Least-squares Regression Model Research Articles

Determination of antioxidant capacity and phenolic content of haskap berries (Lonicera caerulea L.) by attenuated total reflectance-Fourier transformed-infrared spectroscopy

The total phenolic content (TPC) and antioxidant capacity (TAC) of haskap berries cultivated in various locations across Alberta were analyzed using attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy. The Folin–Ciocalteu assay was used to determine TPC, while TAC was quantified by 2,2-diphenyl-1-picrylhydrazl radicals (DPPH) assay and oxygen radical absorbance capacity (ORAC) assay. Three tenfold cross-validated partial least-squares regression (PLSR) models and three fivefold cross-validated deep learning models were developed separately based on FT-IR spectra collected from 22 haskap berry samples and their corresponding reference values determined through Folin-Ciocalteu, DPPH, and ORAC assays. The deep learning models (R2 = 0.95, 0.93, and 0.90 for Folin-Ciocalteu, DPPH, and ORAC assays, respectively) demonstrated better prediction capability compared to the PLSR models (R2 = 0.74, 0.72, and 0.66 for Folin-Ciocalteu, DPPH, and ORAC assays, respectively). In addition, PLS loading plots indicated that phenolic contents and polysaccharides in haskap berries could contribute to their antioxidant capacity. Using ATR-FTIR to estimate the TPC and TAC of fruits offers a rapid alternative to the conventional chemical assays.

Characterization and Quantification of Naphthenic Acids in Produced Water by Orbitrap MS and a Multivariate Approach.

Naphthenic acids (NAs) naturally occur in crude oil and its associated produced water, presenting significant challenges, such as corrosion, in refinery apparatus and ecotoxicity in aquatic habitats. This study delineates a multivariate method to quantify NAs in produced water via electrospray ionization coupled with high-resolution Orbitrap mass spectrometry (ESI-Orbitrap MS). By employing liquid-liquid extraction, followed by direct infusion ESI(-)-Orbitrap MS, we characterized and quantified NAs employing a partial least-squares regression (PLS) model enhanced by the ordered predictor selection (OPS) algorithm. Thirty-six produced water samples were utilized, with 24 allocated for calibration and 12 designated for validation. The PLS-OPS model demonstrated notable accuracy in predicting NA concentrations in simulated and actual produced water samples ranging from ∼30 to 300 mg·L-1. This methodology offers a rapid yet robust alternative for quantifying NAs using mass spectrometry augmented by PLS and the OPS. Its significance is underscored by its potential to equip the petroleum industry with a swift and reliable monitoring mechanism for NAs in produced water, thereby aiding in mitigating environmental and operational risks.

In Situ Eutectic Formation in a Polymeric Matrix via Hot-Melt Reactive Extrusion and the Use of Partial Least Squares Regression Modeling for Reaction Yield Determination.

There has been a significant volume of work investigating the design and synthesis of new crystalline multicomponent systems via examining complementary functional groups that can reliably interact through the formation of noncovalent bonds, such as hydrogen bonds (H-bonds). Crystalline multicomponent molecular adducts formed using this approach, such as cocrystals, salts, and eutectics, have emerged as drug product intermediates that can lead to effective drug property modifications. Recent advancement in the production for these multicomponent molecular adducts has moved from batch techniques that rely upon intensive solvent use to those that are solvent-free, continuous, and industry-ready, such as reactive extrusion. In this study, a novel eutectic system was found when processing albendazole and maleic acid at a 1:2 molar ratio and successfully prepared using mechanochemical methods including liquid-assisted grinding and hot-melt reactive extrusion. The produced eutectic was characterized to exhibit a 100 °C reduction in melting temperature and enhanced dissolution performance (>12-fold increase at 2 h point), when compared to the native drug compound. To remove handling of the eutectic as a formulation intermediate, an end-to-end continuous-manufacturing-ready process enables feeding of the raw parent reagents in their respective natural forms along with a chosen polymeric excipient, Eudragit EPO. The formation of the eutectic was confirmed to have taken place in situ in the presence of the polymer, with the reaction yield determined using a multivariate calibration model constructed by combining spectroscopic analysis with partial least-squares regression modeling. The ternary extrudates exhibited a dissolution profile similar to that of the 1:2 prepared eutectic, suggesting a physical distribution (or suspension) of the in situ synthesized eutectic contents within the polymeric matrix.

