One-class Model Research Articles

Predicting video engagement using heterogeneous DeepWalk

Video engagement is important in online advertisements where there is no physical interaction with the consumer. Engagement can be directly measured as the number of seconds after which a consumer skips an advertisement. In this paper, we propose a model to predict video engagement of an advertisement using only a few samples. This allows for early identification of poor quality videos. This can also help identify advertisement frauds where a robot runs fake videos behind the name of well-known brands. We leverage on the fact that videos with high engagement have similar viewing patterns over time. Hence, we can create a similarity network of videos and use a graph-embedding model called DeepWalk to cluster videos into significant communities. The learned embedding is able to identify viewing patterns of fraud and popular videos. In order to assess the impact of a video, we also consider how the view counts increase or decrease over time. This results in a heterogeneous graph where an edge indicates similar video engagement or history of view counts between two videos. Since it is difficult to find labelled samples for ‘fraud’ video, we leverage on a one-class model that can determine ‘fraud’ videos with outlier or abnormal behavior. The proposed model outperforms baselines in F-measure by over 20%.

Classification of Breast Cancer and Breast Neoplasm Scenarios Based on Machine Learning and Sequence Features from lncRNAs-miRNAs-Diseases Associations.

The influence of non-coding RNAs, such as lncRNAs (long non-coding RNAs) and miRNAs (microRNAs), is undeniable in several diseases, for example, in the formation of neoplasms and cancer scenarios. However, there are challenges due to the scarcity of validated datasets and the imbalance in the data. We found that the research of associations between miRNAs-lncRNAs and diseases is limited or done separately. In addition, those investigations, which use Machine Learning models joined with genomic sequence features extracted from miRNAs and lncRNAs, are few compared with using some methods such as genomic expression or Deep Learning techniques. In this paper, we propose a structure of using supervised and unsupervised machine learning models with genomic sequence features, such as k-mers, sequence alignments, and energy folding values, to validate miRNAs and lncRNAs association with breast cancer and neoplasms scenarios. Using One-Class SVM for outlier detection and comparing two supervised models such as SVM and Random Forest, we manage to obtain accuracy results of 95.44% for the One-class model, with 88.79% and 99.65% for the SVM and Random Forest models, respectively. The results showed a promising path for the study of sequence features interactions joined with Machine Learning models comparable to those found in the existing literature.

1H NMR spectroscopy, one-class classification and outlier diagnosis: A powerful combination for adulteration detection in paprika powder

Paprika powder is a widely consumed high-priced product and therefore particularly prone to fraudulent practices. Nuclear magnetic resonance spectroscopy combined with one-class classification was applied for non-targeted detection of paprika adulteration and further chemometric tools were tested for diagnostic purposes. The 1H NMR spectra of 186 commercial paprika powders and 216 spiked samples were used to develop and comprehensively validate a data-driven soft independent modelling of class analogy model. The established one-class model yielded 92% sensitivity and exemplary adulterants were detected with 100% specificity at concentration levels of 0.1%, 0.1%, 10% and 20% by weight for Azorubine, Ponceau 4R, beetroot and sumac powder, respectively. After successful classification, visualization tools of robust principal component analysis and orthogonal partial least squares analysis were explored to uncover fingerprints of unusual (atypical and spiked) paprika powders. One-class classifiers based on 1H NMR spectroscopic data seem to be suitable for adulteration screening and, in combination with an outlier diagnosis, may improve prioritization of suspicious products for further confirmatory analysis in food authentication.

Improved One-Class Modeling of High-Dimensional Metabolomics Data via Eigenvalue-Shrinkage.

