Showed Classification Accuracy Research Articles

Attenuated Total Reflectance–Fourier Transform Infrared (ATR-FTIR) Spectroscopy Combined With Deep Learning for Classification of Idiopathic Recurrent Spontaneous Miscarriage (IRSM)

Idiopathic recurrent spontaneous miscarriage (IRSM) is characterized by the repeated losses of two or more clinically detected pregnancies before 24 week of gestation with the cause remaining unknown. We aim to ascertain whether attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy with deep learning (DL) can accurately distinguish IRSM from controls. Endometrial tissue (internal lining of the uterine cavity) was collected from 20 women with IRSM and 20 controls during window of implantation. ATR-FTIR spectroscopy was used to acquire spectra (400–4000 cm−1) from three random locations across each specimen and averaged. Multivariate analysis was used to assess group differentiation. Furthermore, DL classifiers were implemented to predict IRSM. Hierarchical cluster analysis (HCA) demonstrated distinct clustering of the two groups. The principal component analysis (PCA) loading plot indicated contributions from amide I, II and III bands, symmetric methyl (CH3) bending of proteins, lipids, nucleic acids, asymmetric stretching of CH2 due to lipids, collagen, and glucose in discriminating between IRSM and controls. Supervised orthogonal projections to latent structures discriminant analysis (OPLS-DA) exhibited optimized discrimination between the groups. Convolutional neural network (CNN) and variational autoencoder-artificial neural network (VAE-ANN) classifiers exhibited good performance and showed classification accuracy of 94% and 94% and F1 scores (mean of recall and precision) of 95% and 93%, respectively. The optimal area under the receiver operating characteristic curve (AUC) value of both CNN and VAE-ANN was 0.94. ATR-FTIR spectroscopy combined with DL approach is promising for understanding complex disease pathophysiology and enabling early prediction.

Chlorophyll Fluorescence Imaging for Environmental Stress Diagnosis in Crops.

The field of plant phenotype is used to analyze the shape and physiological characteristics of crops in multiple dimensions. Imaging, using non-destructive optical characteristics of plants, analyzes growth characteristics through spectral data. Among these, fluorescence imaging technology is a method of evaluating the physiological characteristics of crops by inducing plant excitation using a specific light source. Through this, we investigate how fluorescence imaging responds sensitively to environmental stress in garlic and can provide important information on future stress management. In this study, near UV LED (405 nm) was used to induce the fluorescence phenomenon of garlic, and fluorescence images were obtained to classify and evaluate crops exposed to abiotic environmental stress. Physiological characteristics related to environmental stress were developed from fluorescence sample images using the Chlorophyll ratio method, and classification performance was evaluated by developing a classification model based on partial least squares discrimination analysis from the image spectrum for stress identification. The environmental stress classification performance identified from the Chlorophyll ratio was 14.9% in F673/F717, 25.6% in F685/F730, and 0.209% in F690/F735. The spectrum-developed PLS-DA showed classification accuracy of 39.6%, 56.2% and 70.7% in Smoothing, MSV, and SNV, respectively. Spectrum pretreatment-based PLS-DA showed higher discrimination performance than the existing image-based Chlorophyll ratio.

Multi-class deep learning architecture for classifying lung diseases from chest X-Ray and CT images

Medical imaging is considered a suitable alternative testing method for the detection of lung diseases. Many researchers have been working to develop various detection methods that have aided in the prevention of lung diseases. To better understand the condition of the lung disease infection, chest X-Ray and CT scans are utilized to check the disease’s spread throughout the lungs. This study proposes an automated system for the detection multi lung diseases in X-Ray and CT scans. A customized convolutional neural network (CNN) and two pre-trained deep learning models with a new image enhancement model are proposed for image classification. The proposed lung disease detection comprises two main steps: pre-processing, and deep learning classification. The new image enhancement algorithm is developed in the pre-processing step using k-symbol Lerch transcendent functions model which enhancement images based on image pixel probability. While, in the classification step, the customized CNN architecture and two pre-trained CNN models Alex Net, and VGG16Net are developed. The proposed approach was tested on publicly available image datasets (CT, and X-Ray image dataset), and the results showed classification accuracy, sensitivity, and specificity of 98.60%, 98.40%, and 98.50% for the X-Ray image dataset, respectively, and 98.80%, 98.50%, 98.40% for the CT scans dataset, respectively. Overall, the obtained results highlight the advantages of the image enhancement model as a first step in processing.

