigonometric etc. The objective of this investigation was to determine if this program might be useful in the investigation and development of spectroscopic calibrations. Two sets (173 or 241 samples) of forages (hays) and by-products (hulls, stalks etc.), some of which had been chemically treated with sodium chlorite to increase digestibility, were studied using Eureqa. Data for six analytes were available, but based on previous work using partial least squares regression, only crude protein, a well-determined analyte, and lignin, a much poorly determined analyte, were examined. Results indicated that standard spectral pre-treatments such as normalisation, mean centring, variance scaling, multiplicative scatter correction and derivatives might be beneficial in more rapidly obtaining the best calibrations, but were not necessary unless needed for data scaling. While overall results were comparable to those from partial least squares, but never as quite as good, several aspects of the program led to the conclusion that it is a wonderful exploration and learning tool for spectroscopy, even if one is not interested in developing MLR-based calibrations. As just one example, the results for all samples are displayed for each equation as developed and any of the best dozen or so can be displayed with a simple mouse click. Thus, it is easy to see how the developing equation effects the predicted versus actual fit, outliers etc. in nearly real time. Similarly, the many function fitting options (different error measures, both least- and non-least squares) allow one to explore and see how these affect both the final results and equation evolution. In conclusion, Eureqa is an exceedingly interesting and useful program, both for developing MLR calibrations and for studies of calibration development in general.

The use of near infrared spectroscopy to determine the fat, caffeine, theobromine and ...

Cocoa is an important raw material in international trade, for which the highest quality is demanded. Venezuelan Criollo cocoa beans from the south of Lake Maracaibo or Chuao are of very high quality and are considered to be Venezuela‘s finest beans. It is therefore important to characterise and define that quality, or its precursors, in the different primary cocoa processing stages. Given the large number of samples to be analysed, new analytical techniques providing fast and reliable quality data are essential. Near infrared (NIR) spectroscopy utilises wavelengths from 780–2500 nm to measure the absorbance by a sample, compute organic functional groups and quantitatively predict a particular factor. It has been extensively used to analyse food quality and to determine the main alkaloids of coffee. Fat, caffeine, theobromine and (–)- epicatechin contents are related to cocoa bean “flavour” quality. These parameters are usually determined using conventional methods, which are time consuming, destructive and expensive. The goals of this study were to use NIR spectroscopy to develop a fast and non-destructive tool to determine compounds in unfermented and sun-dried cocoa beans of high quality. Calibration coefficients of determination (R²) and standard errors of cross-validation and per cent dry matter, were 0.94 and 0.89%, 0.94 and 0.05%, 0.88 and 0.08% and 0.96 and 0.18% for fat, caffeine, theobromine and (–)-epicatechin contents, respectively. The results confirmed the good predictability of the models and showed that NIR spectroscopy can be used as a rapid method for determining these compounds in cocoa beans.

Near infrared quantitative analysis of total flavonoid content in fresh Ginkgo biloba leaves ...

Total flavonoid concentration is often considered an important quality attribute of Ginkgo biloba leaf. Near infrared spectroscopy was used to determine total flavonoid concentration in fresh G. biloba leaf. The spectra of 120 leaf samples were acquired in the wavelength range of 10,000 cm^–1 to 4000 cm^–1. After pre-processing, interval partial least squares (iPLS), synergy interval partial least squares (SiPLS), genetic algorithm interval partial least squares (GA-iPLS) and simulation annealing algorithm interval partial least squares (SAA-iPLS) were used to select the most informative wavelength regions correlated with total flavonoid concentration. The number of wavelength regions and the number of PLS factors were optimised by cross-validation. The performance of the SAA-iPLS model developed in this study was better than PLS, iPLS and GA-iPLS models. The coefficient of determination (r²) and the root mean square error of prediction (RMSEP) for the prediction set samples using the SAA-iPLS model were 0.89 mg g^–1 and 3.0 mg g^–1, respectively. These results show that near infrared spectroscopy combined with SAA-iPLS has significant potential for the non-destructive quantitative analysis of total flavonoids in G. biloba leaf.

Development of near infrared calibrations for physical and mechanical properties of eucalypt ...

