Supervised Pattern Recognition Research Articles

Evaluation of Classical Least Squares Discriminant Analysis (CLS‐DA) as a Novel Supervised Pattern Recognition Technique

ABSTRACTMultivariate calibration techniques and machine learning algorithms are inextricably linked within the realm of chemometrics and data analysis. Classical least squares (CLS) modeling, a fundamental multivariate regression approach, has traditionally been utilized for quantitative analysis tasks, establishing relationships between predictor variables (e.g., spectroscopic data) and response variables (e.g., chemical concentrations). However, a unique feature of CLS is its ability to handle scenarios with partial knowledge of the independent variable matrix, making it an intriguing candidate for qualitative pattern recognition and discriminant analysis applications. This study proposes a novel approach, Classical Least Squares Discriminant Analysis (CLS‐DA), which combines the principles of CLS modeling with discriminant analysis objectives. The performance of CLS‐DA is comprehensively evaluated using two real‐world datasets: chemical analysis of three wine cultivars and mid‐infrared spectroscopy of minced meat samples (pork, chicken, and turkey). The results are compared against the well‐established Partial Least Squares Discriminant Analysis (PLS‐DA) method, a widely adopted technique for classification tasks in chemometrics. For both sets of experimental data, CLS‐DA and PLS‐DA showed comparable efficiency. For the classification of three types of wine, the accuracy of the proposed method was 94.3%, while the accuracy of the reference method was 98.1%. For the classification of minced meat samples, the accuracies of CLS‐DA and PLS‐DA were 97.2% and 94%, respectively for all three groups. The findings demonstrate the potential of CLS‐DA as a straightforward and interpretable supervised pattern recognition technique, exhibiting comparable classification performance to PLS‐DA. The study highlights the advantages of CLS‐DA, including its ability to operate within the original data space and its flexibility in accommodating partial knowledge scenarios. The proposed CLS‐DA approach presents a promising alternative for discriminant analysis, offering new perspectives on the applications of classical least squares modeling in chemometrics.

Rapid identification of white tea based on colorimetric indicator displacement assay (IDA) sensor array

The identification of white tea mainly relies on sensory evaluation and mass spectrometry methods, which have drawbacks such as strong subjectivity and complex analysis. Thus, this study established a novel approach based on colorimetric indicator displacement assay (IDA) sensor to achieve rapid and accurate white tea identification. Cost-effective dyes and phenylboronic acids with different substituents were selected as the indicator and receptor, respectively. A color-signal-responsive IDA sensor array was constructed to enable the rapid and sensitive identification of white tea. Additionally, three supervised pattern recognition algorithms were utilized to process image data. The results demonstrate that combining the IDA sensor with the random forest algorithm accurately differentiates white tea samples by origin and type, achieving a 100% recognition accuracy in both training set (n = 84) and prediction set (n = 56) when ntree = 1000. This work offers an economical and efficient method for the rapid identification of white tea products.

Evaluating the cost of classifier discrimination choices for IoT sensor attack detection

The intrusion detection of IoT devices through the classification of malicious traffic packets have become more complex and resource intensive as algorithm design and the scope of the problems have changed. In this research, we compare the cost of a traditional supervised pattern recognition algorithm (k-Nearest Neighbor (KNN)), with the cost of a current deep learning (DL) unsupervised algorithm (Convolutional Neural Network (CNN)) in their simplest forms. The classifier costs are calculated based on the attributes of design, computation, scope, training, use, and retirement. We find that the DL algorithm is applicable to a wider range of problem-solving tasks, but it costs more to implement and operate than a traditional classifier. This research proposes an economic classifier model for deploying suitable AI-based intrusion detection classifiers in IoT environments. The model was empirically validated on the IoT-23 dataset using KNN and CNN. This study closes a gap in prior research that mostly concentrated on technical elements by incorporating economic factors into the evaluation of AI algorithms for IoT intrusion detection. This research thus evaluated the economic implications of deploying AI-based intrusion detection systems in IoT environments, considering performance metrics, implementation costs, and the cost of classifier discrimination choices. Researchers and practitioners should focus on the cost–benefit trade-offs of any artificial intelligence application for intrusion detection, recommending an economic evaluation and task fit assessment before adopting automated solutions or classifiers for IoT intrusion detection, particularly in large-scale industrial settings that involve active attacks.