Machine Learning Methods for Predicting League of Legends Game Outcome

The video game League of Legends has several professional leagues and tournaments that offer prizes reaching several million dollars, making it one of the most followed games in the Esports scene. This article addresses the prediction of the winning team in professional matches of the game, using only pre-game data. We propose to improve the accuracy of the models trained with the features offered by the game API. To this end, new features are built to collect interesting information, such as the skills of a player handling a certain champion, the synergies between players of the same team or the ability of a player to beat another player. Then, we perform feature selection and train different classification algorithms aiming at obtaining the best model. Experimental results show classification accuracy above 0.70, which is comparable to the results of other proposals presented in the literature, but with the added benefit of using few samples and not requiring the use of external sources to collect additional statistics.

Assessment of the effectiveness of a convolutional autoencoder for digital image-based automated core logging

Widely applicable convolutional neural network (CNN)-based lithology classification models are limited to interpret soundness of a trained model and require high computational cost. This study proposes a convolutional autoencoder (CAE)-based classification model to improve the efficiency of rock-core-logging analysis based on digital rock-core images (CAE model I) and suggest a more understandable classification model by extracting the lithological feature information from the network structure of the model. Moreover, input digital-image data transformation using the Canny edge-detection algorithm was implemented (CAE model II). The applicability of the developed models is validated by comparing them with a CNN model and RGBE model. While implementing the proposed model, digital images of 85-m-long cores were used, which were acquired from the Satyr 5 well in the Northern Carnarvon Basin, Western Australia. It was found that CAE models I and II showed classification accuracy comparable to the CNN model (CAE model I: 96.4%, CAE model II: 94.9%, CNN model: 97.5%, and RGBE model: 94.4%). The extracted features of the trained CAE models are effective in interpreting the trained network structure. In addition, in terms of computational efficiency and training time, the CAE-based models are significantly better than the CNN model (CAE model I: 7 min 21 s, CAE model II: 7 min 34 s, CNN model: 12 min 6 s, and RGBE model: 24 min 35 s). Overall, the results demonstrate the feasibility of the proposed CAE-based models for automated rock-core logging.

A two-stage hybrid biomarker selection method based on ensemble filter and binary differential evolution incorporating binary African vultures optimization

BackgroundIn the field of genomics and personalized medicine, it is a key issue to find biomarkers directly related to the diagnosis of specific diseases from high-throughput gene microarray data. Feature selection technology can discover biomarkers with disease classification information.ResultsWe use support vector machines as classifiers and use the five-fold cross-validation average classification accuracy, recall, precision and F1 score as evaluation metrics to evaluate the identified biomarkers. Experimental results show classification accuracy above 0.93, recall above 0.92, precision above 0.91, and F1 score above 0.94 on eight microarray datasets.MethodThis paper proposes a two-stage hybrid biomarker selection method based on ensemble filter and binary differential evolution incorporating binary African vultures optimization (EF-BDBA), which can effectively reduce the dimension of microarray data and obtain optimal biomarkers. In the first stage, we propose an ensemble filter feature selection method. The method combines an improved fast correlation-based filter algorithm with Fisher score. obviously redundant and irrelevant features can be filtered out to initially reduce the dimensionality of the microarray data. In the second stage, the optimal feature subset is selected using an improved binary differential evolution incorporating an improved binary African vultures optimization algorithm. The African vultures optimization algorithm has excellent global optimization ability. It has not been systematically applied to feature selection problems, especially for gene microarray data. We combine it with a differential evolution algorithm to improve population diversity.ConclusionCompared with traditional feature selection methods and advanced hybrid methods, the proposed method achieves higher classification accuracy and identifies excellent biomarkers while retaining fewer features. The experimental results demonstrate the effectiveness and advancement of our proposed algorithmic model.

