Accuracy Of Classifier Research Articles

Non-destructive pre-symptomatic detection of gray mold infection in kiwifruit using hyperspectral data and chemometrics

Application of hyperspectral imaging (HSI) and data analysis algorithms was investigated for early and non-destructive detection of Botrytis cinerea infection. Hyperspectral images were collected from laboratory-based contaminated and non-contaminated fruits at different day intervals. The spectral wavelengths of 450 nm to 900 nm were pretreated by applying moving window smoothing (MWS), standard normal variates (SNV), multiplicative scatter correction (MSC), Savitzky–Golay 1st derivative, and Savitzky–Golay 2nd derivative algorithms. In addition, three different wavelength selection algorithms, namely; competitive adaptive reweighted sampling (CARS), uninformative variable elimination (UVE), and successive projection algorithm (SPA), were executed on the spectra to invoke the most informative wavelengths. The linear discriminant analysis (LDA), developed with SNV-filtered spectral data, was the most accurate classifier to differentiate the contaminated and non-contaminated kiwifruits with accuracies of 96.67% and 96.00% in the cross-validation and evaluation stages, respectively. The system was able to detect infected samples before the appearance of disease symptoms. Results also showed that the gray-mold infection significantly influenced the kiwifruits’ firmness, soluble solid content (SSC), and titratable acidity (TA) attributes. Moreover, the Savitzky–Golay 1st derivative-CARS-PLSR model obtained the highest prediction rate for kiwifruit firmness, SSC, and TA with the determination coefficient (R2) values of 0.9879, 0.9644, 0.9797, respectively, in calibration stage. The corresponding cross-validation R2 values were equal to 0.9722, 0.9317, 0.9500 for firmness, SSC, and TA, respectively. HSI and chemometric analysis demonstrated a high potential for rapid and non-destructive assessments of fungal-infected kiwifruits during storage.

Semantic variant Primary Progressive Aphasia: Additional semantic features in Chinese language users

AbstractBackgroundTo date, most literature on Primary Progressive Aphasia (PPA) has been based on English language users. Yet, there are more than 7,000 living languages worldwide with vast heterogeneities in their linguistic features. Contrary to English language, Chinese language is a classifier language in which it is grammatically mandatory to have classifiers present between numerals and nouns. In addition, Chinese characters typically have an abundance of homophones (i.e., words with same pronunciation but vary in orthography and meaning) and heteronyms (i.e., words with same orthography but vary in meaning and pronunciation). Chinese language users rely on semantic knowledge to identify the accurate classifiers, homophones, and heteronym pronunciations. Thus, we hypothesize that Chinese‐speaking semantic variant (sv) PPA patients have difficulties producing classifiers, homophones, and heteronyms.MethodTo assess classifier production, participants are presented with pictures of nouns. Participants are first tasked with producing noun‐specific classifiers spontaneously. If failed, they are presented with four classifiers to select. To evaluate homophone selection skills, participants are shown short sentences with one character/word removed. They are then tasked to complete the sentences by choosing one of the four homophones presented. For heteronym reading, participants are asked to read out loud Chinese compound words with heteronyms (e.g., “windmill” and “wind up”).ResultWe recruited 22 controls and 47 PPA individuals (12 svPPA, 10 nonfluent variant (nfv) PPA and 25 logopenic variant (lv) PPA). In classifier production test, svPPA and lvPPA patients produce fewer accurate classifiers (p<0.0001, p = 0.001), with svPPA showing least improvement when presented with multiple choices (p = 0.003). In homophone selection test, svPPA and lvPPA scored lower than controls (p<0.0001 and p = 0.007). For heteronym reading test, svPPA showed the lowest performance when tasked with reading heteronym characters (p = 0.001). Using voxel‐based morphometry, we noted that the ability to produce classifiers, read heteronyms and identify heteronyms are correlated with grey matter volumetric changes over the anterior portion of the left temporal lobe and fusiform gyri, brain regions associated with language semantic functions.ConclusionPPA symptomatology may vary across languages. Thus, studying PPA symptoms via a linguistically diverse approach potentially provides a more comprehensive understanding to the syndrome.