Fuzzy Neighborhood-Based Manifold Learning and Feature Weight Matrix for Multilabel Feature Selection

In recent years, it has been difficult to determine neighborhood radii with almost all feature selection methodologies, and the existing multilabel feature selection approaches neglect these correlations among labels, which results in poor classification efficiency. To address these flaws, a fuzzy neighborhood-based manifold learning model and a feature weight matrix are developed to study multilabel feature selection in this article. First, the difference coefficient of the given samples and the new classification margin of each sample are defined to construct an adaptive fuzzy neighborhood radius. While considering the influences of features and labels on the similarity relationships of samples, a novel fuzzy neighborhood similarity relationship is developed by combining it with the label similarity approach that uses the Jaccard similarity coefficient. Second, the least-squares regression model is introduced to design a loss function for learning the mapping relationship between the feature and label spaces. To constrain the objective function, a regularization term based on the fuzzy neighborhood similarity relationship matrix is developed to establish a manifold learning model. To reduce the complexity of the loss function and to reduce the induced structural risk, the final objective function is constructed by incorporating two graph Laplacian matrices. Third, positive and negative confidence values are presented to incorporate the frequency parameter and to form the dependencies between labels into a new label correlation matrix. A new feature weight matrix integrated with the dependencies between labels is defined to learn the correlations between features and labels. Finally, an embedded feature selection method with an objective function is designed, where the objective function is optimized by the gradient descent scheme. Experiments conducted on 12 multilabel datasets reveal that the proposed approach yields better classification results than do other algorithms.

A Complementary Multitechnique Approach to Assess the Bias in Molecular Weight Determination of Lignin by Derivatization-Free Gel Permeation Chromatography.

The growing interest in lignin valorization in the past decades calls for analytical techniques for lignin characterization, ranging from wet chemistry techniques to highly sophisticated chromatographic and spectroscopic methods. One of the key parameters to consider is the molecular weight profile of lignin, which is routinely determined by size-exclusion chromatography; however, this is by no means straightforward and is prone to being hampered by considerable errors. Our study expands the fundamental understanding of the bias-inducing mechanisms in gel permeation chromatography (GPC), the magnitude of error originating from using polystyrene standards for mass calibration, and an evaluation of the effects of the solvent and type of lignin on the observed bias. The developed partial least-squares (PLS) regression model for lignin-related monomers revealed that lignin is prone to association mainly via hydrogen bonding. This hypothesis was supported by functional group-based analysis of the bias as well as pulse field gradient (pfg) diffusion NMR spectroscopy of model compounds in THF-d8. Furthermore, although the lack of standards hindered drawing conclusions based on functionalities, direct infusion electrospray ionization mass spectrometry indicated that the relative bias decreases considerably for higher molecular weight species. The results from pfg-diffusion NMR spectroscopy on whole lignin samples were comparable when the same solvents were used in both experiments; in addition, the comparison between results obtained by pfg-diffusion NMR in different solvents gives some additional insights into the aggregation.

Data-Driven Modeling of the Spray Drying Process. Process Monitoring and Prediction of the Particle Size in Pharmaceutical Production.

Spray drying is used in the pharmaceutical industry for particle engineering of amorphous solid dispersions (ASDs). The particle size of the spray-dried (SD) powders is one of their key attributes due to its impact on the downstream processes and the drug product's functional properties. Offline and inline laser diffraction methods can be used to estimate the product's particle size; however, the final release of these ASDs is based on offline instruments. This paper presents a novel data-driven modeling approach for predicting the particle size of SD products. The model-based characterization of the process and the product's particle size, as a critical quality attribute, follows the quality by design principles. The resulting model can be used for online process monitoring, reducing the risks of out-of-specifications products and supporting their real-time release. A Tucker3 model is trained to capture and factorize the deterministic variability of the process. Subsequently, a partial least-squares regression model is calibrated to model the impact that variability in the input material properties, the process parameters, and the spray nozzle have on the products' particle size. This strategy has been calibrated and validated using large scale production data for two intermediate drug products under high sparsity of particle size data. Despite the challenges, high accuracy was obtained in predicting the median particle size (dv50) for release. The 99% confidence interval results in an error of maximum 2.5 μm, which is less than 10% of the allowed range of variation.