Near infrared (NIR) spectroscopy has been used in several studies to predict the physical and mechanical properties of pulp handsheets. In most of these studies, wood samples were pulped in a laboratory under different regimes and/or refined to introduce variability into the data set. This study investigates the potential of NIR spectroscopy to create calibrations for eucalyptus pulp properties of mill-line origin. Seven mechanical properties (air resistance, compressibility, drainability, hygro-expansivity, stretch, tensile index and tensile stiffness) and three physical properties (bulk density, specific volume, and surface area) were investigated. Coefficients of determination (R²) for all 10 properties were poor. The R²_C value exceeded 0.70 for only one property (tensile index), while the R²_CV values exceeded 0.40 for only two properties (drainability and surface area). Ratios of performance to deviation were equally poor, ranging from 0.87 for bulk density to 1.28 for drainability. These statistics indicate that none of the calibrations could be used to accurately predict the properties of unknown samples. The poor performance of the calibrations is likely due to the low variability of our dataset, which is generally inherent in samples of mill-line origin.

Determination of Eucalyptus globulus wood extractives content by near infrared-based partial ...

Wood extractives, the non-cell wall components that can be removed by solvents, can play an important role in the protection of the living tree as well as derived wood products. On the other hand they can be detrimental for pulp and paper, paint and varnish films and adhesives. The objective of this work was to develop near infrared-based partial least squares regression models for the prediction of wood extractives. The developed models are well suited for screening of the ethanol and total extractives content of Eucalyptus globulus wood. The models for the prediction of ethanol extractives with residual prediction deviations above 5 are also suited for quality control. It is shown that samples with high extractives content always have a more intense OH combination band than the samples with low extractives content and that near infrared can be used for a rough estimation of the relative performance of the reference methods.

Estimation of wood basic density of Acacia melanoxylon (R. Br.) by near infrared spectroscopy

Wood basic density is one of the most important wood quality properties and one of the simplest to assess but it is too time consuming to be really useful for the screening of populations or for improvement programmes where large numbers of samples need to be assessed. Although the usefulness of near infrared (NIR) spectroscopy to assess wood properties, including wood density, is well established only a few of the published models are suitable for screening. The NIR-based partial least squares regression models obtained in this study can be used for screening the basic density of the Portuguese blackwood [Acacia melanoxylon (R. Br.)] population with standard errors of cross-validation of only 11 kg m^–3 and values for the residual prediction deviation well above the 2.5 limit. It was also concluded that at least 45 samples for calibration and a further 16 samples for validation are necessary to obtain acceptable models for screening. Even using a very small number of spectra per disc, accurate estimates of wood basic density were obtained.

A non-destructive method of predicting the particle size of the bulk drug powder in an ...

This paper describes a non-destructive method for determining the particle size of the bulk drug powder in an acetaminophen suppository using near infrared (NIR) spectrometry combined with chemometrics. Acetaminophen bulk powder samples were obtained by sieving through 75–1400 µm screens. Suppository samples containing the bulk powder with various particle sizes were obtained at 50°C. They consisted of the bulk powder (4.5 w/w%), hard fat and plastic suppository container. Spectra of 36 standard suppository samples were recorded using a NIR spectrometer, fitted with a fibre-optic probe, in the range of 4000–12500 cm^–1. The spectra were divided to a calibration set and a validation data set. Various pre-treatments were tested and partial least squares regression used to develop the calibration models which were tested against the samples in the validation set. Chemometric calculations were also performed on long range (4000–12,500 cm^–1) and short range (5000–10,000 cm^–1) data sets, respectively. The best model was obtained when using the short range, the pre-treatment of smoothing, standard normal variate plus area normalisation and eight principal components, to achieve a standard error of cross validation of 100 µm. The plot of the predicted values versus the actual values for a validation set of suppositories gave a straight line with a regression coefficient of determination of 0.957 and standard error of prediction of 98 µm. The correlation spectrum of the best model suggested that the specific positive and negative peaks were due to hard fat and drug particle size, respectively. These results demonstrate that it was possible to predict the particle size of acetaminophen bulk powder in a suppository without destroying the sample.