An application of raman spectroscopy in combination with machine learning to determine gastric cancer spectroscopy marker

Background and ObjectiveGlobally, gastric carcinoma (Gca) ranks fifth in terms of incidence and third in terms of mortality. Higher serum tumor markers (TMs) than those from healthy individuals, led to TMs clinical application as diagnostic biomarkers for Gca. Actually, there is no accurate blood test to diagnose Gca. MethodsRaman spectroscopy is applied as an efficient, credible, minimally invasive technique to evaluate the serum TMs levels in blood samples. After curative gastrectomy, serum TMs levels are important in predicting the recurrence of gastric cancer, which must be detected early. The experimentally assesed TMs levels using Raman measurements and ELİSA test were used to develop a prediction model based on machine learning techniques. A total of 70 participants diagnosed with gastric cancer after surgery (n = 26) and healthy (n = 44) were comrpised in this study. ResultsIn the Raman spectra of gastric cancer patients, an additional peak at 1182 cm−1 was observed and, the Raman intensity of amide III, II, I, and CH2 proteins as well as lipids functional group was higher. Furthermore, Principal Component Analysis (PCA) showed, that it is possible to distinguish between the control and Gca groups using the Raman range between 800 and 1800 cm−1, as well as between 2700 and 3000 cm−1. The analysis of Raman spectra dynamics in gastric cancer and healthy patients showed, that the vibrations at 1302 and 1306 cm−1 were characteristic for cancer patients. In addition, the selected machine learning methods showed classification accuracy of more than 95%, while obtaining an AUROC of 0.98. Such results were obtained using Deep Neural Networks and the XGBoost algorithm. ConclusionsThe obtained results suggest, that Raman shifts at 1302 and 1306 cm−1 could be spectroscopic markers of gastric cancer.

Heterogeneous recurrent spiking neural network for spatio-temporal classification.

Spiking Neural Networks are often touted as brain-inspired learning models for the third wave of Artificial Intelligence. Although recent SNNs trained with supervised backpropagation show classification accuracy comparable to deep networks, the performance of unsupervised learning-based SNNs remains much lower. This paper presents a heterogeneous recurrent spiking neural network (HRSNN) with unsupervised learning for spatio-temporal classification of video activity recognition tasks on RGB (KTH, UCF11, UCF101) and event-based datasets (DVS128 Gesture). We observed an accuracy of 94.32% for the KTH dataset, 79.58% and 77.53% for the UCF11 and UCF101 datasets, respectively, and an accuracy of 96.54% on the event-based DVS Gesture dataset using the novel unsupervised HRSNN model. The key novelty of the HRSNN is that the recurrent layer in HRSNN consists of heterogeneous neurons with varying firing/relaxation dynamics, and they are trained via heterogeneous spike-time-dependent-plasticity (STDP) with varying learning dynamics for each synapse. We show that this novel combination of heterogeneity in architecture and learning method outperforms current homogeneous spiking neural networks. We further show that HRSNN can achieve similar performance to state-of-the-art backpropagation trained supervised SNN, but with less computation (fewer neurons and sparse connection) and less training data.

Automatic Voice Disorder Detection Using Self-Supervised Representations

Many speech features and models, including Deep Neural Networks (DNN), are used for classification tasks between healthy and pathological speech with the Saarbruecken Voice Database (SVD). However, accuracy values of 80.71% for phrases or 82.8% for vowels /aiu/ are the highest reported for audio samples in SVD when the evaluation includes the wide amount of pathologies in the database, instead of a selection of some pathologies. This paper targets this top performance in the state-of-the-art Automatic Voice Disorder Detection (AVDD) systems. In the framework of a DNN-based AVDD system we study the capability of Self-Supervised (SS) representation learning for describing discriminative cues between healthy and pathological speech. The system processes the SS temporal sequence of features with a single feed-forward layer and Class-Token (CT) Transformer for obtaining the classification between healthy and pathological speech. Furthermore, there is evaluated a suitable data extension of the training set with out-of-domain data is also evaluated to deal with the low availability of data for using DNN-based models in voice pathology detection. Experimental results using audio samples corresponding to phrases in the SVD dataset, including all pathologies available, show classification accuracy values until 93.36%. This means that the proposed AVDD system achieved accuracy improvements of 4.1% without the training data extension, and 15.62% after the training data extension compared to the baseline system. Beyond the novelty of using SS representations for AVDD, the fact of obtaining accuracies over 90% in these conditions and using the whole set of pathologies in the SVD is a milestone for voice disorder-related research. Furthermore, the study on the amount of in-domain data in the training set related to the system performance show guidance for the data preparation stage. Lessons learned in this work suggest guidelines for taking advantage of DNN, to boost the performance in developing automatic systems for diagnosis, treatment, and monitoring of voice pathologies.

Near-infrared hyperspectral imaging for online measurement of the viability detection of naturally aged watermelon seeds.

