Performance Of Classifier Research Articles

Detection and Classification of Obstructive Sleep Apnea Using Audio Spectrogram Analysis

Sleep disorders are steadily increasing in the population and can significantly affect daily life. Low-cost and noninvasive systems that can assist the diagnostic process will become increasingly widespread in the coming years. This work aims to investigate and compare the performance of machine learning-based classifiers for the identification of obstructive sleep apnea–hypopnea (OSAH) events, including apnea/non-apnea status classification, apnea–hypopnea index (AHI) prediction, and AHI severity classification. The dataset considered contains recordings from 192 patients. It is derived from a recently released dataset which contains, amongst others, audio signals recorded with an ambient microphone placed ∼1 m above the studied subjects and apnea/hypopnea accurate events annotations performed by specialized medical doctors. We employ mel spectrogram images extracted from the environmental audio signals as input of a machine-learning-based classifier for apnea/hypopnea events classification. The proposed approach involves a stacked model which utilizes a combination of a pretrained VGG-like audio classification (VGGish) network and a bidirectional long short-term memory (bi-LSTM) network. Performance analysis was conducted using a 5-fold cross-validation approach, leaving out patients used for training and validation of the models in the testing step. Comparative evaluations with recently presented methods from the literature demonstrate the advantages of the proposed approach. The proposed architecture can be considered a useful tool for supporting OSAHS diagnoses by means of low-cost devices such as smartphones.

ROBUSTNESS EVALUATION OF MACHINE LEARNING ALGORITHMS FOR NEUROCOMPUTER INTERFACE SOFTWARE USING DISTRIBUTED AND PARALLEL COMPUTING

This article highlights the field of information technology for brain-computer interaction, and the main goal is to use it to determine patterns of human brain activity using electroencephalography (EEG) data. During the execution of the article, machine learning methods were used, namely such classifiers as Random Forest, Multi-Layer Perceptron (MLP), and Logistic Regression. The investigation begins with real-world experiments recording EEG signals during finger movement tasks, providing valuable insight into the complex dynamics of brain operation and interaction. Utilizing 10-fold cross-validation, the performance of each classifier is rigorously evaluated across various metrics, including accuracy, f1_weighted та roc_auc_ovr_weighted. Through this process, the robustness and consistency of classifier performance are assessed, with dispersion values computed to gauge variability across iterations. The findings reveal nuanced differences among the classifiers, with MLP demonstrating the highest robustness, followed by Logistic Regression and Random Forest. The main goal of the article was to find out the importance of such a classifier performance parameter as robustness. Software robustness is a key characteristic, especially in medical applications, where consistent and reliable performance of information technology is paramount. Neural interfaces offer many avenues for solving various limb problems, spinal cord injuries, and neurological diseases in humans. These devices contribute to improving the quality of life by minimizing these problems in people, which leads to increased mobility and functional capabilities of people. The article also emphasizes the potential associated with the transformation of neurointerface technologies in expanding human capabilities and revolutionizing human-machine interaction. In conclusion, the research contributes to advancing the field of brain-computer interaction by leveraging machine learning algorithms to decode neural signals and uncover hidden patterns within EEG data. By identifying the most stable classifier, the study lays the groundwork for the development of robust neurointerface technologies with practical applications in healthcare, rehabilitation, and beyond. Through interdisciplinary collaboration and innovative methodologies, the journey towards unlocking the full potential of brain-computer interaction continues, promising new horizons in human augmentation and technological innovation.