One-class modelling is a useful approach in metabolomics for the untargeted detection of abnormal metabolite profiles, when information from a set of reference observations is available to model “normal” or baseline metabolite profiles. Such outlying profiles are typically identified by comparing the distance between an observation and the reference class to a critical limit. Often, multivariate distance measures such as the Mahalanobis distance () or principal component-based measures are used. These approaches, however, are either not applicable to untargeted metabolomics data, or their results are unreliable. In this paper, five distance measures for one-class modeling in untargeted metabolites are proposed. They are based on a combination of the and five so-called eigenvalue-shrinkage estimators of the covariance matrix of the reference class. A simple cross-validation procedure is proposed to set the critical limit for outlier detection. Simulation studies are used to identify which distance measure provides the best performance for one-class modeling, in terms of type I error and power to identify abnormal metabolite profiles. Empirical evidence demonstrates that this method has better type I error (false positive rate) and improved outlier detection power than the standard (principal component-based) one-class models. The method is illustrated by its application to liquid chromatography coupled to mass spectrometry (LC-MS) and nuclear magnetic response spectroscopy (NMR) untargeted metabolomics data from two studies on food safety assessment and diagnosis of rare diseases, respectively.

One-class models for validation of miRNAs and ERBB2 gene interactions based on sequence features for breast cancer scenarios

One challenge in miRNA–genes–diseases interaction studies is that it is challenging to find labeled data that indicate a positive or negative relationship between miRNA and genes. The use of one-class classification methods shows a promising path for validating them. We have applied two one-class classification methods, Isolation Forest and One-class SVM, to validate miRNAs interactions with the ERBB2 gene present in breast cancer scenarios using features extracted via sequence-binding. We found that the One-class SVM outperforms the Isolation Forest model, with values of sensitivity of 80.49% and a specificity of 86.49% showing results that are comparable to previous studies. Additionally, we have demonstrated that the use of features extracted from a sequence-based approach (considering miRNA and gene sequence binding characteristics) and one-class models have proven to be a feasible method for validating these genetic molecule interactions.

Viral Pneumonia Screening on Chest X-Rays Using Confidence-Aware Anomaly Detection.

Clusters of viral pneumonia occurrences over a short period may be a harbinger of an outbreak or pandemic. Rapid and accurate detection of viral pneumonia using chest X-rays can be of significant value for large-scale screening and epidemic prevention, particularly when other more sophisticated imaging modalities are not readily accessible. However, the emergence of novel mutated viruses causes a substantial dataset shift, which can greatly limit the performance of classification-based approaches. In this paper, we formulate the task of differentiating viral pneumonia from non-viral pneumonia and healthy controls into a one-class classification-based anomaly detection problem. We therefore propose the confidence-aware anomaly detection (CAAD) model, which consists of a shared feature extractor, an anomaly detection module, and a confidence prediction module. If the anomaly score produced by the anomaly detection module is large enough, or the confidence score estimated by the confidence prediction module is small enough, the input will be accepted as an anomaly case (i.e., viral pneumonia). The major advantage of our approach over binary classification is that we avoid modeling individual viral pneumonia classes explicitly and treat all known viral pneumonia cases as anomalies to improve the one-class model. The proposed model outperforms binary classification models on the clinical X-VIRAL dataset that contains 5,977 viral pneumonia (no COVID-19) cases, 37,393 non-viral pneumonia or healthy cases. Moreover, when directly testing on the X-COVID dataset that contains 106 COVID-19 cases and 107 normal controls without any fine-tuning, our model achieves an AUC of 83.61% and sensitivity of 71.70%, which is comparable to the performance of radiologists reported in the literature.