The viability status of seeds before sowing is important to farmers as it allows them to make yield predictions. Monitoring the seed quality in a rapid and nondestructive manner may create a perfect solution, especially for industrial sorting applications. However, current offline laboratory-based strategies employed for the monitoring of seed viability are time-consuming and thus cannot satisfy industrial needs where there is a substantial number of seeds to be analyzed. In this study, we describe a prototype online near-infrared (NIR) hyperspectral imaging system that can be used for the rapid detection of seed viability. A wavelength range of 900-1700 nm was employed to obtain spectral images of three different varieties of naturally aged watermelon seed samples. The partial least square discriminant analysis (PLS-DA) model was employed for real-time viability prediction for seed samples moving through a conveyor unit at a speed of 49 mm/sec. A suction unit was further incorporated to develop the online system and it was programmatically controlled to separate the detected viable seeds from nonviable ones. For an external validation sample set showed classification accuracy levels of 91.8%, 80.7%, and 77.8% in relation to viability for the three varieties of watermelon seed with healthy seedling growth. The regression coefficients of the classification model distinguished some chemical differences in viable and nonviable seed which was verified by the chromatographic analysis after the detection of the proposed online system. The results demonstrated that the developed online system with the viability prediction model has the potential to be used in the seed industry for the quality monitoring of seeds.

Motor imagery signal classification using Wavelet packet decomposition and modified binary grey wolf optimization

Nowadays, Electroencephalogram (EEG) signals are widely used in brain-computer interfaces (BCIs), including the identification of motor imagery (MI) activities and prostheses. In this proposed scheme, wavelet packet decomposition (WPD) is used with approximate entropy (ApEn) as a mechanism for generating features of Electroencephalogram (EEG) signals for the MI-BCI system. Usually, features extracted from an EEG signal include redundant and irrelevant features. This paper proposes a novel scheme of feature selection using Modified Binary grey wolf optimization (MBGWO) for MI-BCI with EEG. Finally, performance of the proposed approach is evaluated using machine learning technique K-nearest Neighbors (KNN). In this study, KNN showed classification accuracy for subjects “a" and “b" is 92.86% and 91.53% respectively. Where “a" and “b" are two participants in Dataset I from BCI competition IV. The results show that the suggested algorithm considerably impacts classification performance. Classification results suggest that the proposed method used in this study can achieve better performance in BCI applications for understanding MI activities in patients with physical disabilities.

Multispectral image analysis for monitoring by IoT based wireless communication using secure locations protocol and classification by deep learning techniques

Land surface data is one of the entry points for a enormous resource management and human activity such as to monitor farming, ecology, urban management, and territories. This research aimed to improve the performance of the multispectral image processing technique using a novel deep learning based classification approaches used in classifying the surfaces in multispectral satellite images. The monitoring of Internet of Things (IoT) based wireless communication using secure locations sensor network protocol (SLSNP). The multispectral satellite images are collected based on developing and developed countries. Upon the collection of multispectral image data using deep convolution neural network (DCNN) deep learning system landmarks of the multispectral images are estimated. This will be effective for the identification of resources in a particular location. The simulation results show classification accuracy, precision, recall and the network lifetime, efficiency, throughput based on data transmission from satellite for secure location based wireless sensor networks (WSNs).

Novel cellular signatures for determining time since deposition for trace DNA evidence

With the increase in sensitivity of DNA profiling, questions about how and when the DNA was deposited have become a driving issue in forensic cases. To address this, we propose a novel method to determine time since deposition of trace DNA samples based on morphological and autofluorescence properties of individual epithelial cells which can change as the sample ages. To develop this signature, a series of trace DNA samples were generated by contact/handling a substrate and then allowed to age anywhere between one day and more than one year prior to collection. Imaging flow cytometry (IFC) was then used to characterize the morphology and autofluorescence profiles of individual cells within each sample followed by multivariate modelling and predictive classification. Resultsshowed that epithelial cell populations could be classified with high accuracy (∼90%) into one of three time-since-deposition groups: < 1 week, between 1 week and 2 months, and > 2months. Differences across age groups were largely driven by decreases in brightfield contrast and increases in the intensity of autofluorescence. To further test this approach for forensic casework, 47 individual donor cell populations spanning each time deposition group were classified blindly against the remaining data set. Samples containing at least 75 cells and a posterior probability greater than 0.90 showed classification accuracies ∼95%. Accuracies for individual time groups were 97% (<1 week), 92% (1week-2months), 98% (>2 months) with an average posterior probability for all time groups ∼0.96. This indicates that autofluorescence and morphological analyses may provide probative information regarding time since deposition for many types of trace DNA samples in forensic casework.