Optimization of electronic nose measurements and discrimination of apple juices in combination with supervised pattern recognition

Optimization of electronic nose measurements and discrimination of apple juices in combination with supervised pattern recognition

Identification of Euphorbiae pekinensis Radix and its counterfeit and adulterated products based on DNA barcode, UPLC-Q-TOF-MS, UPLC fingerprint, and chemometrics.

Euphorbiae pekinensis Radix (EPR) is a traditional Chinese herb commonly used to treat edema, pleural effusion, and ascites. However, counterfeit and adulterated products often appear in the market because of the homonym phenomenon, similar appearance, and artificial forgery of Chinese herbs. This study comprehensively evaluated the quality of EPR using multiple methods. The DNA barcode technique was used to identify EPR, while the UPLC-Q-TOF-MS technique was utilized to analyze the chemical composition of EPR. A total of 15 tannin and phenolic acid components were identified. Furthermore, UPLC fingerprints of EPR and its common counterfeit products were established, and unsupervised and supervised pattern recognition models were developed using these fingerprints. The backpropagation artificial neural network and counter-propagation artificial neural network models accurately identified counterfeit and adulterated products, with a counterfeit ratio of more than 25%. Finally, the contents of the chemical markers 3,3'-di-O-methyl ellagic acid-4'-O-β-D-glucopyranoside, ellagic acid, 3,3'-di-O-methyl ellagic acid-4'-O-β-d-xylopyranoside, and 3,3'-di-O-methyl ellagic acid were determined to range from 0.05% to 0.11%, 1.95% to 8.52%, 0.27% to 0.86%, and 0.10% to 0.42%, respectively. This proposed strategy offers a general procedure for identifying Chinese herbs and distinguishing between counterfeit and adulterated products.

Feature analysis and recognition of fiber breakage AE signals after propagation

Acoustic emission (AE) is a non-destructive testing technique, and establishing correlations between AE signals and material damage modes is one of its primary challenges. However, it is difficult to identify damage modes in ceramic matrix composites (CMCs) due to AE signal attenuation occurring after propagation and complex damage modes. In this study, AE signals generated by the breakage of C and SiC fibers were monitored at different distances and angles on the C/SiC plate. The attenuation of energy and the frequency spectra were analyzed. The Mel-frequency cepstral coefficient (MFCC) method was used to analyze the waveform data of AE signals and extract MFCC features. To identify the damage caused by C and SiC fiber breakage, AE parameter features and MFCC features were selected as inputs, and a fully connected neural network was constructed to train a supervised pattern recognition model. The results show that the MFCC feature has higher recognition accuracy than the traditional feature when AE is used for damage identification.

The combination of Fourier-transform infrared spectroscopy with pattern recognition techniques for classification and discrimination of red snapper fish oils.

Fish oils are good sources for essential fatty acids such as omega-3 and omega-6 fatty acids needed to human growth. Indonesia is rich in fish species and among this, red snapper fish (Lutjanus sp.) can be extracted to get red snapper fish oils (RSFOs). The aim of this study was to classify and discriminate RSFO from different origins using Fourier-transform infrared (FTIR) spectra and pattern recognition techniques. All of the RSFO's FTIR spectra were very similar. The FTIR vibrations showed the presence of triglycerides as the main composition in fish oils. Principal component analysis (PCA) could separate the RSFO according to sample origin. Supervised pattern recognition of partial least square-discriminant analysis (PLS-DA) and sparse PLS-DA (sPLS-DA) successfully discriminated and classified different Lutjanus species of fish oils obtained from different origins. The vibration of functional groups at 1711, 1653, 1745, and 3012 per cm were considered for their important contributions in discriminating of Lutjanus species (variable importance in projection, variable importance in the projection score >1). Fish oils obtained from the same species were classified into the same class indicating similar chemical compositions. Among the three pattern recognition techniques used, sPLS-DA offers the best model for the discrimination and classification of Lutjanus fish oils. It can be concluded that FTIR spectroscopy in combination with the pattern recognition technique is the potential to be used for of fish oil authentication to verify the quality of the fish oils. It can be further developed as a rapid and effective method for fish oil authentication.