Cleaner and faster method to detect adulteration in cassava starch using Raman spectroscopy and one-class support vector machine

Due to food adulteration concerns, analytical assays are routinely performed in labs to evaluate and ensure food quality control. However, classical analytical methods used to acquire reliable results are lengthy and costly. Therefore, we aim to propose a new approach to detect adulterants in cassava starch in a clean, green, cheap, and quick way. Raman spectroscopy meets all these requirements and presents great potential to perform such routine analyses. Data treatment is also an important step in authentication problems, and we propose the use of one-class models to do so. One-class support vector machine (OC-SVM) and soft independent modelling by class analogy (SIMCA) were the two approaches to one-class classifiers assessed in this study. Cassava starch samples were modified in the lab with adulteration ranging from 0.5 to 50%, with adulterants such as wheat flour, sodium bicarbonate, and others. The two chemometric models were statistically compared and OC-SVM was found to outperform SIMCA, reaching higher values of sensitivity (87.1%), specificity (86.8%), and accuracy (86.9%) in the prediction of known data samples. This better performance also resulted in the possibility of detecting adulterations over 2% by OC-SVM, compared to only 5% by SIMCA.

Network Anomaly Detection Using Memory-Augmented Deep Autoencoder

In recent years, attacks on network environments continue to rapidly advance and are increasingly intelligent. Accordingly, it is evident that there are limitations in existing signature-based intrusion detection systems. In particular, for novel attacks such as Advanced Persistent Threat (APT), signature patterns have problems with poor generalization performance. Furthermore, in a network environment, attack samples are rarely collected compared to normal samples, creating the problem of imbalanced data. Anomaly detection using an autoencoder has been widely studied in this environment, and learning is through semi-supervised learning methods to overcome these problems. This approach is based on the assumption that reconstruction errors for samples that are not used for training will be large, but an autoencoder is often over-generalized and this assumption is often broken. In this paper, we propose a network intrusion detection method using a memory-augmented deep auto-encoder (MemAE) that can solve the over-generalization problem of autoencoders. The MemAE model is trained to reconstruct the input of an abnormal sample that is close to a normal sample, which solves the generalization problem for such abnormal samples. Experiments were conducted on the NSL-KDD, UNSW-NB15, and CICIDS 2017 datasets, and it was confirmed that the proposed method is better than other one-class models.

ROC curves for the optimization of one-class model parameters. A case study: Authenticating extra virgin olive oil from a Catalan protected designation of origin

This paper proposes a ROC curve-based methodology to find optimal classification model parameters. ROC curves are implemented to set the optimal number of PCs to build a one-class SIMCA model and to set the threshold class value that optimizes both the sensitivity and specificity of the model.The authentication of the geographical origin of extra-virgin olive oils of Arbequina botanical variety is presented. The model was developed for samples from Les Garrigues, target class, Samples from Siurana were used as the non-target class.Samples were measured by FT-Raman with no pretreatment. PCA was used as exploratory technique. Spectra underwent pre-treatment and variables were selected based on their VIP score values.ROC curve and others already known criteria were applied to set the threshold class value. The results were better when the ROC curve was used, obtaining performance values higher than 82%, 75% and 77% for sensitivity, specificity and efficiency, respectively.

Handling imbalanced medical image data: A deep-learning-based one-class classification approach.

In clinical settings, a lot of medical image datasets suffer from the imbalance problem which hampers the detection of outliers (rare health care events), as most classification methods assume an equal occurrence of classes. In this way, identifying outliers in imbalanced datasets has become a crucial issue. To help address this challenge, one-class classification, which focuses on learning a model using samples from only a single given class, has attracted increasing attention. Previous one-class modeling usually uses feature mapping or feature fitting to enforce the feature learning process. However, these methods are limited for medical images which usually have complex features. In this paper, a novel method is proposed to enable deep learning models to optimally learn single-class-relevant inherent imaging features by leveraging the concept of imaging complexity. We investigate and compare the effects of simple but effective perturbing operations applied to images to capture imaging complexity and to enhance feature learning. Extensive experiments are performed on four clinical datasets to show that the proposed method outperforms four state-of-the-art methods.