Galaxy merger challenge: A comparison study between machine learning-based detection methods

Aims. Various galaxy merger detection methods have been applied to diverse datasets. However, it is difficult to understand how they compare. Our aim is to benchmark the relative performance of merger detection methods based on machine learning (ML). Methods. We explore six leading ML methods using three main datasets. The first dataset consists of mock observations from the IllustrisTNG simulations, which acts as the training data and allows us to quantify the performance metrics of the detection methods. The second dataset consists of mock observations from the Horizon-AGN simulations, introduced to evaluate the performance of classifiers trained on different, but comparable data to those employed for training. The third dataset is composed of real observations from the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) survey. We also compare mergers and non-mergers detected by the different methods with a subset of HSC-SSP visually identified galaxies. Results. For the simplest binary classification task (i.e. mergers vs. non-mergers), all six methods perform reasonably well in the domain of the training data. At the lowest redshift explored 0.1 < ɀ < 0.3, precision and recall generally range between ~70% and 80%, both of which decrease with increasing ɀ as expected (by ~5% for precision and ~10% for recall at the highest ɀ explored 0.76 < ɀ < 1.0). When transferred to a different domain, the precision of all classifiers is only slightly reduced, but the recall is significantly worse (by ~20–40% depending on the method). Zoobot offers the best overall performance in terms of precision and F1 score. When applied to real HSC observations, different methods agree well with visual labels of clear mergers, but can differ by more than an order of magnitude in predicting the overall fraction of major mergers. For the more challenging multi-class classification task to distinguish between pre-mergers, ongoing-mergers, and post-mergers, none of the methods in their current set-ups offer good performance, which could be partly due to the limitations in resolution and the depth of the data. In particular, ongoing-mergers and post-mergers are much more difficult to classify than pre-mergers. With the advent of better quality data (e.g. from JWST and Euclid), it is of great importance to improve our ability to detect mergers and distinguish between merger stages.

Data oversampling and feature selection for class imbalanced datasets

Introduction: Significant advancements and modifications have been implemented in data classification (DC) in the past few decades. Due to their infinite quantity and imbalance, data becomes challenging for classification. The biggest concern in DM (Data Mining) is Class Imbalance (CI). To avoid these issues in recent work proposed map reduce based data parallelization of class imbalanced datasets. Methods: A novel Over Sampling (OS) technique called Minority Oversampling in Kernel Canonical Correlation Adaptive Subspaces (MOKCCAS) has been suggested with the objective to minimize data loss throughout (FSP) Feature Space Projections. This technique takes advantage of the constant Feature Extraction (FE) capability of a version of the ASSOM (Adaptive Subspace Self-Organizing Maps) that is derived from Kernel Canonical Correlation Analysis (KCCA). And in classification, Feature Selection (FS) plays an important role because the acquired dataset might contain large volume of samples, utilizing all features of samples from the dataset for classification will decrease the classifier performance. And then data parallelization will be done by using map reduce framework to solve this computation requirement problem.Result: Then proposes a feature selection model using Mutated whale optimization (MWO) methods and produces features and reduces the time consumption. Finally proposed class balancing model will be tested using uniform distribution based enhanced adaptive neuro fuzzy inference system (UDANFIS). Test outcomes validate the efficiency of the suggested technique by precision, recall, accuracy and Error Rate (ER).Conclusion: The study subsequently suggests a novel OS approach called MOKCCAS to lessen the loss of data throughout feature space projection.

An improved DNN model for WLAN intrusion detection

Abstract Intrusion detection represents an efficacious approach for addressing security concerns. However, given the substantial volume and high-dimensional nature of WLAN dataset features, existing methods exhibit limited effectiveness in feature extraction, thereby impacting classification performance. To address above problems, an improved deep neural network (DNN) model for WLAN intrusion detection was proposed. Firstly, the activation function and loss function of a single sparse autoencoders (SAE) were determined through experiments, followed by the addition of regularization terms to the autoencoder, to prevent the model from overfitting. Subsequently, multiple SAEs were employed for a stacked architecture. This configuration served the purpose of feature dimension reduction and facilitated the selection of suitable feature dimensions for training the dataset. The chosen features were then utilized as the input layer for a DNN, with a SoftMax classifier serving as the output layer. Secondly, to obtain better DNN model parameters, the grid search method was adopted to optimize the parameters of the DNN model, namely activation, epochs, batch_size, init_mode, and optimizer. The results were visualized for assessment and analysis. Finally, the receiver operating characteristic curves were generated to assess the performance of various models, the analysis results show that the model exhibited better classifier performance.