Gun identification from gunshot audios for secure public places using transformer learning

Increased mass shootings and terrorist activities severely impact society mentally and physically. Development of real-time and cost-effective automated weapon detection systems increases a sense of safety in public. Most of the previously proposed methods were vision-based. They visually analyze the presence of a gun in a camera frame. This research focuses on gun-type (rifle, handgun, none) detection based on the audio of its shot. Mel-frequency-based audio features have been used. We compared both convolution-based and fully self-attention-based (transformers) architectures. We found transformer architecture generalizes better on audio features. Experimental results using the proposed transformer methodology on audio clips of gunshots show classification accuracy of 93.87%, with training loss and validation loss of 0.2509 and 0.1991, respectively. Based on experiments, we are convinced that our model can effectively be used as both a standalone system and in association with visual gun-detection systems for better security.

Hybrid Deep Learning Model for Fake News Detection in Social Networks (Student Abstract)

The proliferation of fake news has grown into a global concern with adverse socio-political and economical impact. In recent years, machine learning has emerged as a promising approach to the automation of detecting and tracking fake news at scale. Current state of the art in the identification of fake news is generally focused on semantic analysis of the text, resulting in promising performance in automated detection of fake news. However, fake news campaigns are also evolving in response to such new technologies by mimicking semantic features of genuine news, which can significantly affect the performance of fake news classifiers trained on contextually limited features. In this work, we propose a novel hybrid deep learning model for fake news detection that augments the semantic characteristics of the news with features extracted from the structure of the dissemination network. To this end, we first extend the LIAR dataset by integrating sentiment and affective features to the data, and then use a BERT-based model to obtain a representation of the text. Moreover, we propose a novel approach for fake news detection based on Graph Attention Networks to leverage the user-centric features and graph features of news residing social network in addition to the features extracted in the previous steps. Experimental evaluation of our approach shows classification accuracy of 97% on the Politifact dataset. We also examined the generalizability of our proposed model on the BuzzFeed dataset, resulting in an accuracy 89.50%.

Multi-classification of arrhythmias using ResNet with CBAM on CWGAN-GP augmented ECG Gramian Angular Summation Field

Cardiovascular diseases are the leading cause of death globally. Arrhythmias are the most common symptoms and can cause sudden cardiac death. Accurate and reliable detection of arrhythmias from large amount of ECG signals remains a challenge. We here propose to use ResNet with convolutional block attention modules (CBAM-ResNet) to classify the major types of cardiac arrhythmias. To facilitate the classifier in extracting the rich information in the ECG signals, we transform the time series into Gramian angular summation field (GASF) images. In order to overcome the imbalanced data problem, we employ the conditional Wasserstein generative adversarial network with gradient penalty (CWGAN-GP) model to augment the minor categories. Tested using the MIT-BIH arrhythmia database, our method shows classification accuracy of 99.23%, average precision of 99.13%, sensitivity of 97.50%, specificity of 99.81% and the average F1 score of 98.29%. Compared with the performance of the state-of-the-art algorithms in the extant literature, our method is highest accuracy and specificity, comparable in precision, sensitivity and F1 score. These results suggest that transforming the ECG time series into GASF images is a valid approach to representing the rich ECG features for arrhythmia classification, and that CWGAN-GP based data augmentation provides effective solution to the imbalanced data problem and helps CBAM-ResNet to achieve excellent classification performance.

Machine Learning Based Healthcare System for Investigating the Association Between Depression and Quality of Life.

New technological innovations are changing the future of healthcare system. Identification of factors that are responsible for causing depression may lead to new experiments and treatments. Because depression as a disease is becoming a leading community health concern worldwide. Using machine learning techniques this article presents a complete methodological framework to process and explore the heterogenous data and to better understand the association between factors related to quality of life and depression. Subsequently, the experimental study is mainly divided into two parts. In the first part, a data consolidation process is presented. The relationship of data is formed and to uniquely identify each relation in data the concept of the Secure Hash Algorithm is adopted. Hashing is used to locate and index the actual items in the data. The second part proposed a model using both unsupervised and supervised machine learning techniques. The consolidation approach helped in providing a base for formulation and validation of the research hypothesis. The Self organizing map provided 08 cluster solution and the classification problems were taken from the clustered data to further validate the performance of the posterior probability multi-class Support Vector Machine. The expectations of the importance sampling resulted in factors responsible for causing depression. The proposed model was adopted to improve the classification performance, and the result showed classification accuracy of 91.16%.