Discrimination/Classification of Edible Vegetable Oils from Raman Spatially Solved Fingerprints Obtained on Portable Instrumentation.

Currently, the combination of fingerprinting methodology and environmentally friendly and economical analytical instrumentation is becoming increasingly relevant in the food sector. In this study, a highly versatile portable analyser based on Spatially Offset Raman Spectroscopy (SORS) obtained fingerprints of edible vegetable oils (sunflower and olive oils), and the capability of such fingerprints (obtained quickly, reliably and without any sample treatment) to discriminate/classify the analysed samples was evaluated. After data treatment, not only unsupervised pattern recognition techniques (as HCA and PCA), but also supervised pattern recognition techniques (such as SVM, kNN and SIMCA), showed that the main effect on discrimination/classification was associated with those regions of the Raman fingerprint related to free fatty acid content, especially oleic and linoleic acid. These facts allowed the discernment of the original raw material used in the oil's production. In all the models established, reliable qualimetric parameters were obtained.

Forensic Characterization of Mobile Phone Cases Using Infrared Spectroscopy with Deep Learning Methods

The fast and nondestructive identification of mobile phone cases is a crucial task for law enforcement personnel in the field of forensic science. Attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR) has been employed in conjunction with feature extraction and deep learning algorithms to develop a method for distinguishing mobile phone cases. As preprocessing steps, multivariate scatter correction and standard normal variate were performed. Feature variables were extracted using the competitive adaptive reweighted sampling and shuffled frog leaping algorithm. Several supervised pattern recognition methods, namely ResNet50, VGG16, and multilayer perceptron (MLP), were utilized for constructing classifiers. The results indicated that the model’s performance based on multivariate scattering correction (MSC) with competitive adaptive re-weighted sampling (CARS) surpassed that of other approaches. Notably, the ResNet50 model exhibited superior discriminative ability compared to the VGG16 and MLP models, achieving 100% accuracy in the differentiation of all samples. This designed method represents a potentially rapid and nondestructive approach for identifying mobile phone cases in forensic science.

A deep neural network with electronic nose for water stress prediction in Khasi Mandarin Orange plants

Water stress is a significant environmental factor that hampers plant productivity and leads to various physiological and biological changes in plants. These include modifications in stomatal conductance and distribution, alteration of leaf water potential & turgor loss, altered chlorophyll content, and reduced cell expansion and growth. Additionally, water stress induces changes in the emission of volatile organic compounds across different parts of the plants. This study presents the development of an electronic nose (E-nose) system integrated with a deep neural network (DNN) to detect the presence and levels of water stress induced in Khasi Mandarin Orange plants. The proposed approach offers an alternative to conventional analytical methods that demand expensive and complex laboratory facilities. The investigation employs the leaf relative water content (RWC) estimation, a conventional technique, to evaluate water stress induction in the leaves of 20 plants collected over a span of 9 days after stopping irrigation. Supervised pattern recognition algorithms are trained using the results of RWC measurement, categorising leaves into non-stressed or one of four stress levels based on their water content. The dataset used for training and optimising the DNN model consists of 27 940 samples. The performance of the DNN model is compared to traditional machine learning methods, including linear and radial basis function support vector machines, k-nearest neighbours, decision tree, and random forest. From the results, it is seen that the optimised DNN model achieves the highest accuracy of 97.59% in comparison to other methods. Furthermore, the model is validated on an unseen dataset, exhibiting an accuracy of 97.32%. The proposed model holds the potential to enhance agricultural practices by enabling the detection and classification of water stress in crops, thereby aiding in water management improvements and increased productivity.