Effect of Using Numerical Data Scaling on Supervised Machine Learning Performance

Before building machine learning models, the dataset should be prepared to be a high quality dataset, we should give the model the best possible representation of the data. Different attributes may have different scales which possibly will increase the difficulty of the problem that is modeled. A model with varying scale values may suffers from poor performance during learning. Our study explores the usage of Numerical Data Scaling as a data pre-processing step with the purpose of how effectively these methods can be used to improve the accuracy of learning algorithms. In particular, three numerical data Scaling methods with four machine learning classifiers to predict disease severity were compared. The experiments were built on Coronavirus 2 (SARS-CoV-2) datasets which included 1206 patients who were admitted during the period between June 2020 and April 2021. The diagnosis of all cases was confirmed with RT-PCR. Basic demographic data and medical characteristics of all participants was collected. The reported results indicate that all techniques are performing well with Numerical Data Scaling and there are significant improvement in the models for unseen data. lastly, we can conclude that there are increase in the classifier performance while using scaling techniques. However, these methods help the algorithms to better understand learn the patterns in the dataset which help making accurate models

Classification of Parkinson's disease severity using gait stance signals in a spatiotemporal deep learning classifier.

Parkinson's disease (PD) is a degenerative nervous system disorder involving motor disturbances. Motor alterations affect the gait according to the progression of PD and can be used by experts in movement disorders to rate the severity of the disease. However, this rating depends on the expertise of the clinical specialist. Therefore, the diagnosis may be inaccurate, particularly in the early stages of PD where abnormal gait patterns can result from normal aging or other medical conditions. Consequently, several classification systems have been developed to enhance PD diagnosis. In this paper, a PD gait severity classification algorithm was developed using vertical ground reaction force (VGRF) signals. The VGRF records used are from a public database that includes 93 PD patients and 72 healthy controls adults. The work presented here focuses on modeling each foot's gait stance phase signals using a modified convolutional long deep neural network (CLDNN) architecture. Subsequently, the results of each model are combined to predict PD severity. The classifier performance was evaluated using ten-fold cross-validation. The best-weighted accuracies obtained were 99.296(0.128)% and 99.343(0.182)%, with the Hoehn-Yahr and UPDRS scales, respectively, outperforming previous results presented in the literature. The classifier proposed here can effectively differentiate gait patterns of different PD severity levels based on gait signals of the stance phase.

Performance of clinical and histological prognostic scores for kidney survival in ANCA-associated vasculitis.

Integrating clinical and histological parameters into prognostic scores may enhance the prediction of progression to kidney failure in anti-neutrophil cytoplasm antibodies-associated vasculitis (AAV). This study aimed to evaluate the prognostic performance of histological classifications and scoring systems for kidney survival in AAV. This retrospective cohort study included 101 AAV patients with kidney involvement diagnosed by biopsy and followed for ≥12 months. The main outcome was the time to kidney failure. The prognostic performance of each histological and prognostic score was evaluated using Harrell's C statistic and Akaike's Information Criteria. Among the 101 patients, 37 progressed to kidney failure over a median follow-up of 75 months (IQR 39-123). The Harrell's C statistic was 0.702 (0.620-0.784), 0.606 (0.473-0.738), 0.801 (0.736-0.867), 0.782 (0.706-0.858), and 0.817 (0.749-0.885) for the EUVAS/Berden classification, Mayo Clinic Chronicity Score, Percentage of ANCA Crescentic Score (PACS), ANCA renal risk score (ARRS), and the improved ANCA kidney risk score (AKRiS), respectively. The AKRiS best discriminated the risk of kidney failure progression among subgroups. The AKRiS performance decreased with longer follow-up intervals. Adding the peak estimated glomerular filtration rate attained post-therapy improved the AKRiS performance at all follow-up intervals. Kidney relapses precipitated kidney failure in 71% of cases that progressed after the first year of follow-up. The novel AKRiS enhances the prediction of kidney failure in AAV with kidney involvement. As the prognostic yield of AKRiS decreases over time, a second calculation of AKRiS, including post-therapy kidney function, may improve its long-term performance.