Rapid monitoring of the spatial distribution of soil organic matter using unmanned aerial vehicle imaging spectroscopy

ABSTRACT An existing technique for quickly obtaining the spatial distribution data of soil organic matter (SOM) involves combining an unmanned aerial vehicle (UAV) platform and imaging spectroscopy technology; this technique is suitable for precision agricultural management and real-time monitoring. In this study, the average spectral data were first extracted via circular regions of interest (ROIs) of different radii and modelled using the measured SOM to determine the optimal extraction range for the spectral data. Partial least-squares regression (PLSR) and support vector machine (SVM) models were constructed based on the raw spectra, standard normal transform spectra, multiple scattering-corrected spectra, and first-order differential spectra. The difference index, ratio index, and normalized index were calculated and analysed for their correlation with SOM content, and the spectral indices were screened to establish the SOM prediction model. Finally, the accuracy of the model was evaluated, and we selected the optimal model to map the spatial distribution of SOM. The results of our study show that 1) the radius of the ROI significantly affects the model’s accuracy; 2) the accuracy of the nonlinear SVM model is much higher than that of the linear PLSR model; 3) the prediction model based on spectral indices is better than full-band model. This study can serve as a model for the application of UAV imaging spectroscopy technology to field-scale SOM estimations and rapid monitoring, as well as provide technical support for precision agriculture and climate change studies.

Predicting Molecular Weight Characteristics of Reductively Depolymerized Lignins by ATR-FTIR and Chemometrics.

Recent scientific advances in the valorization of lignin, through e.g., (partial-)catalytic depolymerization, require equally state-of-the-art approaches for the analysis of the obtained depolymerized lignins (DLs) or lignin bio-oils. The use of chemometrics in combination with infrared (IR) spectroscopy is one avenue to provide rapid access to pertinent lignin parameters, such as molecular weight (MW) characteristics, which typically require analysis via time-consuming size-exclusion methods, or diffusion-ordered NMR spectroscopy. Importantly, MW serves as a marker for the degree of depolymerization (or recondensation) that the lignin has undergone, and thus probing this parameter is essential for the optimization of depolymerization conditions to achieve DLs with desired properties. Here, we show that our ATR-IR-based chemometrics approach used previously for technical lignin analysis can be extended to analyze these more processed, lignin-derived samples as well. Remarkably, also at this lower end of the MW scale, the use of partial least-squares (PLS) regression models well-predicted the MW parameters for a sample set of 57 depolymerized lignins, with relative errors of 9.9-11.2%. Furthermore, principal component analysis (PCA) showed good correspondence with features in the regression vectors for each of the biomass classes (hardwood, herbaceous/grass, and softwood) obtained from PLS-discriminant analysis (PLS-DA). Overall, we show that the IR spectra of DLs are still amenable to chemometric analysis and specifically to rapid, predictive characterization of their MW, circumventing the time-consuming, tedious, and not generally accessible methods typically employed.

Alien range size, habitat breadth, origin location, and domestication of alien species matter to their impact risks.

Invasive alien species are a major driver of biodiversity loss. Currently, the process of biological invasions is experiencing a constant acceleration, foreshadowing a future increase in the threat posed by invasive alien species to global biodiversity. Therefore, it is necessary to assess the impact risks of invasive alien species and related factors. Here, we constructed a dataset of negative environmental impact events to evaluate the impact risks of alien species. We collected information on 1071 established alien terrestrial vertebrates and then gathered negative environmental impacts for 108 of those species. Generalized linear mixed-effects model and phylogenetic generalized least-squares regression model were used to examine the characteristic (including life-history traits, characteristics related to distribution, and introduction event characteristics) correlates of species' impact risks at the global scale for 108 established alien terrestrial vertebrates (mammals, birds, reptiles and amphibians). Our results showed that a total of 3158 negative environmental impacts were reported for 108 harmful species across 71 countries worldwide. Factors associated with impact risks varied slightly among taxa, but alien range size, habitat breadth, origin location, and domestication were significantly correlated with impact risks. Our study aims to identify the characteristics of alien species with high-impact risks to facilitate urgent assessment of alien species and to protect the local ecological environment and biodiversity.

Minorities' Diminished Psychedelic Returns: Income and Educations Impact on Whites, Blacks, Hispanics, and Asians.

