Reliable Model Research Articles

Deep Active Learning-Enabled Cost-Effective Electricity Theft Detection in Smart Grids

In industrial informatics-enabled smart grids, machine learning approaches have exhibited high potential in data-driven electricity theft detection (ETD), whereas none of the existing studies pay sufficient attention to the high costs of manually labeling massive sensing data during learning data preparation. To address this defect, this paper develops a cost-effective data-driven ETD approach that significantly reduces the data labeling costs without sacrificing the reliability of ETD. Specifically, the approach is systematically realized via an intelligent deep active learning (DAL) scheme. By seamlessly incorporating convolutional neural network (CNN) learning with Monte Carlo dropout-based Bayesian active query, the DAL scheme efficiently selects the most valuable instances for ETD model training. In this way, the proposed approach is able to derive a reliable CNN-based ETD model with limited labeled learning instances, thus largely reducing the data labeling costs. Experimental test results on an actual ETD dataset provided by the State Grid Corporation of China extensively illustrate the efficacy of the proposed approach.

From animal testing to in vitro systems: advancing standardization in microphysiological systems.

Limitations with cell cultures and experimental animal-based studies have had the scientific and industrial communities searching for new approaches that can provide reliable human models for applications such as drug development, toxicological assessment, and in vitro pre-clinical evaluation. This has resulted in the development of microfluidic-based cultures that may better represent organs and organ systems in vivo than conventional monolayer cell cultures. Although there is considerable interest from industry and regulatory bodies in this technology, several challenges need to be addressed for it to reach its full potential. Among those is a lack of guidelines and standards. Therefore, a multidisciplinary team of stakeholders was formed, with members from the US Food and Drug Administration (FDA), the National Institute of Standards and Technology (NIST), European Union, academia, and industry, to provide a framework for future development of guidelines/standards governing engineering concepts of organ-on-a-chip models. The result of this work is presented here for interested parties, stakeholders, and other standards development organizations (SDOs) to foster further discussion and enhance the impact and benefits of these efforts.

Use of comprehensive datasets to estimate the Slovak National Transportation Model

Developing accurate and reliable transportation models capable of simulating real traffic behaviour is essential to making sound engineering and investment decisions. However, the construction of transport models is typically data- and time-intensive, opaque, and the results are not reproducible. This paper describes the development of the passenger component of the new macroscopic four-step National Transport Model (NTM) for Slovakia. Efforts were made to limit the number of dimensions of the model as far as possible, beginning with one demand stratum and increasing complexity only with improved model fit. Traditional data sets, such as national registries, censuses, and travel use surveys, were compared to novel data, such as public transport ticket sales, mobile phone GPS signals (call detail records), and open-access data sets for cross-border travel. Finally, validation against several existing models was conducted. The result is a passenger traffic model with 25 demand strata, 109 domestic zones, and 68 international zones. This is a significant simplification relative to the previous National Model. The aim of the new NTM and this paper is to increase transparency and reproducibility of transport models.

Establishment and Validation of a Mouse Bacterial Vaginosis Model.

Bacterial vaginosis (BV) is a common vaginal infection without a reliable animal model. To establish a novel mouse BV model, we evaluated multiple parameters of various identified bacteria-infected mice, including Staphylococcus aureus (SA), Escherichia coli (EC), Streptococcus agalactiae, β-Hemolytic streptococcus, and Gardnerella vaginalis (GV). Mature female KM mice were randomly allocated to a vehicle group (group A, without any treatment) and experimental groups. After vaginal secretions were harvested, experimental groups were divided into phosphate buffer solution group (PBS, group B), control group including SA, and EC with a 1:1 ratio (group C), SA, EC, and Streptococcus agalactiae with a 1:2:1 ratio group (group D), SA, EC, and β-Hemolytic streptococcus with a 1:2:1 ratio group (group E), and GV group (group F). The vaginal secretions of experimental mice were collected by flushing with 100 mL sterile PBS on days 2, 4, and 6. Vaginal secretions were examined by Gram staining, sialidase assay, ammonia test, and pH value measurement. IL-6 and IL-10 levels in mouse serum were detected by enzyme-linked immunosorbent assay. Hematoxylineosin staining and mouse cervicovaginal tissue histopathological scores were observed. The diagnostic test results were analyzed by logistic regression analysis and receiver operating characteristic curves. The Shapiro-Wilk analysis of variance, or rank-sum test, was used for normal distribution analysis. Pearson's correlation and chi-squared test determined the correlation and comparison data expressed as a percentage or frequency. There was less severe vaginal morphology in GV-infected mice compared to other bacteria-infected mice. The sialidase assay, the ammonia test, and the pH values of vaginal secretions showed significant differences between GV-infected and uninfected mice. Serum IL-6 and IL-10 levels and vaginal histological scoring increased in other bacteria-infected mice, but GV-infected mice showed only a mildly increasing trend of IL-10 levels and vaginal histological scoring compared to control mice. GV-infected mice showed clinical features similar to human BV infection, including vaginal anatomical and pathological indices, and biochemical and immune parameters. Serum IL-10 level has potential for use in BV diagnosis.