Investigating the Performance of a Novel Modified Binary Black Hole Optimization Algorithm for Enhancing Feature Selection

High-dimensional datasets often harbor redundant, irrelevant, and noisy features that detrimentally impact classification algorithm performance. Feature selection (FS) aims to mitigate this issue by identifying and retaining only the most pertinent features, thus reducing dataset dimensions. In this study, we propose an FS approach based on black hole algorithms (BHOs) augmented with a mutation technique termed MBHO. BHO typically comprises two primary phases. During the exploration phase, a set of stars is iteratively modified based on existing solutions, with the best star selected as the “black hole”. In the exploration phase, stars nearing the event horizon are replaced, preventing the algorithm from being trapped in local optima. To address the potential randomness-induced challenges, we introduce inversion mutation. Moreover, we enhance a widely used objective function for wrapper feature selection by integrating two new terms based on the correlation among selected features and between features and classification labels. Additionally, we employ a transfer function, the V2 transfer function, to convert continuous values into discrete ones, thereby enhancing the search process. Our approach undergoes rigorous evaluation experiments using fourteen benchmark datasets, and it is compared favorably against Binary Cuckoo Search (BCS), Mutual Information Maximization (MIM), Joint Mutual Information (JMI), and minimum Redundancy Maximum Eelevance (mRMR), approaches. The results demonstrate the efficacy of our proposed model in selecting superior features that enhance classifier performance metrics. Thus, MBHO is presented as a viable alternative to the existing state-of-the-art approaches. We make our implementation source code available for community use and further development.

Optical coherence tomography-derived texture-based radiomics features identify eyes with intraocular inflammation in the HAWK clinical trial

Background and objectiveIntravitreal injection of anti-VEGF agents is the first-line treatment for patients with neovascular-age related macular degeneration (nAMD). One unique serious adverse event that may be associated with these agents is intraocular inflammation (IOI). The main purpose of this analysis was to evaluate the potential presence of texture-based radiomics features characterizing heterogeneity within the vitreous compartment of spectral domain optical coherence tomography (SD-OCT) images that may precede or develop in association with IOI and might serve as OCT biomarkers for IOI. MethodsThis is a post-hoc analysis of a subset of cases (N = 67) involving IOI, endophthalmitis, and/or retinal vascular occlusion in the phase 3 HAWK trial. These were investigator determined diagnoses that were also confirmed by the safety review committee. Intraocular inflammation was any signs of inflammation within the eye, endophthalmitis was inflammation associated with presumed infection, and retinal vascular occlusions consisted of intraocular inflammation with concurrent vascular occlusions/vasculitis. Out of 67 eyes, 34 belonged to the Safety group with an IOI event and 33 were propensity-matched Controls. A total of 481 texture-based radiomics features were extracted from the vitreous compartment of the SD-OCT scans at pre-IOI time point (i.e., much earlier than the actual event). Most discriminating five features, selected by the Wilcoxon Rank Sum feature selection were evaluated using Random Forest (RF) classifier on the training set (Str, N = 47) to differentiate between the two patient groups. Classifier performance was subsequently validated on the independent test set (St, N = 20). Additionally, the classifier performance in discriminating the Control and Safety group was also validated on St at the IOI event timepoint. ResultsThe RF classifier yielded area under the Receiver Operating Characteristics curve (AUC) of 0.76 and 0.81 on St using texture-based radiomics features at pre-IOI and event time-point, respectively. ConclusionsIn this analysis, the presence of a pre-IOI safety signal was detected in the form of textural heterogeneity within the vitreous compartment even prior to the actual event being identified by the investigator. This finding may help the clinicians to assess for underlying posterior inflammation.