Growing evidence suggests that the race and ethnic minority population may experience fewer protective effects of psychedelics on mental health. The minority diminished psychedelic returns theory proposes that racism, manifested in socioeconomic inequality, could partially account for the smaller health gains observed. Therefore, it is important to investigate whether socioeconomic inequality reduces the impact of psychedelics on health outcomes for minority populations. Additionally, despite having higher socioeconomic status, it remains unclear whether psychedelic use among minorities is associated with the same level of health benefits as observed in non-Hispanic whites. This study utilizes data from the National Survey of Drug Use (N = 2008 to 2019), which involved 458,372 participants aged 18 and above. The objective is to examine the impact of various psychedelics (MDMA, psilocybin, DMT, ayahuasca, peyote/mescaline, and LSD), as well as lifetime classic psychedelics use (LCPU), on psychological distress in the past month, taking into account socioeconomic factors (education level and family income) and race/ethnic differences (White, Black, Hispanic, and Asian). The analysis employed a series of nested ordinary least-square regression models using Stata 18. The results indicate that, after controlling for socioeconomic status, there is no association between Black and Hispanic psychedelic use and distress. However, white psychedelic use remains associated with lower levels of distress. Additionally, despite having higher levels of education and income, psychedelic use among minority groups does not appear to be linked to reduced stress. In fact, for Asians with higher education and income, certain psychedelic use is associated with increased distress.

Effects of Chinese “double carbon strategy” on soil polycyclic aromatic hydrocarbons pollution

Polycyclic aromatic hydrocarbons (PAHs) and carbon dioxide primarily originate from the combustion of fossil fuels and biomass. The implementation of the Chinese “double carbon strategy” is expected to impact the distribution of PAH emissions, consequently influencing the spatial distribution trend of PAHs in surface soil. Therefore, it is crucial to quantitatively evaluate the effectiveness of the Chinese “double carbon strategy” on soil PAH pollution for the purpose of “the reduction of pollution and carbon emissions”. This study utilized 15,088 individual PAH concentration data from 943 soil samples collected between 2003 and 2020 in China, in conjunction with PAH emissions at a 10 km resolution, for meta-analysis. The calculated PAH emissions in this study are in line with the global PAH emission inventory (PKU-PAH-2007), with a relative standard deviation at the provincial level of less than 25 %. Subsequently, a novel method was developed using emission density and Kow of PAHs to predict PAH concentrations in surface soil based on a least-squares regression model. Compared to other environmental models, the method established in this study significantly reduced the percent sample deviation to less than 70 %. Furthermore, energy consumption data for China were simulated based on the implementation plan of the “double carbon strategy” to project PAH emissions and soil PAH levels for the years 2030 and 2060. The predicted PAH emissions in China were estimated to decrease to 41,300 t in 2030 and 10,406.5 t in 2060 from 78,815 t in 2020. Moreover, the heavily contaminated areas of soil PAHs (i.e., total PAH concentrations in soil exceeding 1000 μg kg−1) were projected to decrease by 45 % and 82 % in 2030 and 2060, respectively, compared to levels in 2020. These findings suggest that the implementation of the “double carbon strategy” can fundamentally reduce the pollution of PAHs in surface soil of China.

Estimating the Aboveground Fresh Weight of Sugarcane Using Multispectral Images and Light Detection and Ranging (LiDAR)

This study aimed to develop a remote sensing method for estimating the aboveground fresh weight (AGFW) of sugarcane using multispectral images and light detection and ranging (LiDAR). Remotely sensed data were acquired from an unmanned aerial vehicle (drone). Sample plots were harvested and the AGFW of each plot was measured. Sugarcane crown heights and volumes were obtained by isolating individual tree crowns using a LiDAR-derived digital surface model of the area. Multiple linear regression (MLR) and partial least-squares regression (PLSR) models were tested for the field-sampled AGFWs (dependent variable) and individual canopy heights and volumes, and spectral indices were used as independent variables or predictors. The PLSR model showed more promising results than the MLR model when predicting the AGFW over the study area. Although PLSR is well-suited to a large number of collinear predictor variables and a limited number of field samples, this study showed moderate results (R2 = 0.5). The visual appearance of the spatial distribution of the AGFW map is satisfactory. The limited no. of field samples overfitted the MLR prediction results. Overall, this research highlights the potential of integrating remote sensing technologies in the sugarcane industry, thereby improving yield estimation and effective crop management.