Modeling of the Acute Lymphoblastic Leukemia Detection by Convolutional Neural Networks (CNNs).

The techniques differed in many of the literature on the detection of Acute Lymphocytic Leukemia from the blood smear pictures, as the cases of infection in the world and the Kingdom of Saudi Arabia were increasing and the causes of this disease were not known, especially for children, which is a serious and fatal disease. Through this work we seek to contribute to discover the blood cells affected by Acute Lymphocytic Leukem and to find an effective and fast method and to have the correct diagnosis as the time factor is important in the diagnosis and the initiation of treatment. which is based on one of the deep learning techniques that specialize in very deep networks, the use of one of the CNNs is VGG16. Detection scheme is implemented by pre-processing, feature extraction, model building, fine tuning method, classification are executed. By using VGG16 pre-trained, and using SVM and MLP classification algorithms in Machine Learning. Our results are evaluated based on criteria, such as Accuracy, Precision, Recall, and F1-Score. The accuracy results for SVM classifier MLP of 77% accuracy at 0.001 learning rate and the accuracy for SVM classifier 75% at 0.005 learning rate. Whereas, the best accuracy result for VGG16 model was 92.27% at 0.003 learning rate. The best validation accuracy result was 85.62% at 0.001 learning rate.

Role of Multiparametric Ultrasound in Evaluating Hepatic Acute Graft-versus-Host Disease: An Animal Study.

Hepatic acute graft-versus-host disease (aGVHD) is a serious complication of allogeneic hematopoietic stem cell transplantation (allo-HSCT) and is one of the leading causes of early non-recurrent death. The current diagnosis is based mainly based on clinical diagnosis, and there is a lack of non-invasive quantitative diagnosis methods. We propose a multiparametric ultrasound (MPUS) imaging method and explore its effectiveness in evaluating hepatic aGVHD. In this study, 48 female Wistar rats were used as receptors and 12 male Fischer 344 rats were used as donors for allo-HSCT to establish aGVHD models. After transplantation, 8 rats were randomly selected for ultrasonic examination weekly, including color Doppler ultrasound, contrast-enhanced ultrasound (CEUS) and shear wave dispersion (SWD) imaging. The values of nine ultrasonic parameters were obtained. Hepatic aGVHD was subsequently diagnosed by histopathological analysis. A classification model for predicting hepatic aGVHD was established using principal component analysis and support vector machines. According to the pathological results, the transplanted rats were categorized into the hepatic aGVHD and non-GVHD (nGVHD) groups. All parameters obtained by MPUS differed statistically between the two groups. The first three contributing percentages of principal component analysis results were resistivity index, peak intensity and shear wave dispersion slope, respectively. The accuracy of classifying aGVHD and nGVHD using support vector machines reached 100%. The accuracy of the multiparameter classifier was significantly higher than that of the single parameter. The MPUS imaging method has proven to be useful in detecting hepatic aGVHD.

Analysis of tumor- and circulating-free DNA methylation to identify clinically relevant small cell lung cancer subtypes.

8577 Background: Small-cell lung cancer (SCLC) is an aggressive malignancy composed of distinct transcriptional subtypes, defined by the predominant expression of one of the three transcription factors ASCL1 (SCLC-A), NEUROD1 (SCLC-N) and POU2F3 (SCLC-P) as well as an inflamed subtype (SCLC-I; see Gay et al. Cancer Cell. 2021), each with potential therapeutic vulnerabilities. Implementing subtyping in the clinic has remained challenging due to limited tissue availability, particularly for longitudinal monitoring. Given the known epigenetic regulation of critical SCLC transcriptional programs, we hypothesized that there would be subtype-specific patterns of DNA methylation that could be detected in tumor or blood from SCLC patients. Methods: We included 179 patients with SCLC and performed RNA sequencing and genomic-wide reduced-representation bisulfite sequencing (RRBS). We further analyzed DNA methylation in 68 plasma samples including longitudinal samples to track SCLC subtype evolution over time. Results: Using machine learning approaches, we developed a highly accurate DNA methylation-based classifier (SCLC-DMC) that could distinguish SCLC subtypes using clinical tumor samples with 95.8% accuracy in the testing set compared to mRNA-based profiling. We further adjusted the classifier for circulating-free DNA (cfDNA) to subtype SCLC from plasma. Using the cfDNA classifier (cfDMC), we could demonstrate that SCLC subtypes evolve frequently during disease progression, highlighting the need for longitudinal tracking of SCLC during clinical treatment. Furthermore, methylation-based subtyping predicted response to a wide variety of drugs in preclinical models like CDK and AURK inhibitors, and clinical outcomes were indistinguishable in cohorts of patients subtyped using mRNA or SCLC-DMC (p = 0.95). Conclusions: These data establish that tumor and cfDNA methylation can be used to identify SCLC subtypes and guide precision SCLC therapy.