New Insights into Gas-in-Oil-Based Fault Diagnosis of Power Transformers

The dissolved gas analysis of insulating oil in power transformers can provide valuable information about fault diagnosis. Power transformer datasets are often imbalanced, worsening the performance of machine learning-based fault classifiers. A critical step is choosing the proper evaluation metric to select features, models, and oversampling techniques. However, no clear-cut, thorough guidance on that choice is available to date. In this work, we shed light on this subject by introducing new tailored evaluation metrics. Our results and discussions bring fresh insights into which learning setups are more effective for imbalanced datasets.

Comparison of the classifiers based on mRNA, microRNA and lncRNA expression and DNA methylation profiles for the tumor origin detection.

Tumor tissue origin detection is of great importance in determining the appropriate course of treatment for cancer patients. Classifiers based on gene expression and DNA methylation profiles have been confirmed to be feasible and reliable to predict the tumor primary. However, few works have been performed to compare the performance of these classifiers based on different profiles. Using gene expression and DNA methylation profiles from The Cancer Genome Atlas (TCGA) project, eight machine learning methods were employed for the tumor tissue origin detection. We then evaluated the predictive performance using DNA methylation, mRNA, microRNA (miRNA) and long non-coding RNA (lncRNA) expression profiles in a comparative manner. A statistical method was introduced to select the most informative CpG sites. We found that LASSO is the most predictive models based on various profiles. Further analyses indicated that the results derived from DNA methylation (overall accuracy: 97.77%) are better than those derived from mRNA expression (overall accuracy: 88.01%), microRNA expression (overall accuracy: 91.03%) and lncRNA expression (overall accuracy: 95.7%). It has been suggested that we can achieve an overall accuracy >90% using only 1,000 methylated CpG sites for prediction. In this work, we comprehensively evaluated the performance of classifiers based on different profiles for the tumor origin detection. Our findings demonstrated the effectiveness of DNA methylation as biomarker for tracing tumor tissue origin using LASSO and neural network.

Consistency of the Signature of Phonotraumatic Vocal Hyperfunction Across Different Ambulatory Voice Measures.

Although different factors and voice measures have been associated with phonotraumatic vocal hyperfunction (PVH), it is unclear what percentage of individuals with PVH exhibit such differences during their daily lives. This study used a machine learning approach to quantify the consistency with which PVH manifests according to ambulatory voice measures. Analyses included acoustic parameters of phonation as well as temporal aspects of phonation and rest, with the goal of determining optimally consistent signatures of PVH. Ambulatory neck-surface acceleration signals were recorded over 1 week from 116 female participants diagnosed with PVH and age-, sex-, and occupation-matched vocally healthy controls. The consistency of the manifestation of PVH was defined as the percentage of participants in each group that exhibited an atypical signature based on a target voice measure. Evaluation of each machine learning model used nested 10-fold cross-validation to improve the generalizability of findings. In Experiment 1, we trained separate logistic regression models based on the distributional characteristics of 14 voice measures and durations of voicing and resting segments. In Experiments 2 and 3, features of voicing and resting duration augmented the existing distributional characteristics to examine whether more consistent signatures would result. Experiment 1 showed that the difference in the magnitude of the first two harmonics (H1-H2) exhibited the most consistent signature (69.4% of participants with PVH and 20.4% of controls had an atypical H1-H2 signature), followed by spectral tilt over eight harmonics (73.6% participants with PVH and 32.1% of controls had an atypical spectral tilt signature) and estimated sound pressure level (SPL; 66.9% participants with PVH and 27.6% of controls had an atypical SPL signature). Additionally, 77.6% of participants with PVH had atypical resting duration, with 68.9% exhibiting atypical voicing duration. Experiments 2 and 3 showed that augmenting the best-performing voice measures with univariate features of voicing or resting durations yielded only incremental improvement in the classifier's performance. Females with PVH were more likely to use more abrupt vocal fold closure (lower H1-H2), phonate louder (higher SPL), and take shorter vocal rests. They were also less likely to use higher fundamental frequency during their daily activities. The difference in the voicing duration signature between participants with PVH and controls had a large effect size, providing strong empirical evidence regarding the role of voice use in the development of PVH.