Rapid estimates of leaf litter chemistry using reflectance spectroscopy

Measuring the chemical traits of leaf litter is important for understanding plants’ influence on nutrient cycles, including through nutrient resorption and litter decomposition, but conventional leaf trait measurements are often destructive and labor-intensive. Here, we develop and evaluate the performance of partial least-squares regression models that use reflectance spectra of intact or ground leaves to estimate leaf litter traits, including carbon and nitrogen concentration, carbon fractions, and leaf mass per area (LMA). Our analyses included more than 300 samples of senesced foliage from 11 species of temperate trees, including both needleleaf and broadleaf species. Across all samples, we could predict each trait with moderate-to-high accuracy from both intact-leaf litter spectra (validation R2 = 0.543–0.941; %root mean squared error (RMSE) = 7.49–18.5) and ground-leaf litter spectra (validation R2 = 0.491–0.946; %RMSE = 7.00–19.5). Notably, intact-leaf spectra yielded better predictions of LMA. Our results support the feasibility of building models to estimate multiple chemical traits from leaf litter of a range of species. In particular, intact-leaf spectral models allow non-destructive trait estimation in a matter of seconds, which could enable researchers to measure the same leaves over time in studies of nutrient resorption.

The Effect of Corporate Governance on the Degree of Agency Cost in the Korean Market

This study examines the relationship between corporate governance (CG) and agency costs using Korean market data, particularly for chaebol firms. The final sample includes 660 firm-year observations between 2016 and 2020 for Korean non-financial firms listed on the Korean Composite Stock Price Index (KOSPI). This study employs an ordinary least-squares panel data regression model using two proxies for agency costs, namely, asset utilization ratio and operating expense ratio, and six CG individual metrics as independent variables (CG score, protection of shareholder rights, board structure, disclosure, audit organization, and managerial discretion and error management). We find that firms with high CG experience lower agency costs than those with low CG. Moreover, our evidence suggests that firms can decrease agency costs by improving the quality of CG. The results of our regression model also support the idea that CG is effective in reducing agency costs for chaebol firms but not for non-chaebol firms. Finally, our findings suggest that the implementation of effective CG mechanisms in firms might improve managerial behavior through better decision-making to maximize the value of firms.

Estimated Sustainable Cost-Based Prices for Diabetes Medicines

The burden of diabetes is growing worldwide. The costs associated with diabetes put substantial pressure on patients and health budgets, especially in low- and middle-income countries. The prices of diabetes medicines are a key determinant for access, yet little is known about the association between manufacturing costs and current market prices. To estimate the cost of manufacturing insulins, sodium-glucose cotransporter 2 inhibitors (SGLT2Is), and glucagonlike peptide 1 agonists (GLP1As), derive sustainable cost-based prices (CBPs), and compare these with current market prices. In this economic evaluation, the cost of manufacturing insulins, SGLT2Is, and GLP1As was modeled. Active pharmaceutical ingredient cost per unit (weighted least-squares regression model using data from a commercial database of trade shipments, data from January 1, 2016, to March 31, 2023) was combined with costs of formulation and other operating expenses, plus a profit margin with an allowance for tax, to estimate CBPs. Cost-based prices were compared with current prices in 13 countries, collected in January 2023 from public databases. Countries were selected to provide representation of different income levels and geographic regions based on the availability of public databases. Estimated CBPs; lowest current market prices (2023 US dollars). In this economic evaluation of manufacturing costs, estimated CBPs for treatment with insulin in a reusable pen device could be as low as $96 (human insulin) or $111 (insulin analogues) per year for a basal-bolus regimen, $61 per year using twice-daily injections of mixed human insulin, and $50 (human insulin) or $72 (insulin analogues) per year for a once-daily basal insulin injection (for type 2 diabetes), including the cost of injection devices and needles. Cost-based prices ranged from $1.30 to $3.45 per month for SGLT2Is (except canagliflozin: $25.00-$46.79) and from $0.75 to $72.49 per month for GLP1As. These CBPs were substantially lower than current prices in the 13 countries surveyed. High prices limit access to newer diabetes medicines in many countries. The findings of this study suggest that robust generic and biosimilar competition could reduce prices to more affordable levels and enable expansion of diabetes treatment globally.