Using cell-free DNA to predict response to immunotherapy in non-small cell lung cancer pre- and post-treatment.

e21198 Background: Despite revolutionizing cancer therapy, immune checkpoint inhibitors (ICI) still do not benefit a significant proportion of patients. The risks associated with ICI-related adverse events, mixed performance of PD-L1 staining in predicting treatment response and its high cost, present a clinical need for more precise methods to define disease states in the context of ICI treatment. ICI response is thought to depend on the phenotype of the tumor and the associated immune response, especially functional state of CD8+ T cells in the tumor microenvironment (TME). Methods: Here we show that the CD8+ T cell signatures from plasma cell-free DNA (cfDNA) could be used to predict response of patients to anti-PD-1 therapy, using samples collected before and shortly after start of treatment. Blood samples were drawn just before the first dose, 1 day to 1 week before the start of treatment and 45-60 days since start of treatment. The treatment duration varied depending on response. Response was evaluated by CT scans every 8-12 weeks. 11 of the samples are from patients with no or minor response ( < 6 months of treatment), and 12 are from patients with prolonged benefit of the medication ( > 1 year of treatment). Exhausted CD8+ T cells, one of the targets of immunotherapy treatment are epigenetically regulated. So, we utilized naïve and PD-1high CD8+ T cell ATAC-seq data to define the repertoire of accessible chromatin specific to these cell types. We found an enrichment of short cfDNA fragments at a significant fraction of these sites, enabling us to define responder-specific and non-responder-specific accessible PD-1high regions from cfDNA. Results: We used the enrichment of short- versus long-cfDNA fragments (reflecting transcription factor-nucleosome dynamics in tissue of origin of cfDNA) as a scoring function, and repeated-cross-validation as a classifier model to identify differentially enriched features between responders and non-responders. Our model accurately predicts the two classes of response both pre-treatment (mean test AUC = 0.86; SD = 0.14). Conclusions: Notably, in addition to generating an accurate classifier, our analysis enabled us to identify predominant transcription factor motifs from predictive ATAC-seq peaks that characterize both response and lack of response, paving the way for further understanding the mechanistic basis of patient-specific response to ICI. Our results suggest the possibility of personalized prediction of treatment response that is independent of specific tumor genotype.

Physics verification and validation for transferring data between bearings

Bearings are critical components for transferring load and motion between subsystems with reduced friction for rotational equipment. Manufacturers implement condition monitoring technology to prevent failures of bearings by using different data acquisition methods, such as vibration, acoustics, temperature, motor current, and ultrasonic sensing, to monitor and predict changes that could indicate early equipment degradation. However, the data quality and availability varies depending on the application, and low data availability from physical environments can lead to poorly trained models. This paper explores how to transfer data and information about failure propensity between bearings of different sizes using a combined physics and data-driven approach. Though the approach is exemplified with roller bearings, the method is extensible to other types of failures in different-sized equipment. Data are generated using an experimental test stand measuring failure of one component, with the intent to scale findings to represent a real-world system; findings are verified with a physics-based model. Data from three different bearing sizes, i.e. 6205, 6206, and 6207, are used to train the algorithms to identify similarity among the sets. Classifiers trained with the raw data provide over 90% accuracy, leading to the conclusion that the data classes are separable based on bearing size. 6205 and 6207 data were scaled to simulate 6206 and tested to see if a classifier could differentiate between the true 6206 bearing data and the simulated bearing data derived from the 6205 and 6207 data. In the simulated data, the classifier accuracy for each algorithm dropped below 90% to as low as 50% (Naive Bayes case). The lower accuracy implied a greater overlap in the vibration features, increasing the data similarity between the different bearing sizes. Future work will further investigate defining dimensionless numbers as scaling parameters for bearing data.