The classification performance of multivariate curve resolution-discriminant analysis: A comparative study

Pattern recognition models establish the relationship between the selected variables and objects' belonging to different classes. Variations in measured variables, different object patterns, and relationship models explain the need to develop different algorithms. Partial least squares-discriminant analysis (PLS-DA) is based on the projection of object responses in the space of latent vectors of the PLS calibration model. Recently, Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) method has been introduced for discriminant analysis by our group. The possibility of incorporation of real physical and chemical information in the MCR decomposition of data leads to solutions that directly have meaning and simplifies the interpretation of results. Considering the new application of MCR in supervised pattern recognition, there is still a need to evaluate and compare the efficiency of MCR-DA with tested methods such as PLS-DA.Detection and discrimination of extra virgin olive oil from four other classes of pure edible oils and also their mixture with extra virgin olive oil, as adulteration was chosen just as a typical system to evaluate and compare the efficiency of MCR-DA and PLS-DA methods. Raman spectroscopy was used forresponse measurement of oil samples. Different classifications were made between extra virgin olive oil and diverse combinations of four other edible oils, and similar sets were used for the comparative evaluation of the two discrimination methods. Various criteria evaluated the performance of the two methods and in most cases the new MCR-DA method was comparable to PLS-DA. However, MCR-DA can assign pure Raman spectra resulting from decomposition to different classes, which can provide direct information about the components of the studied system.

Characterization of Animal and Vegetable Oil by Attenuated Total Reflectance – Fourier Transform Infrared (ATR-FTIR) Spectroscopy with Supervised Pattern Recognition and Filter Algorithm

The potential of filter algorithms in improving spectral model performance was evaluated. A total of 329 animal and vegetable oil samples were used to collect attenuated total reflectance – Fourier transform infrared (ATR-FTIR) spectra. Fisher discriminant analysis (FDA), support vector machine (SVM), decision tree (DT), K-nearest neighbor analysis (KNN), and multilayer perceptron neural network (MLP) were considered to build models. Three filter algorithms, finite length unit impulse response filter (FIR), infinite length impulse response filter (IIR) and wavelet transform (WT), were evaluated to enhance the performance of the models. The Morlet basis function was the most suitable wavelet transform, accurately classifying 90.3% of the training set and 94.4% of the test set. The MLP algorithms were demonstrated to be superior to the others. The best performance was obtained using low-pass or band-stop filters that provided 100% accuracy with the MLP model. The reported method is demonstrated to be affordable and easy-to-use in forensic analysis.

Discriminating the viscoelastic properties of cellulose textile fibers for recycling

The viscoelastic properties of cellulose fibers play an important role in chemical recycling of textiles. Here we discriminated the intrinsic viscosity of cotton roll towels and bed linens using near-infrared imaging spectroscopy and supervised pattern recognition. The classification results showed training and test set accuracies of 84–97% and indicated that the relevant spectral features were related to water, cellulose, and cellulose crystallinity. We hypothesized that the decreasing intrinsic viscosity of cotton was associated with changes in cellulose crystallinity and water adsorption, which was supported by additional X-ray and sorption measurements. These results are important as they indicate the potential to non-invasively estimate the degree of polymerization and the suitability of different cotton materials for chemical recycling. We propose that changes in the degree of polymerization and cellulose crystallinity could be used as an indicator of the chemical quality of cellulose fibers, which would have wider impacts for textile recycling.

Pattern Recognition of GC-FID Profiles of Volatile Compounds in Brandy de Jerez Using a Chemometric Approach Based on Their Instrumental Fingerprints

Brandy de Jerez is a unique spirit produced in Southern Spain under Protected Geographical Indication “Brandy de Jerez” (PGI). Two key factors for the production of quality brandies are the original wine spirit and its aging process. They are significantly conditioned by specific variables related to the base wine and the distillation method employed to produce the wine spirit used to obtain a finally aged brandy. This final beverage is therefore strongly influenced by its production process. The chromatographic instrumental fingerprints (obtained by GC FID) of the major volatile fraction of a series of brandies have been examined by applying a chemometric approach based on unsupervised (hierarchical cluster analysis and principal component analysis) and supervised pattern recognition tools (partial least squares–discriminant analysis and support vector machine). This approach was able to identify the fermentation conditions of the original wine, the distillation method used to produce the wine spirit, and the aging process as the most influential factors on the volatile profile.