Non-Destructive Near-Infrared Technology for Efficient Cannabinoid Analysis in Cannabis Inflorescences.

In the evolving field of cannabis research, scholars are exploring innovative methods to quantify cannabinoids rapidly and non-destructively. This study evaluates the effectiveness of a hand-held near-infrared (NIR) device for quantifying total cannabidiol (total CBD), total delta-9-tetrahydrocannabinol (total THC), and total cannabigerol (total CBG) in whole cannabis inflorescences. Employing pre-processing techniques, including standard normal variate (SNV) and Savitzky-Golay (SG) smoothing, we aim to optimize the portable NIR technology for rapid and non-destructive cannabinoid analysis. A partial least-squares regression (PLSR) model was utilized to predict cannabinoid concentration based on NIR spectra. The results indicated that SNV pre-processing exhibited superior performance in predicting total CBD concentration, yielding the lowest root mean square error of prediction (RMSEP) of 2.228 and the highest coefficient of determination for prediction (R2P) of 0.792. The ratio of performance to deviation (RPD) for total CBD was highest (2.195) with SNV. In contrast, raw data exhibited the least accurate predictions for total THC, with an R2P of 0.812, an RPD of 2.306, and an RMSEP of 1.651. Notably, total CBG prediction showed unique characteristics, with raw data yielding the highest R2P of 0.806. SNV pre-processing emerges as a robust method for precise total CBD quantification, offering valuable insights into the optimization of a hand-held NIR device for the rapid and non-destructive analysis of cannabinoid in whole inflorescence samples. These findings contribute to ongoing efforts in developing portable and efficient technologies for cannabinoid analysis, addressing the increasing demand for quick and accurate assessment methods in cannabis cultivation, pharmaceuticals, and regulatory compliance.

Comparison of Machine Learning Models in Simulating Glacier Mass Balance: Insights from Maritime and Continental Glaciers in High Mountain Asia

Accurately simulating glacier mass balance (GMB) data is crucial for assessing the impacts of climate change on glacier dynamics. Since physical models often face challenges in comprehensively accounting for factors influencing glacial melt and uncertainties in inputs, machine learning (ML) offers a viable alternative due to its robust flexibility and nonlinear fitting capability. However, the effectiveness of ML in modeling GMB data across diverse glacier types within High Mountain Asia has not yet been thoroughly explored. This study addresses this research gap by evaluating ML models used for the simulation of annual glacier-wide GMB data, with a specific focus on comparing maritime glaciers in the Niyang River basin and continental glaciers in the Manas River basin. For this purpose, meteorological predictive factors derived from monthly ERA5-Land datasets, and topographical predictive factors obtained from the Randolph Glacier Inventory, along with target GMB data rooted in geodetic mass balance observations, were employed to drive four selective ML models: the random forest model, the gradient boosting decision tree (GBDT) model, the deep neural network model, and the ordinary least-square linear regression model. The results highlighted that ML models generally exhibit superior performance in the simulation of GMB data for continental glaciers compared to maritime ones. Moreover, among the four ML models, the GBDT model was found to consistently exhibit superior performance with coefficient of determination (R2) values of 0.72 and 0.67 and root mean squared error (RMSE) values of 0.21 m w.e. and 0.30 m w.e. for glaciers within Manas and Niyang river basins, respectively. Furthermore, this study reveals that topographical and climatic factors differentially influence GMB simulations in maritime and continental glaciers, providing key insights into glacier dynamics in response to climate change. In summary, ML, particularly the GBDT model, demonstrates significant potential in GMB simulation. Moreover, the application of ML can enhance the accuracy of GMB modeling, providing a promising approach to assess the impacts of climate change on glacier dynamics.