A Cloud Approach for Melanoma Detection Based On Deep Learning Networks

Abstract: Utilizing PC vision, machine learning, and deep learning , the objective is to track down new data and concentrate data from advanced pictures. Images can now be used for both early illness detection and treatment. Dermatology use deep neural network to tell the difference between images with and without melanoma. Two important melanoma location research topics have been emphasized in this essay. Classifier accuracy is impacted by even minor alterations to dataset’s bounds, the primary variable under investigation. We examined the exchange learning issues in this example. We propose using continuous preparation test cycles to create trustworthy prediction models on the basis of this initial evaluation’s findings . Seconds, a more flexible design philosophy that can oblige changes in the preparation datasets is fundamental. We recommended the creation and utilization of a half breed plan in view of cloud, dimness, and edge figuring to give Melanoma Area the board in light of clinical and dermoscopic pictures. By lessing the span of the consistent retrain, this designing must continually adjust to the quantity of data the should be investigated. This aspect has been highlighted in experiments coduted on a dingle Pc using various conveyance method, demonstrating how a distributed system guarantees yield fulfillment in an unquestionably more acceptable amount of time

Gut and oral microbiome modulate molecular and clinical markers of schizophrenia-related symptoms: A transdiagnostic, multilevel pilot study

Although increasing evidence links microbial dysbiosis with the risk for psychiatric symptoms through the microbiome-gut-brain axis (MGBA), the specific mechanisms remain poorly characterized. In a diagnostically heterogeneous group of treated psychiatric cases and nonpsychiatric controls, we characterized the gut and oral microbiome, plasma cytokines, and hippocampal inflammatory processes via proton magnetic resonance spectroscopic imaging (1H-MRSI). Using a transdiagnostic approach, these data were examined in association with schizophrenia-related symptoms measured by the Positive and Negative Syndrome Scale (PANSS). Psychiatric cases had significantly greater heterogeneity of gut alpha diversity and an enrichment of pathogenic taxa, like Veillonella and Prevotella, in the oral microbiome, which was an accurate classifier of phenotype. Cases exhibited significantly greater positive, negative, and general PANSS scores that uniquely correlated with bacterial taxa. Strong, positive correlations of bacterial taxa were also found with cytokines and hippocampal gliosis, dysmyelination, and excitatory neurotransmission. This pilot study supports the hypothesis that the MGBA influences psychiatric symptomatology in a transdiagnostic manner. The relative importance of the oral microbiome in peripheral and hippocampal inflammatory pathways was highlighted, suggesting opportunities for probiotics and oral health to diagnose and treat psychiatric conditions.

Pulse repetition interval modulation recognition using deep CNN evolved by extreme learning machines and IP-based BBO algorithm

Pulse repetition interval modulation (PRIM) recognition is a critical task in electronic intelligence (ELINT) and electronic support measure (ESM) systems for detecting radar threats accurately. However, PRI recognition is a complex issue due to missing and spurious pulses, resulting in noisy PRI pattern changes in real environments. To address this problem, this paper proposes a novel approach that recognizes the five common types of PRIM through a four-phase process. In the first phase, a deep convolutional neural network (DCNN) is used as a feature extractor. Then, extreme learning machines (ELMs) are used for real-time recognition of the PRIM patterns in the second phase. In the third phase, we employ the biogeography-based optimizer (BBO) to enhance the network’s robustness by optimizing the connection weights and biases. To address the increasing complexity of the model, we introduce an optimized variable-length internet protocol-based BBO (VBBO) in the fourth phase. In this approach (i.e., DCNN-VBBO-ELM), each layer of DCNN is encoded by an IP address into a habitat of VBBO in the same sequence as the DCNN layers. To evaluate the proposed method, we develop a real experimental dataset consisting of five common PRI patterns. Our approach achieves a final accuracy of 97.05%, which is better than other ELM-based benchmark models. Moreover, the proposed model requires only 27 s of training time to process 50,000 training images, confirming its real-time capabilities. In conclusion, our proposed approach improves PRI recognition by leveraging DCNN, ELM, and VBBO, resulting in a more accurate and robust real-time radar PRI classifier.