A one-class classification decision tree based on kernel density estimation

One-class Classification (OCC) is an important field of machine learning which aims at predicting a single class on the basis of its lonely representatives and potentially some additional counter-examples. OCC is thus opposed to traditional classification problems involving two or more classes, and addresses the issue of class unbalance. There is a wide range of one-class models which give satisfaction in terms of performance. But at the time of explainable artificial intelligence, there is an increasing need for interpretable models. The present work advocates a novel one-class model which tackles this challenge. Within a greedy and recursive approach, our proposal for an explainable One-Class decision Tree (OC-Tree) rests on kernel density estimation to split a data subset on the basis of one or several intervals of interest. Thus, the OC-Tree encloses data within hyper-rectangles of interest which can be described by a set of rules. Against state-of-the-art methods such as Cluster Support Vector Data Description (ClusterSVDD), One-Class Support Vector Machine (OCSVM) and isolation Forest (iForest), the OC-Tree performs favorably on a range of benchmark datasets. Furthermore, we propose a real medical application for which the OC-Tree has demonstrated effectiveness, through the ability to tackle interpretable medical diagnosis aid based on unbalanced datasets.

Untargeted identification of adulterated Sanqi powder by near-infrared spectroscopy and one-class model

Sanqi is a widely used traditional Chinese medicines (TCM) for its outstanding efficacy. In Chinese market, Sanqi powder is the goal of counterfeiting for a long time. Investigation of Sanqi authenticity is very important in both economic and public health terms. The present work aims at studying the feasibility of combining near-infrared (NIR) spectroscopy with relief-based variable selection and class-modeling for identifying adulterated Sanqi powder. A total of 209 samples including pure and mixed samples, were prepared. Principal component analysis (PCA) was applied for exploratory analysis. The relief algorithm was used to rank all variables, based on which only the first 100 most informative variables were picked out for subsequent class-modeling. By optimizing the parameters such as the number of components, type I and type II errors, the final one-class models were constructed on the training set and evaluated on the test set. Such a procedure is simple and is more in line with actual need. The performance of the models is acceptable. The results indicate that NIR spectroscopy combined with class-modeling and relief-based variable selection is feasible for identifying the adulteration of Sanqi powder.

Towards explaining anomalies: A deep Taylor decomposition of one-class models

Detecting anomalies in the data is a common machine learning task, with numerous applications in the sciences and industry. In practice, it is not always sufficient to reach high detection accuracy, one would also like to be able to understand why a given data point has been predicted to be anomalous. We propose a principled approach for one-class SVMs (OC-SVM), that draws on the novel insight that these models can be rewritten as distance/pooling neural networks. This ‘neuralization’ step lets us apply deep Taylor decomposition (DTD), a methodology that leverages the model structure in order to quickly and reliably explain decisions in terms of input features. The proposed method (called ‘OC-DTD’) is applicable to a number of common distance-based kernel functions, and it outperforms baselines such as sensitivity analysis, distance to nearest neighbor, or edge detection.

Predictive Modeling of Antibiotic Susceptibility in E. Coli Strains Using the U-Net Network and One-Class Classification

The antibiotic resistance of bacterial pathogens has become one of the most serious global health issues due to misusing and overusing of antibiotics. Recently, different technologies were developed to determine bacteria susceptibility towards antibiotics; however, each of these technologies has its advantages and limitations in clinical applications. In this contribution, we aim to assess and automate the detection of bacterial susceptibilities towards three antibiotics; i.e. ciprofloxacin, cefotaxime and piperacillin using a combination of image processing and machine learning algorithms. Therein, microscopic images were collected from different E. coli strains, then the convolutional neural network U-Net was implemented to segment the areas showing bacteria. Subsequently, the encoder part of the trained U-Net was utilized as a feature extractor, and the U-Net bottleneck features were utilized to predict the antibiotic susceptibility of E. coli strains using a one-class support vector machine (OCSVM). This one-class model was always trained on images of untreated controls of each bacterial strain while the image labels of treated bacteria were predicted as control or non-control images. If an image of treated bacteria is predicted as control, we assume that these bacteria resist this antibiotic. In contrast, the sensitive bacteria show different morphology of the control bacteria; therefore, images collected from these treated bacteria are expected to be classified as non-control. Our results showed 83% area under the receiver operating characteristic (ROC) curve when OCSVM models were built using the U-Net bottleneck features of control bacteria images only. Additionally, the mean sensitivities of these one-class models are 91.67% and 86.61% for cefotaxime and piperacillin; respectively. The mean sensitivity for the prediction of ciprofloxacin is only 59.72% as the bacteria morphology was not fully detected by the proposed method.

