Search

This study presents a method for predicting location classes of a room such as a kitchen, and restroom, where a user is located by discovering location-specific sensor data motifs in sensor data observed by user’s sensor devices, such as smartwatch, without requiring labeled training data collected in a target environment. For example, we can observe similar waveforms corresponding to kitchen knife chopping actions using body-worn accelerometers in kitchens and can also observe similar sound features by active sound probing in bathrooms because of their water-resistant walls. This indicates that such location-specific sensor data motifs can be inherent information for location class prediction in almost every environment. This study proposes a novel method that automatically detects location-specific motifs from time series sensor data by calculating a score that represents the “location specificity” of each motif in a time series. Previous studies on location class prediction assume that location-specific sensor data are always observed in a room or use handcrafted rules and templates to detect location-specific sensor data resulting in difficulties in applying them to several realistic environments. In contrast, our method, named IndoLabel, can automatically discover short sensor data motifs, specific to a location class, and can automatically build an environment-independent location classifier without requiring handcrafted rules and templates. The proposed method was evaluated in real house environments using leave-one-environment-out cross-validation and achieved a state-of-the-art performance although labeled training data in the target environment was unavailable.

OBJECTIVE Subjective memory decline (SMD) has been identified as a potential early marker of nonnormal and accelerated cognitive decline. We performed data-driven analyses that integrated trajectory classification with prediction modeling to test declining trajectory class prediction by SMD facets, pulse pressure (PP; i.e., a robust proxy for vascular health), and sex. METHOD The longitudinal design featured memory trajectories across a 40-year band (55-95 years) of nondemented aging (N = 580; Mage = 70.2 years; 65% female) from the Victoria Longitudinal Study. First, latent class growth analyses identified distinct classes of memory trajectories. Second, we used the three-step method (R3STEP) to predict membership in the declining memory classes using six measures: memory complaints, memory concerns, memory compensation, memory self-efficacy, PP, and sex. RESULTS First, we identified four classes of memory aging trajectories: (a) stable memory aging (STABLE), (b) typical memory aging (TYPICAL), (c) slowly declining memory aging (SLOW), and (d) rapidly declining memory aging (RAPID). Second, more memory concerns predicted membership in the SLOW and RAPID classes. Higher PP predicted membership in the SLOW class. Male sex predicted membership in the declining (TYPICAL, SLOW, RAPID) classes. CONCLUSION Among SMD facets, memory concerns represent the most severe degree of apprehension about subjectively experienced memory losses. The present integrative data-driven analysis indicated that such concerns predicted membership in declining memory trajectory classes in addition to worse vascular health (higher PP) and sex (male). In nondemented aging, concerns about increasing memory failures may be veridical indicators of memory loss, especially when coupled with vascular comorbidity and being male. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

COVID-19 is an infectious disease, which has adversely affected public health and the economy across the world. On account of the highly infectious nature of the disease, rapid automated diagnosis of COVID-19 is urgently needed. A few recent findings suggest that chest X-rays and CT scans can be used by machine learning for the diagnosis of COVID-19. Herein, we employed semi-supervised learning (SSL) approaches to detect COVID-19 cases accurately by analyzing digital chest X-rays and CT scans. On a relatively small COVID-19 radiography dataset, which contains only 219 COVID-19 positive images, 1341 normal and 1345 viral pneumonia images, our algorithm, COVIDCon, which takes advantage of data augmentation, consistency regularization, and multicontrastive learning, attains 97.07% average class prediction accuracy, with 1000 labeled images, which is 7.65% better than the next best SSL method, virtual adversarial training. COVIDCon performs even better on a larger COVID-19 CT Scan dataset that contains 82,767 images. It achieved an excellent accuracy of 99.13%, at 20,000 labels, which is 6.45% better than the next best pseudo-labeling approach. COVIDCon outperforms other state-of-the-art algorithms at every label that we have investigated. These results demonstrate COVIDCon as the benchmark SSL algorithm for potential diagnosis of COVID-19 from chest X-rays and CT-Scans. Furthermore, COVIDCon performs exceptionally well in identifying COVID-19 positive cases from a completely unseen repository with a confirmed COVID-19 case history. COVIDCon, may provide a fast, accurate, and reliable method for screening COVID-19 patients.