RedTell: an AI tool for interpretable analysis of red blood cell morphology.

Introduction: Hematologists analyze microscopic images of red blood cells to study their morphology and functionality, detect disorders and search for drugs. However, accurate analysis of a large number of red blood cells needs automated computational approaches that rely on annotated datasets, expensive computational resources, and computer science expertise. We introduce RedTell, an AI tool for the interpretable analysis of red blood cell morphology comprising four single-cell modules: segmentation, feature extraction, assistance in data annotation, and classification. Methods: Cell segmentation is performed by a trained Mask R-CNN working robustly on a wide range of datasets requiring no or minimum fine-tuning. Over 130 features that are regularly used in research are extracted for every detected red blood cell. If required, users can train task-specific, highly accurate decision tree-based classifiers to categorize cells, requiring a minimal number of annotations and providing interpretable feature importance. Results: We demonstrate RedTell's applicability and power in three case studies. In the first case study we analyze the difference of the extracted features between the cells coming from patients suffering from different diseases, in the second study we use RedTell to analyze the control samples and use the extracted features to classify cells into echinocytes, discocytes and stomatocytes and finally in the last use case we distinguish sickle cells in sickle cell disease patients. Discussion: We believe that RedTell can accelerate and standardize red blood cell research and help gain new insights into mechanisms, diagnosis, and treatment of red blood cell associated disorders.

Non-specific myocardial fibrosis in young competitive athletes, predicted by a powerful machine learning-based model

Abstract Funding Acknowledgements Type of funding sources: None. Background Non-specific myocardial fibrosis (NSMF) is a heterogeneous entity with potentially significant clinical implications. We aimed to create a machine-learning (ML) based model for its prediction, based on thorough baseline investigations in a cohort of young competitive athletes with and without NSMF. Methods We analysed data from 328 young athletes referred to our Sports Cardiology service for a variety of reasons. All athletes underwent an evaluation with ECG, Holter, cardiopulmonary exercise test and cardiac magnetic resonance (CMR). We identified 61 athletes with NSMF (80% male, 72% white, 65% endurance sport) and compared them with a matched group of 75 athletes with no fibrosis (controls). Athletes with NSMF were further divided into Group 1 (n=28) with minimal (‘minor’=insertion point) fibrosis and Group 2 (n=33), (all other patterns, ‘major’) fibrosis. Selected baseline athletes’ characteristics including demographics, ECG findings , exercise intensity and type, arrhythmia, echo and CMR findings were tested to train the model. Finally, we tested various machine learning (ML) algorithms to create a model which classifies the individuals under observance into two distinct classes; Class A which contains the control group plus the athletes with ‘minor’ fibrosis, while the second, Class B which contains the athletes with ‘major’ fibrosis. We created 4 different classifiers; a logistic regression classifier, a random forest classifier, a naive Bayes classifier, and a voting classifier which chooses the class predicted by the majority of the previous three. Results Given that ~75% of athletes belonged to either no fibrosis or Group 1 (class A), and ~25% to Group 2, it can be inferred that a dummy classifier who would always predict class A would achieve an accuracy of 75%. We therefore used such a classifier method, which a-priori predicts class A, as a baseline to which we compared the results of our trained classifiers. We split the dataset, trained the model, and measured its accuracy (and the accuracy of the benchmark a-priori classifier) on the test set for each classifier. Machine learning classifiers had the following accuracy classification results; ȃȃA-priori classifier: 75% [74%, 76] ȃȃNaive Bayes: 83% [81%, 85%] <0.001 ȃȃLogistic Regression: 87% [86%, 88%] <0.001 ȃȃVoting classifier: 89% [88%, 90%] <0.001 ȃȃRandom forest: 90% [89%, 91%] <0.001 Testing the models derived by machine learning, all our models significantly outperformed the benchmark method and the best accuracy was achieved by the random forest classifier with 90% mean accuracy. Conclusions This study showed that ECG findings, demographics and comprehensive baseline assessment can be used to develop a powerful ML-based prediction model of NSMF. Testing ML-derived models, they all significantly outperformed the benchmark method and the best accuracy was achieved by the random forest classifier with 90% mean accuracy.