Paper-based SERS substrate and one-class classifier to monitor thiabendazole residual levels in extracts of mango peels.

The assessment of pesticide residue levels demands fast, low cost and easy-to-use procedures which are not found in conventional methods. In this work, SERS substrates based on the deposition of gold nanoparticles (GNPs) on common office paper were prepared using a wax printer. These substrates combined with Data Driven Soft Independent Modelling of Class Analogies (DD-SIMCA), a one-class classifier algorithm, were used for detection of pesticide residues in water extracts of mango peels. Paper-based substrates made sample collection easier compared with conventional SERS methods, since few microliters of the pesticide aqueous extract from fruit peels needed to be deposited onto the substrate. Moreover, one-class classifiers dismiss the need for quantification or calibration curves. Classification of a fruit with residue levels in accordance to regulatory bodies' limits is based on a mathematical threshold. Just as in an authentication problem, all the possibilities for a given analysed fruit are now restricted to agreeing or not agreeing with current regulations. The performance of the one-class model was demonstrated by detecting thiabendazole (TBZ) residues at various mango samples, with all results being confirmed by HPLC-DAD analysis. The final model could distinguish samples with TBZ levels above the ones allowed by the Brazilian Health Regulatory Agency with 94% of selectivity and 92% of sensitivity, even in the presence of other pesticides.

Host-Response Subphenotypes Offer Prognostic Enrichment in Patients With or at Risk for Acute Respiratory Distress Syndrome.

Classification of patients with acute respiratory distress syndrome into hyper- and hypoinflammatory subphenotypes using plasma biomarkers may facilitate more effective targeted therapy. We examined whether established subphenotypes are present not only in patients with acute respiratory distress syndrome but also in patients at risk for acute respiratory distress syndrome (ARFA) and then assessed the prognostic information of baseline subphenotyping on the evolution of host-response biomarkers and clinical outcomes. Prospective, observational cohort study. Medical ICU at a tertiary academic medical center. Mechanically ventilated patients with acute respiratory distress syndrome or ARFA. None. We performed longitudinal measurements of 10 plasma biomarkers of host injury and inflammation. We applied unsupervised latent class analysis methods utilizing baseline clinical and biomarker variables and demonstrated that two-class models (hyper- vs hypoinflammatory subphenotypes) offered improved fit compared with one-class models in both patients with acute respiratory distress syndrome and ARFA. Baseline assignment to the hyperinflammatory subphenotype (39/104 [38%] acute respiratory distress syndrome and 30/108 [28%] ARFA patients) was associated with higher severity of illness by Sequential Organ Failure Assessment scores and incidence of acute kidney injury in patients with acute respiratory distress syndrome, as well as higher 30-day mortality and longer duration of mechanical ventilation in ARFA patients (p < 0.0001). Hyperinflammatory patients exhibited persistent elevation of biomarkers of innate immunity for up to 2 weeks postintubation. Our results suggest that two distinct subphenotypes are present not only in patients with established acute respiratory distress syndrome but also in patients at risk for its development. Hyperinflammatory classification at baseline is associated with higher severity of illness, worse clinical outcomes, and trajectories of persistently elevated biomarkers of host injury and inflammation during acute critical illness compared with hypoinflammatory patients. Our findings provide strong rationale for examining treatment effect modifications by subphenotypes in randomized clinical trials to inform precision therapeutic approaches in critical care.