The constant evolution of the illicit drug market makes the identification of unknown compounds problematic. Obtaining certified reference materials for a broad array of new analogues can be difficult and cost prohibitive. Machine learning provides a promising avenue to putatively identify a compound before confirmation against a standard. In this study, machine learning approaches were used to develop class prediction and retention time prediction models. The developed class prediction model used a Naïve Bayes architecture to classify opioids as belonging to either the fentanyl analogues, AH series or U series, with an accuracy of 89.5%. The model was most accurate for the fentanyl analogues, most likely due to their greater number in the training data. This classification model can provide guidance to an analyst when determining a suspected structure. A retention time prediction model was also trained for a wide array of synthetic opioids. This model utilised Gaussian Process Regression to predict the retention time of analytes based on multiple generated molecular features with 79.7% of the samples predicted within ± 0.1 min of their experimental retention time. Once the suspected structure of an unknown compound is determined, molecular features can be generated and input for the prediction model to compare with experimental retention time. The incorporation of machine learning prediction models into a compound identification workflow can assist putative identifications with greater confidence and ultimately save time and money in the purchase and/or production of superfluous certified reference materials.

Structure elucidation of chemical compounds is a complex and challenging activity that requires expertise and well-suited tools. To assign the molecular structure of a given compound, 13C NMR is one of the most widely used techniques because of its broad range of structural information. Taking into account that molecules found in nature can be grouped into natural product (NP) classes because of structural similarities, we explore the possibility of NP class prediction via 13C NMR data. Employing freely available 13C NMR data of NPs, we trained four classifiers for the prediction of eight common NP classes. The best performance was obtained with the XGBoost classifier reaching f1-scores of above 0.82. We also performed experiments with different percentages of positive samples, including the glycoside presence. Furthermore, we tested cases outside the data set, yielding performances above 80% for most classes. For the chromans case, we restricted the test examples to the coumarin subclass, and the prediction accuracy increased to 100%.

Antibiotic resistance in bacteria limits the effect of corresponding antibiotics, and the classification of antibiotic resistance genes (ARGs) is important for the treatment of bacterial infections and for understanding the dynamics of microbial communities. Although several methods have been developed to classify ARGs, none of them work well when the ARGs diverge from those in the reference ARG databases. We develop a novel method, ARG-SHINE, for ARG classification. ARG-SHINE utilizes state-of-the-art learning to rank machine learning approach to ensemble three component methods with different features, including sequence homology, protein domain/family/motif and raw amino acid sequences for the deep convolutional neural network. Compared with other methods, ARG-SHINE achieves better performance on two benchmark datasets in terms of accuracy, macro-average f1-score and weighted-average f1-score. ARG-SHINE is used to classify newly discovered ARGs through functional screening and achieves high prediction accuracy. ARG-SHINE is freely available at https://github.com/ziyewang/ARG_SHINE.

Simultaneous use of the drugs may lead to undesirable Drug-Drug Interactions (DDIs) in the human body. Many DDIs are associated with changes in drug metabolism that performed by Drug-Metabolizing Enzymes (DMEs). In this case, DDI manifests itself as a result of the effect of one drug on the biotransformation of other drug(s), its slowing down (in the case of inhibiting DME) or acceleration (in case of induction of DME), which leads to a change in the pharmacological effect of the drugs combination. We used OpeRational ClassificAtion (ORCA) system for categorizing DDIs. ORCA divides DDIs into five classes: contraindicated (class 1), provisionally contraindicated (class 2), conditional (class 3), minimal risk (class 4), no interaction (class 5). We collected a training set consisting of several thousands of drug pairs. Algorithm of PASS program was used for the first, second and third classes DDI prediction. Chemical descriptors called PoSMNA (Pairs of Substances Multilevel Neighbourhoods of Atoms) were developed and implemented in PASS software to describe in a machine-readable format drug substances pairs instead of the single molecules. The average accuracy of DDI class prediction is about 0.84. A freely available web resource for DDI prediction was developed (http://way2drug.com/ddi/).

Seq2Enz method is a new way to identify whether a query protein sequence is an enzyme and to assign an enzyme class to the protein sequence. The method is based on mask BLAST fortified with some novel structural-chemical properties (NCL) of the building blocks of proteins. All available reviewed enyme sequences (267,276 in number) in Uniprot/SwissProt and most recent depositions (7062) not used for training in ECPred, a state of the art software for enzyme class prediction, are taken for assessment and the results are compared with those from conventional BLAST and ECPred respectively. Seq2Enz is seen to perform consistently better for all the enzyme classes to all the four levels. Seq2Enz methodology is converted into an easy to use web-server and made freely accessible at http://www.scfbio-iitd.res.in Seq2Enz/.

Logistic regression models are a popular and effective method to predict the probability of categorical response data. However, inference for these models can become computationally prohibitive for large datasets. Here we adapt ideas from symbolic data analysis to summarize the collection of predictor variables into histogram form, and perform inference on this summary dataset. We develop ideas based on composite likelihoods to derive an efficient one-versus-rest approximate composite likelihood model for histogram-based random variables, constructed from low-dimensional marginal histograms obtained from the full histogram. We demonstrate that this procedure can achieve comparable classification rates to the standard full data multinomial analysis and against state-of-the-art subsampling algorithms for logistic regression, but at a substantially lower computational cost. Performance is explored through simulated examples, and analyses of large supersymmetry and satellite crop classification datasets. Supplementary materials for this article are available online.