Generative adversarial network based synthetic data training model for lightweight convolutional neural networks.

Inadequate training data is a significant challenge for deep learning techniques, particularly in applications where data is difficult to get, and publicly available datasets are uncommon owing to ethical and privacy concerns. Various approaches, such as data augmentation and transfer learning, are employed to address this problem, which help to some extent in removing this limitation. However, after a certain amount of data augmentation, the quality of the generated data stalls, and transfer learning suffers from the issue of negative transfer. This paper proposes a novel generative adversarial network-based synthetic data training (GAN-ST) model to generate synthetic data for training a lightweight convolutional neural network (CNN). An enhanced generator is proposed to quickly saturate and cover the colour space of the training distribution. The GAN-ST model is based on Deep Convolutional Generative Adversarial Network(s) (DCGAN) and Conditional Generative Adversarial Network(s) (CGAN) models, which consist of an enhanced generator. The study evaluates the accuracy of a CNN model on the MNIST and CIFAR 10 datasets using both original and synthetic data. The results revealed an impressive classifier accuracy on the MNIST dataset, achieving an accuracy of 99.38% on GAN-ST-generated synthetic training data, which is only 0.05% lower than the performance on original data-based training. The classifier performance on the CIFAR dataset is also remarkable, achieving an accuracy of 90.23%. The performance of CNN trained using GAN-ST-based synthetic data is notable, with the most considerable improvement of 0.66% and 7.06%, over a single GAN-based synthetic data training for the MNIST and CIFAR datasets, respectively. By training two GANs independently, the GAN-ST model covers different parts of the original data distribution, resulting in a more diverse and realistic training data set for the classifier. This diverse set of synthetic data, when used to train a CNN, shows better generalization to new data, leading to improved classification accuracy.

A highly generalized classifier for osteoporosis radiography based on multiscale fractal, lacunarity, and entropy distributions.

Background: Osteoporosis is a common degenerative disease with high incidence among aging populations. However, in regular radiographic diagnostics, asymptomatic osteoporosis is often overlooked and does not include tests for bone mineral density or bone trabecular condition. Therefore, we proposed a highly generalized classifier for osteoporosis radiography based on the multiscale fractal, lacunarity, and entropy distributions. Methods: We collected a total of 104 radiographs (92 for training and 12 for testing) of lumbar spine L4 and divided them into three groups (normal, osteopenia, and osteoporosis). In parallel, 174 radiographs (116 for training and 58 for testing) of calcaneus from health and osteoporotic fracture groups were collected. The texture feature data of all the radiographs were pulled out and analyzed. The Davies-Bouldin index was applied to optimize hyperparameters of feature counting. Neighborhood component analysis was performed to reduce feature dimension and increase generalization. A support vector machine classifier was trained with only the most effective six features for each binary classification scenario. The accuracy and sensitivity performance were estimated by calculating the area under the curve. Results: Interpretable feature trends of osteoporotic pathological changes were depicted. On the spine test dataset, the accuracy and sensitivity of binary classifiers were 0.851 (95% CI: 0.730-0.922), 0.813 (95% CI: 0.718-0.878), and 0.936 (95% CI: 0.826-1) for osteoporosis diagnosis; 0.721 (95% CI: 0.578-0.824), 0.675 (95% CI: 0.563-0.772), and 0.774 (95% CI: 0.635-0.878) for osteopenia diagnosis; and 0.935 (95% CI: 0.830-0.968), 0.928 (95% CI: 0.863-0.963), and 0.910 (95% CI: 0.746-1) for osteoporosis diagnosis from osteopenia. On the calcaneus test dataset, they were 0.767 (95% CI: 0.629-0.879), 0.672 (95% CI: 0.545-0.793), and 0.790 (95% CI: 0.621-0.923) for osteoporosis diagnosis. Conclusion: This method showed the capacity of resisting disturbance on lateral spine radiographs and high generalization on the calcaneus dataset. Pixel-wise texture features not only helped to understand osteoporosis on radiographs better but also shed new light on computer-aided osteopenia and osteoporosis diagnosis.