Aroma modeling and quality evaluation of spearmint (Mentha spicata subsp. spicata) using electronic nose technology coupled with artificial intelligence algorithms

The present study aimed to evaluate the capability of an electronic nose system coupled with chemometrics tools as a fast, real-time, and non-destructive technology for spearmint grading. The developed e-nose system was utilized to measure 46 samples' volatile organic components (VOCs) procured from different geographical origins. PCA was performed for original data clustering and dimensionality reduction of the aroma data. PLS-DA and fuzzy clustering were performed for differentiating the samples. Fuzzy clustering presented a more reasonable performance due to the overlapping nature of the flowering and vegetative stages clusters. An artificial neural network model was utilized as a supervised pattern recognition algorithm to map the aroma data of the samples to their quality grade based on their essential oils percentage. The developed system predicted the sample’s grade with high correlation coefficients of prediction values (Rp2 = 0.97 and RMSEp= 0.075) and acceptable sensitivity (97.7%) and specificity (86.4%).

Authentication and discrimination of tissue origin of bovine gelatin using combined supervised pattern recognition strategies

Gelatin is a protein-based product derived from bovine and porcine bone and skin. Gelatin is increasingly used in food and non-food products, but there are errors in labeling concerning the origin of commercial gelatin raw materials. These mislabeling may negatively impact religious or safety communities. In gelatin studies, species identification is common, but determining the origin of the tissues, bone and skin, is also crucial. The current study demonstrated the use of Raman spectroscopy as a rapid and economical method coupled with chemometrics to differentiate tissue origins of bovine gelatin. The proposed technique could also detect adulteration of bone origin gelatin with skin gelatin with 100% sensitivity and specificity. Accordingly, Raman spectroscopy, along with DD- SIMCA and PLS-DA methods, provides a reliable technique that can easily be applied to detecting the tissue source of bovine gelatin. Also, two- step classification method with goal of finding a suitable authentication of target classes have been applied. dd-SIMCA as class modeling for authentication and PLS-DA for discrimination of groups have been combined in this strategy.

Influence of the use of sulfur dioxide, the distillation method, the oak wood type and the aging time on the production of brandies

Brandies are spirits produced from wine spirit and wine distillates. The original wines selected to be distilled to produce the wine spirits as well as the distillation method used determine, to a large extent, the organoleptic characteristics of the final products. The young wine spirits evolve during their aging in oak casks, this being another key stage that affects the chemical and sensorial characteristics of the final brandy. In this work, seven different brandies have been studied. They were obtained from wine produced with and without the addition of sulfur dioxide, during their fermentation, using different distillation methods (single, double or serial distillation using pot stills and continuous column distillation) and aged for 14 or 28 months in three different types of oak wood (Quercus alba, Quercus robur and Quercus petraea) previously toasted to two different grades (medium or light).The use of unsupervised pattern recognition methods (HCA and FA) determined that the addition of sulfur dioxide during the fermentation of the base wine has a major influence on the aromatic and phenolic profile of the aged distillates. On the other hand, by means of supervised pattern recognition methods such as LDA and ANNs, the most significant variables that would allow to discriminate between the classes of brandies identified in the study were evaluated. Thus, the results obtained should cast some light on the most significant variables to be taken into account regarding Brandy production processes if a better control over these production processes is to be achieved, so that more exclusive and better quality products are obtained.