In this paper, we have implemented an efficient and novel technique for multi-label class prediction using associative rule mining. Many of the research works for the classification have been carried out on single-label datasets, but it is not useful for all real-world application accounting to multi-label datasets like scene classification, text categorization, etc. Hence, we propose an algorithm for performing multi-label classification and solve the problems which come across in the domain pertaining to single-label classification. Our novel technique (ARM-MLC) will aim in enhancing the accuracy of any decision-making processes. Here, in multi-label classification, based on our work, we aim to predict the multiple characters of the instances.

Domain adaptation aims at learning a predictive model that can generalize to a new target domain different from the source (training) domain. To mitigate the domain gap, adversarial training has been developed to learn domain invariant representations. State-of-the-art methods further make use of pseudo labels generated by the source domain classifier to match conditional feature distributions between the source and target domains. However, if the target domain is more complex than the source domain, the pseudo labels are unreliable to characterize the class-conditional structure of the target domain data, undermining prediction performance. To resolve this issue, we propose a Pairwise Similarity Regularization (PSR) approach that exploits cluster structures of the target domain data and minimizes the divergence between the pairwise similarity of clustering partition and that of pseudo predictions. Therefore, PSR guarantees that two target instances in the same cluster have the same class prediction and thus eliminate the negative effect of unreliable pseudo labels. Extensive experimental results show that our PSR method significantly boosts the current adversarial domain adaptation methods by a large margin on four visual benchmarks. In particular, PSR achieves a remarkable improvement of more than 5% over the state-of-the-art on several hard-to-transfer tasks.

Power system is one of the most fundamental infrastructure systems for human society. The stability of power system greatly affects the economy and security of the city. Therefore, recovering the power system in cities after a severe disaster is the high priority work in a disaster emergency process. Among several kinds of huge disasters, typhoon is a highly valuable research target. Different from the earthquake or tsunami, typhoons often last several days before landing after their generation, so there is much more time for people to prepare and cope with the damage to power system. In this paper, a damage class prediction method for determining the command institution is researched. Based on the actual power system damage data in China from 2004 to 2019, this research proposed the prediction method with machine learning algorithm. The result shows SVM model can reach 77.4 % accuracy in this task, and it is accurate enough to support the emergency decision.

In this manuscript, a deep learning based approach has been investigated for land-cover classification capable of handling situations where there is an imbalance in the class-wise samples in the training set under a novel semi-supervised learning framework. This problem is persistent when the presence of some of the major land-cover classes affects the other recessive classes in a particular region and due to this the generated training set contains a fewer number of samples from the latter group of classes. Here, an adversarial auto-encoder has been used to generate synthetic samples from each of the minority classes so that each of the classes is well represented in the training set. A new training set has then been designed by including the ‘most confident’ artificial samples from the minority classes. This newly formed balanced training set is then shown to be more effective to classify the test samples as compared to the initially available imbalanced training set. The validation of the proposed approach has been performed using patterns collected from two multi-spectral satellite images captured by Ikonos-2 and landsat-8 satellites over different regions of India. The results show significant improvement in test class prediction when compared to that of other state-of-the-art imbalance handling schemes in land-cover classification.

MotivationAntibiotic resistance is a growing public health problem, which affects millions of people worldwide, and if left unchecked is expected to upend many aspects of healthcare as it is practiced today. Identifying the type of antibiotic resistant genes in genome and metagenomic sample is of utmost importance in the prevention, diagnosis, and treatment of infections. Today there are multiple tools available that predict antibiotic resistance class from DNA and protein sequences, yet there is a lack of benchmarks on the performances of these tools.ResultsWe have developed a dataset that is curated from 15 available databases, and annotated with their antibiotic class labels. We also developed a transfer learning approach with neural networks, TRAC, that outperforms existing antiobiotic resistance prediction tools. While TRAC provides the current state-of-the-art performance, we hope our newly developed dataset will also provide the community with a much needed standardized dataset to develop novel methods that can predict antibiotic resistance class with superior prediction performance.AvailabilityTRAC is available at github (https://github.com/nafizh/TRAC) and the datasets are available at figshare (https://doi.org/10.6084/m9.figshare.11413302).Contactmhamid@mgh.harvard.edu, idoerg@iastate.edu

This paper presents a study of associative music classification with sequential patterns. The analysis focuses on class prediction from multiple patterns, considering aggregation of pattern-level measures and evaluation of pattern-set measures. The study complements recent work in music classification which has employed unsupervised pattern discovery followed by instance-based classification.