A prototype selection technique based on relative density and density peaks clustering for k nearest neighbor classification

k-nearest neighbor classifier (KNN) is one of the most famous classification models due to its straightforward implementation and an error bounded by twice the Bayes error. However, it usually degrades because of noise and the high cost in computing the distance between different samples. In this context, hybrid prototype selection techniques have been postulated as a good solution and developed. Yet, they have the following issues: (a) adopted edition methods are susceptible to harmful samples around tested samples; (b) they retain too many internal samples, which contributes little to the classification of KNN classifier and (or) leading to the low reduction; (c) they rely on many parameters. The main contributions of our work are that (a) a novel competitive hybrid prototype selection technique based on relative density and density peaks clustering (PST-RD-DP) are proposed against the above issues at the same time; (b) a new edition method based on relative density and distance (EMRDD) in PST-RD-DP is first proposed to remove harmful samples and smooth the class boundary; (c) a new condensing method based on relative density and density peaks clustering (CMRDDPC) in PST-RD-DP is second proposed to retain representative borderline samples. Intensive experiments prove that PST-RD-DP outperforms 6 popular hybrid prototype selection techniques on extensive real data sets in weighing accuracy and reduction of the KNN classifier. Besides, the running time of PST-RD-DP is also acceptable.

An art painting style explainable classifier grounded on logical and commonsense reasoning

AbstractThis paper presents the art painting style explainable classifier named ANYXI. The classifier is based on art specialists’ knowledge of art styles and human-understandable color traits. ANYXI overcomes the principal flaws in the few art painting style classifiers in the literature. In this way, we first propose, using the art specialists’ studies, categorizations of the Baroque, Impressionism, and Post-Impressionism. Second, we carry out a human survey with the aim of validating the appropriateness of the color features used in the categorizations for human understanding. Then, we analyze and discuss the accuracy and interpretability of the ANYXI classifier. The study ends with an evaluation of the rationality of explanations automatically generated by ANYXI. We enrich the discussion and empirical validation of ANYXI by considering a quantitative and qualitative comparison versus other explainable classifiers. The reported results show how ANYXI is outstanding from the point of view of interpretability while keeping high accuracy (comparable to non-explainable classifiers). Moreover, automated generations are endowed with a good level of rationality.

Identifying Effective Feature Selection Methods for Alzheimer's Disease Biomarker Gene Detection Using Machine Learning.

Alzheimer's disease (AD) is a complex genetic disorder that affects the brain and has been the focus of many bioinformatics research studies. The primary objective of these studies is to identify and classify genes involved in the progression of AD and to explore the function of these risk genes in the disease process. The aim of this research is to identify the most effective model for detecting biomarker genes associated with AD using several feature selection methods. We compared the efficiency of feature selection methods with an SVM classifier, including mRMR, CFS, the Chi-Square Test, F-score, and GA. We calculated the accuracy of the SVM classifier using validation methods such as 10-fold cross-validation. We applied these feature selection methods with SVM to a benchmark AD gene expression dataset consisting of 696 samples and 200 genes. The results indicate that the mRMR and F-score feature selection methods with SVM classifier achieved a high accuracy of around 84%, with a number of genes between 20 and 40. Furthermore, the mRMR and F-score feature selection methods with SVM classifier outperformed the GA, Chi-Square Test, and CFS methods. Overall, these findings suggest that the mRMR and F-score feature selection methods with SVM classifier are effective in identifying biomarker genes related to AD and could potentially lead to more accurate diagnosis and treatment of the disease.

Building an Effective Classifier for Phishing Web Pages Detection: A Quantum-Inspired Biomimetic Paradigm Suitable for Big Data Analytics of Cyber Attacks.