The least-square support vector regression model for the dyes and heavy metal ions removal prediction

Wastewater treatment plants are typically complex because they involve physical, chemical, and biological processes. Meanwhile, the efficiency of the removal of pollutants such as auramine (AO), methylene blue (MB), and ion Cadmium (Cd (II)) from the wastewater are difficult to be measured directly in real-time, as this measurement requires laboratory instruments that are time-consuming. The soft sensors could be the solution to perform the real-time prediction of the AO, MB, and Cd (II) removals’ capability. Hence, this study investigates the performances of a soft sensor, namely least-square support vector regression (LSSVR) to estimate the AO, MB, and Cd (II) removals’ ability. In this study, two wastewater-related case studies involving AO, MB, and Cd (II) removals were used to evaluate the predictive performance of the LSSVR. Additionally, its results were compared and analyzed with other soft sensors. For both case studies, notice that LSSVR gives the best results for AO, MB, and Cd (II) removals as compared to other soft sensor models where its root means square errors, mean absolute errors, and the approximate error, are lowered by 83% to 1,756%. Moreover, its coefficients of determination, denoted R 2 are the highest which are all more than or close to 0.9 for all the AO, MB, and Cd (II) removals even for the testing data for the case studies that were not used to develop the LSSVR model. In conclusion, LSSVR is more suitable for evaluating the effectiveness of the AO, MB, and Cd (II) removals at present.

Generative data augmentation and automated optimization of convolutional neural networks for process monitoring.

Chemometric modeling for spectral data is considered a key technology in biopharmaceutical processing to realize real-time process control and release testing. Machine learning (ML) models have been shown to increase the accuracy of various spectral regression and classification tasks, remove challenging preprocessing steps for spectral data, and promise to improve the transferability of models when compared to commonly applied, linear methods. The training and optimization of ML models require large data sets which are not available in the context of biopharmaceutical processing. Generative methods to extend data sets with realistic in silico samples, so-called data augmentation, may provide the means to alleviate this challenge. In this study, we develop and implement a novel data augmentation method for generating in silico spectral data based on local estimation of pure component profiles for training convolutional neural network (CNN) models using four data sets. We simultaneously tune hyperparameters associated with data augmentation and the neural network architecture using Bayesian optimization. Finally, we compare the optimized CNN models with partial least-squares regression models (PLS) in terms of accuracy, robustness, and interpretability. The proposed data augmentation method is shown to produce highly realistic spectral data by adapting the estimates of the pure component profiles to the sampled concentration regimes. Augmenting CNNs with the in silico spectral data is shown to improve the prediction accuracy for the quantification of monoclonal antibody (mAb) size variants by up to 50% in comparison to single-response PLS models. Bayesian structure optimization suggests that multiple convolutional blocks are beneficial for model accuracy and enable transfer across different data sets. Model-agnostic feature importance methods and synthetic noise perturbation are used to directly compare the optimized CNNs with PLS models. This enables the identification of wavelength regions critical for model performance and suggests increased robustness against Gaussian white noise and wavelength shifts of the CNNs compared to the PLS models.

Application of a short-wave pocket-sized near-infrared spectrophotometer to predict milk quality traits

Portable-hand-held devices based on near-infrared (NIR) technology have improved and are gaining popularity, even if their implementation in milk has been barely evaluated. Thus, the aim of the present study was to assess short-wave pocket-sized NIR devices' feasibility to predict milk quality. A total of 331 individual milk samples from different cow breeds and herds were collected in 2 consecutive days for chemical determination and spectral collection by using 2 pocket-sized NIR spectrophotometers working in the range of 740-1070 nm. The reference data was matched with the corresponding spectrum and modified-partial least-squares regression models were developed. A 5-fold cross-validation was applied to evaluate individual devices' performance and an external-validation with 25% of the data set as the validation set was applied for the final models. Results revealed that both devices' absorbance was highly correlated but greater for instrument A than B. Thus, the final models were built by averaging the spectra from both devices for each sample. The fat content prediction model was adequate for quality control with a coefficient of determination (R2ExV) and a residual predictive deviation (RPDExV) in external validation of 0.93 and 3.73, respectively. Protein and casein content as well as fat-to-protein ratio prediction models might be used for a rough screening (R2ExV > 0.70; RPDExV > 1.73). However, poor prediction models were obtained for all the other traits with an R2ExV between 0.43 (urea) and 0.03 (somatic cell count), and a RPDExV between 1.18 (urea) and 0.22 (somatic cell count). In conclusion, short-wave portable-hand-held NIR devices accurately predicted milk fat content, and protein, casein, and fat-to-protein ratio might be applied for rough screening. It seems that there is not enough information in this NIR region to develop adequate prediction models for lactose, somatic cell count, urea, and freezing point.