Pattern-centric transformation of omics data grounded on discriminative gene associations aids predictive tasks in TCGA while ensuring interpretability.

The increasing prevalence of omics data sources is pushing the study of regulatory mechanisms underlying complex diseases such as cancer. However, the vast quantities of molecular features produced and the inherent interplay between them lead to a level of complexity that hampers both descriptive and predictive tasks, requiring custom-built algorithms that can extract relevant information from these sources of data. We propose a transformation that moves data centered on molecules (e.g., transcripts and proteins) to a new data space focused on putative regulatory modules given by statistically relevant co-expression patterns. To this end, the proposed transformation extracts patterns from the data through biclustering and uses them to create new variables with guarantees of interpretability and discriminative power. The transformation is shown to achieve dimensionality reductions of up to 99% and increase predictive performance of various classifiers across multiple omics layers. Results suggest that omics data transformations from gene-centric to pattern-centric data supports both prediction tasks and human interpretation, notably contributing to precision medicine applications.

Towards the generalization of time series classification: A feature-level style transfer and multi-source transfer learning perspective

Transfer learning-based methods hold promise for enhancing classification task performance. However, a transfer learning mechanism for hard-to-classify time series classification tasks caused by limited samples in training set is still to be accomplished. Drawing inspiration from domain adaptation and style transfer methods in computer vision, we aim to address sample scarcity and enhance classifier generalization by leveraging abundant unrelated time series datasets. In this paper, a transfer learning mechanism called feature-level style transfer, where the feature alignment is conducted in a domain adaptation-inspired manner, has been proposed to target initially hard-to-classify time series tasks with limited training samples. The proposed mechanism eliminates constraints on source dataset selection, enabling the utilization of weakly-related or unrelated datasets to enhance classifier performance. Furthermore, a voting mechanism has been formulated to achieve more accurate predictions by taking transferable information learnt from different source domains into comprehensive consideration.

Balancing act: Tackling organized retail fraud on e-commerce platforms with imbalanced learning text models

As online shopping expands rapidly, so does the prevalence of fraud, resulting in significant losses for retailers. According to the 2020 National Retail Federation (NRF) report, organized retail crime costs retailers nearly $800,000 per billion in sales, with an expected global annual increase of over fourteen percent. This paper introduces a text-based fraud detection framework to mitigate these losses efficiently. The framework comprises four key components: text preprocessing, representation, knowledge extraction via machine learning algorithms, and model evaluation. By integrating data augmentation techniques, the framework enhances classifier performance in detecting fraud. The proposed method, employing a combination of FastText and Random Forest classifiers, achieves an impressive F1 score of 0.833 and AUC score of 0.99 on an augmented dataset, surpassing conventional keyword-based models. Informed by best practices in fraud detection, this scalable framework promises a solution to combat the escalating fraud associated with the exponential growth of online shopping.

GMADV: An android malware variant generation and classification adversarial training framework

Android malware uses anti-reverse analysis and APK shelling technology, which leads to the failure of the classification method based on decompiled features and the reduction of the classification accuracy based on single file features. Moreover, the lack of samples in some families of Android malware makes the classification model based on sample learning ineffective. To solve the above problems, this paper proposes a two-layer general framework for Android malware classification and adversarial training named GMADV, which enhances classifier performance through adversarial training. In the sample classification layer, based on the transformation method of the Markov model, it is proposed for the first time to convert the three files in the APK into RGB Markov images, and use VGG13 to automatically extract features and classification; In the variant amplification layer, the idea of "regression for generation" is firstly proposed, and GMM-GAN based on Gaussian process is designed to amplify the diversity of samples within the family. The experimental results show that RGB Markov images have better classification performance than grayscale images. On the three datasets, the classification effect after amplification has been improved to varying degrees, and all F1_Score reaches 95 %. Compared with other methods, GMADV has stronger family sample amplification ability and greater adversarial intensity.