Application of linear and nonlinear pattern recognition techniques for discrimination of fluoroquinolones using modified AgNPs-based colorimetric sensor array

Surface plasmon resonance-based colorimetric sensor arrays made up of metallic nanoparticles (NPs) are powerful tools that are mostly used for pattern recognition and quantification purposes, especially in the case of pharmaceutical and biologically active substances. Ease of fabrication, simplicity, cost-effectiveness, and huge amounts of collectible data besides the various fabrication techniques, expand their applications in analytical chemistry, especially when combined with chemometrics data analysis tools. In this study, a colorimetric sensor array is fabricated using unmodified citrate-capped together with the modified AgNPs for the discrimination of three fluoroquinolone drugs ofloxacin, ciprofloxacin, and moxifloxacin in real serum samples. The aggregation-induced color changes of the AgNPs were monitored during the time and the collected kinetic-spectrophotometric huge data in the ranges 300–800 nm was compressed by Discrete Wavelet Transform (DWT) to remove the redundant variables and analyzed with different unsupervised and supervised pattern recognition techniques like Principal Components Analysis (PCA), Linear Discriminant Analysis (LDA) and XY-Fused neural networks (XYF). The best classification performance was achieved using the genetic algorithm-optimized XY-Fused networks with 92% accuracy and an error of cross-validation equal to zero. In addition, all of the spiked real serum samples were classified correctly using the generated model according to the assigned class labels.

Stable isotope and elemental profiles determine geographical origin of saffron from China and Iran

Origin verification of high-value saffron is essential for fair trade and to protect consumers' interests and rights. A traceability method using elemental content (% C and % N) and stable isotopes (δ13C, δ2H, δ18O, and δ15N) combined with chemometrics was developed to discriminate saffron from Iran and China and classify major domestic production areas in China. Results showed that Iranian samples had lower % C and % N contents but higher δ13C values than Chinese origin saffron, with δ13C acting as an important variable for origin discrimination. Moreover, δ2H and δ13C isotopes were found to be important variables to classify Chinese regional saffron origin. Two supervised pattern recognition models (PLS-DA) developed to classify Iranian and Chinese saffron, and regional Chinese saffron had a discrimination accuracy of 85.0 % and 80.2 %, respectively. These models provide the basis for a new regulatory inspection procedure to verify saffron origin and label claims, minimizing fraudulent mislabeling and adding value to saffron from specific regions.

Forensic Identification of 3,4-Methylendioxy-N-methylamphetamine (MDMA), Ketamine, and Benzodiazepine by Handheld Infrared Spectroscopy and Chemometrics

The development of simple and cost-effective analytical methods is indispensable for law enforcement officials in the rapid screening for large quantities of seized drugs. Handheld infrared spectroscopy has been used with Bayes discrimination analysis (BDA), k nearest neighbor (KNN), support vector machine (SVM), decision-making tree (DT), and random forest (RF) chemometrics to characterize 3,4-methylenedioxy-N-methylamphetamine (MDMA), ketamine, and benzodiazepine samples. Spectral data were pretreated by normalization, multivariate scatter correction, Savitzky-Golay smoothing algorithm, and automatic baseline correction. Principal component analysis (PCA) was employed to reduce the dimensionality of spectral data, while BDA, KNN, SVM, DT, and RF, as supervised pattern recognition methods, were implemented and their parameters were optimized to identify the samples. The results showed that SVM was optimal among the models, followed by DT, KNN, BDA, and RF. The optimal model, based upon polynomial kernel function, achieved an accuracy of 100% (training set) and 100% (test set) for all samples and was used to further distinguish the additives in MDMA, ketamine, and benzodiazepine samples. In the MDMA samples, 42 samples contained caffeine, 23 contained glucose, and 21 contained starch. In the ketamine samples, 34 contained caffeine, 13 contained glucose, and 12 contained phenacetin. In the benzodiazepine samples, estazolam, diazepam, lorazepam, clonazepam, and nitrazepam were identified. 5 samples were misidentified into other categories. This study demonstrated that handheld infrared spectroscopy with chemometrics is a reliable and desirable approach for the rapid and nondestructive identification of MDMA, ketamine, and benzodiazepine with high potential in practical applications.