To combat malicious domains, which serve as a key platform for a wide range of attacks, domain name service (DNS) data provide rich traces of Internet activities and are a powerful resource. This paper presents new research that proposes a model for finding malicious domains by passively analyzing DNS data. The proposed model builds a real-time, accurate, middleweight, and fast classifier by combining a genetic algorithm for selecting DNS data features with a two-step quantum ant colony optimization (QABC) algorithm for classification. The modified two-step QABC classifier uses K-means instead of random initialization to place food sources. In order to overcome ABCs poor exploitation abilities and its convergence speed, this paper utilizes the metaheuristic QABC algorithm for global optimization problems inspired by quantum physics concepts. The use of the Hadoop framework and a hybrid machine learning approach (K-mean and QABC) to deal with the large size of uniform resource locator (URL) data is one of the main contributions of this paper. The major point is that blacklists, heavyweight classifiers (those that use more features), and lightweight classifiers (those that use fewer features and consume the features from the browser) may all be improved with the use of the suggested machine learning method. The results showed that the suggested model could work with more than 96.6% accuracy for more than 10 million query-answer pairs.

Automated classification of pediatric age from 12-lead ECG using deep residual convolutional neural networks

With increased interest in screening of young people for potential causes of sudden death, accurate automated detection of ventricular pre-excitation (VPE) or Wolff–Parkinson–White syndrome (WPW) in the pediatric resting ECG is important. Several recent studies have shown interobserver variability when reading screening ECGs and thus an accurate automated reading for this potential cause of sudden death is critical. We designed and tested an automated algorithm to detect pediatric VPE optimized for low prevalence.Digital ECGs with 12 leads or 15 leads (12-lead plus V3R, V4R and V7) were selected from multiple hospitals and separated into a testing and training database. Inclusion criterion was age less than 16 years. The reference for algorithm detection of VPE was cardiologist annotation of VPE for each ECG. The training database (n = 772) consisted of VPE ECGs (n = 37), normal ECGs (n = 492) and a high concentration of conduction defects, RBBB (n = 232) and LBBB (n = 11). The testing database was a random sample (n = 763). All ECGs were analyzed with the Philips DXL ECG Analysis algorithm for basic waveform measurements. Additional ECG features specific to VPE, mainly delta wave scoring, were calculated from the basic measurements and the average beat. A classifier based on decision tree bootstrap aggregation (tree bagger) was trained in multiple steps to select the number of decision trees and the 10 best features. The classifier accuracy was measured on the test database.The new algorithm detected pediatric VPE with a sensitivity of 78%, a specificity of 99.9%, a positive predictive value of 88% and negative predictive value of 99.7%.This new algorithm for detection of pediatric VPE performs well with a reasonable positive and negative predictive value despite the low prevalence in the general population.

Vibration Signal for Bearing Fault Detection using Random Forest

Based on the chosen properties of an induction motor, a random forest (RF) classifier, a machine learning technique, is examined in this study for bearing failure detection. A time-varying actual dataset with four distinct bearing states was used to evaluate the suggested methodology. The primary objective of this research is to evaluate the bearing defect detection accuracy of the RF classifier. First, run four loops that cycle over each feature of the data frame corresponding to the daytime index to determine the bearing states. There were 465 repetitions of the inner race fault and the roller element fault in test 1, 218 repetitions of the outer race fault in test 2, and 6324 repetitions of the outer race in test 3. Secondly, the task is to find the data for the typical bearing data procedure to differentiate between normal and erroneous data. Out of 3 tests, (22-23) % normal data was obtained since every bearing beginning to degrade usually exhibits some form of a spike in many locations, or the bearing is not operating at its optimum speed. Thirdly, to display and comprehend the data in a 2D and 3D environment, Principal Component Analysis (PCA) is performed. Fourth, the RF algorithm classifier recognized the data frame’s actual predictions, which were 99% correct for normal bearings, 97% accurate for outer races, 94% accurate for inner races, and 97% accurate for roller element faults. It is thus concluded that the proposed algorithm is capable to identify the bearing faults.