MEDAL: A Multimodality-Based Effective Data Augmentation Framework for Illegal Website Identification

The emergence of illegal (gambling, pornography, and attraction) websites seriously threatens the security of society. Due to the concealment of illegal websites, it is difficult to obtain labeled data with high quantity. Meanwhile, most illegal websites usually disguise themselves to avoid detection; for example, some gambling websites may visually resemble gaming websites. However, existing methods ignore the means of camouflage in a single modality. To address the above problems, this paper proposes MEDAL, a multimodality-based effective data augmentation framework for illegal website identification. First, we established an illegal website identification framework based on tri-training that combines information from different modalities (including image, text, and HTML) while making full use of numerous unlabeled data. Then, we designed a multimodal mutual assistance module that is integrated with the tri-training framework to mitigate the introduction of error information resulting from an unbalanced single-modal classifier performance in the tri-training process. Finally, the experimental results on the self-developed dataset demonstrate the effectiveness of the proposed framework, performing well on accuracy, precision, recall, and F1 metrics.

Machine learning assisted prediction of land surface temperature (LST) based on major air pollutants over the Annamayya District of India

Remote sensing (RS), Geographic information systems (GIS), and Machine learning can be integrated to predict land surface temperatures (LST) based on the data related to carbon monoxide (CO), Formaldehyde (HCHO), Nitrogen dioxide (NO2), Sulphur dioxide (SO2), absorbing aerosol index (AAI), and Aerosol optical depth (AOD). In this study, LST was predicted using machine learning classifiers, i.e., Extra trees classifier (ET), Logistic regressors (LR), and Random Forests (RF). The accuracy of the LR classifier (0.89 or 89%) is higher than ET (82%) and RF (82%) classifiers. Evaluation metrics for each classifier are presented in the form of accuracy, Area under the curve (AUC), Recall, Precision, F1 score, Kappa, and MCC (Matthew’s correlation coefficient). Based on the relative performance of the ML classifiers, it was concluded that the LR classifier performed better. Geographic information systems and RS tools were used to extract the data across spatial and temporal scales (2019 to 2022). In order to evaluate the model graphically, ROC (Receiver operating characteristic) curve, Confusion matrix, Validation curve, Classification report, Feature importance plot, and t- SNE (t-distributed stochastic neighbour embedding) plot were used. On validation of each ML classifier, it was observed that the RF classifier returned model complexity due to limited data availability and other factors yet to be studied post data availability. Sentinel-5-P and MODIS data are used in this study.

THE EFFECT OF FEATURE EXTRACTION ON COVID-19 CLASSIFICATION

X-ray imaging stands as a prominent technique for diagnosing COVID-19, and it also serves as a crucial tool in the medical field for the analysis of various diseases. Numerous approaches are available to facilitate this analysis. Among these techniques, one involves the utilization of a Feature Extractor, which effectively captures pertinent characteristics from X-ray images. In a recent study, a comprehensive examination was conducted using 25 distinct feature extractors on X-ray images specific to COVID-19 cases. These images were categorized into two classes: COVID-19-positive and non-COVID-19. To enable a thorough evaluation, a sequence of machine learning classifiers was employed on these categorized images. The outcomes derived from this experimentation gauged the magnitude of impact that each individual feature exerted on COVID-19-related imagery. This assessment aimed to determine the efficacy levels of various feature extractors in terms of detection capability. Consequently, a distinction emerged between the more effective and less effective feature extractors, shedding light on their varying degrees of contribution to the detection process. Moreover, the comparative performance of different classifiers became evident, revealing the classifiers that exhibited superior performance when measured against their counterparts.