Sentimental Analysis of Twitter Users from Turkish Content with Natural Language Processing.

Artificial Intelligence has guided technological progress in recent years; it has shown significant development with increased academic studies on Machine Learning and the high demand for this field in the sector. In addition to the advancement of technology day by day, the pandemic, which has become a part of our lives since early 2020, has led to social media occupying a larger place in the lives of individuals. Therefore, social media posts have become an excellent data source for the field of sentiment analysis. The main contribution of this study is based on the Natural Language Processing method, which is one of the machine learning topics in the literature. Sentiment analysis classification is a solid example for machine learning tasks that belongs to human-machine interaction. It is essential to make the computer understand people emotional situation with classifiers. There are a limited number of Turkish language studies in the literature. Turkish language has different types of linguistic features from English. Since Turkish is an agglutinative language, it is challenging to make sentiment analysis with that language. This paper aims to perform sentiment analysis of several machine learning algorithms on Turkish language datasets that are collected from Twitter. In this research, besides using public dataset that belongs to Beyaz (2021) to get more general results, another dataset is created to understand the impact of the pandemic on people and to learn about public opinions. Therefore, a custom dataset, namely, SentimentSet (Balli 2021), was created, consisting of Turkish tweets that were filtered with words such as pandemic and corona by manually marking as positive, negative, or neutral. Besides, SentimentSet could be used in future researches as benchmark dataset. Results show classification accuracy of not only up to ∼87% with test data from datasets of both datasets and trained models, but also up to ∼84% with small “Sample Test Data” generated by the same methods as SentimentSet dataset. These research results contributed to indicating Turkish language specific sentiment analysis that is dependent on language specifications.

A two-stage intrusion detection system with auto-encoder and LSTMs

‘Curse of dimensionality’ and the trade-off between low false alarm rate and high detection rate are the major concerns while designing an efficient intrusion detection system. In this study, we propose a hybrid framework comprising deep auto-encoder (AE) with the long short term memory (LSTM) and the bidirectional long short term memory (Bi-LSTM) for intrusion detection system by obtaining optimal features using AE and then LSTMs for classification into normal and anomaly samples. The performance of the proposed models is evaluated on the well-known dataset NSL-KDD in terms of error indices including precision, recall, F-score, accuracy, detection rate (DR), and false alarm rate (FAR). Experimental results indicate that the proposed AE-LSTM performance is significantly better with less prediction error as compared to other deep and shallow machine learning techniques including other recently reported methods. On the NSL-KDD dataset, AE-LSTM shows classification accuracy of 89% with DR of 89.84% and FAR of 11% which demonstrates the enhanced performance of the proposed model over recent state-of-the-art techniques.

Discrimination of electrical tapes using ATR-FTIR spectroscopy and chemometrics.

Electrical tapes are recovered during criminal investigations as physical evidence in cases of rape, kidnapping, and explosion incidents. The analysis of such evidence can provide an evidentiary link between the suspect, the victim, object, or the crime scene. In the present study, 25 brands of electrical tapes have been analyzed using ATR-FTIR (attenuated total reflectance Fourier transform infrared) spectroscopy. Samples (1cm2 ) were analyzed in the mid IR (Infrared) region from 4000-600cm-1 , and the functional groups of various components have been profiled. Chemometric methods-PCA (principal component analysis) and PCA-LDA (linear discriminant analysis) have been employed to interpret the data and classify the samples into its respective classes. Preliminary assessment of sample clustering due to similar chemical composition was visualized using PCA. PCA-LDA applied for classification purpose yielded classification accuracy (calibration) of 92.98% for the adhesive side and 88% for the backing side. The validation results showed classification accuracy of 89.47% for the adhesive side and 84% for the backing side. Blind validation study was carried out using 5 samples, and classification accuracy of 100% and 80% was obtained for the adhesive and the backing side, respectively. In the current study, a preliminary substrate study was carried out, and the results showed that the backing samples could be more accurately matched to their correct source of origin than the adhesive side.