Azoxymethane/Dextran Sodium Sulfate (AOM/DSS) Model of Colorectal Cancer.

Recent progress in developing new vaccination strategies against cancer requires the production of complex and reliable animal models reflecting the complexity of the tumors with their microenvironment. Mice can be considered a good source due to low cost and ease of being genetically modified, inoculated with tumor cell lines or treated by chemicals to induce different cancers. Despite significant limitations in modeling human cancer complexity, preclinical trials conducted in mice can efficiently contribute to understand molecular mechanisms of cancer, to closely resemble and follow carcinogenesis steps impossible to study into humans, and to test new anticancer therapies. In this chapter, we generally describe the different mouse models developed for cancer vaccines' preclinical trials. A particular focus is dedicated to a chemically-induced colorectal cancer model in use in our laboratories.

Robust Framework for Diagnosing Novel Corona Virus from CT images using Support Vector Binary Classifier

The novel coronavirus (commonly abbreviated as CoVID-19) has emerged as a threat to the entire global civilization and has emerged as one of the most infectious and, at times, deadly viruses. Prompt discovery of this infection can assist medical supervisors in taking preventive actions to control the spread. Usually, radiologists and medical specialists require an average time of ~31 minutes to test the CT images and confirm the infection. A large dataset of more than 1000 patients has been gathered and randomly chosen for this experiment. In this research, a ready-to-deploy computer-aided diagnosis (CADx) to detect COVID-19 infection is introduced. A variety of deep learning architectures have been experimented to discover the most reliable predictive model for the diagnosis. This research uses the Densely Connected Convolution Network (DenseNet-121 architecture) along with a boosting support vector binary classifier to tell the difference between someone who has the coronavirus and someone who is healthy. The combination put forward in this work achieved 93% ± 1.8% accuracy, 94.9% ± 2.6% recall, 98% ± 1.5% precision, and an F1 score of 94% ± 1.7%. The model takes less than 1 second to process one image. On the grounds of the above findings, it can be concluded that the proposed approach can be used to diagnose novel coronavirus infections.

Design, 3D-QSAR, molecular docking, ADMET, molecular dynamics and MM-PBSA simulations for new anti-breast cancer agents

Breast cancer is the most frequent form of malignant tumor in women, and represents a major public health problem due to its high mortality rate. Although a multitude of therapeutic options exist for control of this disease, the emergence of resistance to current pharmaceutical treatments underscores the urgency of developing new anti- breast cancer drugs, with a focus on reducing the adverse effects associated with current therapeutic agents. The present study concerns a new series of (23) compounds based on 1,4-quinone and quinoline derivatives to design candidate drugs against breast cancer. For this purpose, integrated computational techniques were applied, including 3D-QSAR, molecular docking and molecular dynamics simulations (MD). CoMFA and CoMSIA were used to build a robust and highly reliable 3D-QSAR models. To validate the model's predictive capabilities, an external validation was carried out. The results of the best model (CoMSIA/SEA) revealed that electrostatic, steric and hydrogen bond acceptor fields had a significant effect on the anti-breast cancer activity of molecules studied. In addition, evaluation of ADMET properties determined whether these newly designed ligands were likely to be selected as drug-candidates. To confirm the binding stability of the selected ligands to aromatase (3S7S) and validate the molecular docking results, molecular dynamics simulations lasting 100 nanoseconds were performed by calculating RMSD, RMSF, RoG, H-bond, SASA and MM-PBSA parameters. As a result, only one designed compound (ligand 5) emerged as the most promising drug candidate for experimental in vitro and in vivo testing, due to its potential inhibition of breast cancer.