Automatic cancer discrimination based on near-infrared spectrum and class-modeling technique

To develop an effective and objective diagnostic method for detecting the malignancy is of great importance. Considering that near-infrared (NIR) spectroscopy has many advantages such as being inexpensive and simple in sample preparation and class-modeling is a rather new strategy, the present paper investigates the feasibility of combining class-modeling technique other than classic classification and NIR spectroscopy for colorectal diagnosis. A total of 162 colorectal tissue slices were prepared and used to collect NIR spectra. A special variable importance (VI) index was defined to pick out 20 most significant variables. The Kennard-Stone (KS) algorithm was used to select representative 57 cancerous samples as the training set for building one-class model and the other samples served as the test set. The results showed that on the independent test set, it can achieve acceptable performance, i.e., the total accuracy of 95.2 %, the sensitivity of 96 %, and the specificity of 94.5 %. It indicates that the combination of NIR spectroscopy and one-class classifier is a potential tool for automatic cancer diagnosis.

From one-class to two-class classification by incorporating expert knowledge: Novelty detection in human behaviour

One-class classification is the standard procedure for novelty detection. Novelty detection aims to identify observations that deviate from a determined normal behaviour. Only instances of one class are known, whereas so called novelties are unlabelled. Traditional novelty detection applies methods from the field of outlier detection. These standard one-class classification approaches have limited performance in many real business cases. The traditional techniques are mainly developed for industrial problems such as machine condition monitoring. When applying these to human behaviour, the performance drops significantly. This paper proposes a method that improves existing approaches by creating semi-synthetic novelties in order to have labelled data for the two classes. Expert knowledge is incorporated in the initial phase of this data generation process. The method was deployed on a real-life test case where the goal was to detect fraudulent subscriptions to a telecom family plan. This research demonstrates that the two-class expert model outperforms a one-class model on the semi-synthetic dataset. In a next step the model was validated on a real dataset. A fraud detection team of the company manually checked the top predicted novelties. The results show that incorporating expert knowledge to transform a one-class problem into a two-class problem is a valuable method.

A Deep One-class Model for Network Anomaly Detection

For traditional network anomaly detection system, the detection performance is related to the selected features and training dataset. But traditional methods adopt handcraft feature selection, which requires heavy human labour and relies on the experts’ knowledge and experience. Besides, the collected dataset for training is not balanced, which makes the prediction of the trained model tends to be biased to the majority class. In this paper, a one-class network anomaly detection model based on the stacked autoencoders was proposed. We use the stacked autoencoders to select the prominent features from the raw collected data, then apply the one-class classification algorithm support vector data description to train a classifier to identify the network traffic into normal data and anomalous data. The experimental results demonstrate the promising results of our approach for network anomaly detection.

Express detection of expired drugs based on near-infrared spectroscopy and chemometrics: A feasibility study.

Levofloxacin is a third-generation fluoroquinolone antimicrobials drug that inhibits bacterial DNA replication. Driven by huge profit, one kind of particular counterfeit, e.g., repackaged expired tablets, becomes very common especially in developing countries. The feasibility of identifying expired levofloxacin tablets by combining NIR spectroscopy with chemometrics was investigated. Five kinds of levofloxacin samples from different manufacturers were collected for experiment. Two types of expired mode were considered and a simple model-independent algorithm was used for feature selection. Principal component analysis (PCA) was used for exploratory analysis and simple discriminant analysis. Taking seventy samples as the target class, a final one-class model based on Data Driven Soft Independent Modeling by Class Analogy with abbreviation DD-SIMCA was constructed, which achieved 97% sensitivity and 100% specificity on the independent set composed of 34 unexpired and 128 expired tablets. These results confirm that the combination of NIR spectroscopy, feature selection and class-modeling is feasible for identifying the expired levofloxacin tablets. Such a method can be extended to the analysis of similar drugs.