Cycle-Retinex: Unpaired Low-Light Image Enhancement via Retinex-Inline CycleGAN

Low-light image enhancement aims to recover normal-light images from the images captured under dim environments. Most existing methods could just improve the light appearance globally whereas failing to handle other degradation such as dense noise, color offset and extremely low-light. Moreover, unsupervised methods proposed in recent years lack reliable physical model as the basis, thus universality is greatly limited. To address these problems, we propose a novel low-light image enhancement method via Retinex-inline cycle-consistent generative adversarial network named Cycle-Retinex, whose training is totally dependent on unpaired datasets. Specifically, we organically combine Retinex theory with CycleGAN, by which we decouple low-light image enhancement task into two sub-tasks, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> illumination map enhancement and reflectance map restoration. Retinex theory helps CycleGAN simplify low-light image enhancement problem and CycleGAN provides synthetic paired images to guide the training of Retinex decomposition network. We further introduce a self-augmented method to address the color distortion and noise problem, thus making the network learn to enhance low-light images adaptively. Extensive experiments show that the proposed method can achieve promising results. The source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/mummmml/Cycle-Retinex</uri> .

Deep-Learning based Melanoma Detection using Cloud Approach

The aim of computer vision techniquesand deep learning in the era of digitalization is to derive valuable insights from them and generate novel understanding. This makes it possible to employ imaging to quickly diagnose and treat a variety of diseases. In the field of dermatology, deep neural networks are utilized to differentiate between images of melanoma and non-melanoma skin lesions. In this paper, we have emphasised two important aspects of melanoma detection research. The accuracy of classifiers is the first thing to take into account, even with very little modifications to the dataset's characteristics there will be a lot of difference in accuracy. We investigated transfer learning issues in this case. We propose that continual training-test iterations are necessary to create reliable prediction models based on the results of the initial study.The second argument is the need for a system with a flexible design that can accommodate changes to training datasets.Our proposal for creating and implementing a melanoma detection service that utilizes clinical and thermoscopic images involves the development and implementing a hybrid architecture that fuses fog, edge and cloud computing. In addition, this design should aim to decrease the duration of the ongoing retraining process, which is necessary to accommodate the large volume of data that requires evaluation. This notion has been reinforced by experiments using a single computer and a variety of distribution techniques, which show how a dispersed strategy ensures output attainment in a noticeably more sufficient amount of time.

Reliable AI models can reveal key processes of heat recovery steam generator operation in air pollutant emission

Artificial intelligence (AI) algorithms have been applied to air pollution prediction modeling, but assessment of their trustworthiness is lacking. This study aims to examine whether AI-based models can identify key operational processes of a Heat Recovery Steam Generator (HRSG) facility and predict the emission of target air pollutants, including nitrogen oxides (NOx), sulfur oxides (SOx), and total suspended particles (TSP). First, key operational variables are selected based on the feature importance of the Random Forest models (that is, data-driven input selection). Secondly, Bidirectional Long Short-Term Memory-based AutoEncoder (BiLSTM-AE) and Random Forest (RF) models with the data-driven input selection are trained and evaluated using multiple predictive performance metrics. Then, the BiLSTM-AE and RF models are trained and evaluated using the operational variables selected by experts. Results demonstrate that the data-driven and expertise-based methods select five common operational variables that are key factors to predict variations of target output variables. The results from the multi-metric evaluations show a skillful prediction of BiLSTM-AE and RF models for NOx and TSP, but not SOx. For TSP prediction, RF is more sensitive to the input selection method than BiLSTM-AE. This study underscores the dependency of the prediction skill of AI algorithm-based models on the target air pollutant and input selection method, and suggests the trustworthiness of selected key operational variables by RF's feature importance.

Atomic structure and water arrangement on K-feldspar microcline (001).

The properties of clouds, such as their reflectivity or their likelihood to precipitate, depend on whether the cloud droplets are liquid or frozen. Thus, understanding the ice nucleation mechanisms is essential for the development of reliable climate models. Most ice nucleation in the atmosphere is heterogeneous, i.e., caused by ice nucleating particles such as mineral dusts or organic aerosols. In this regard, K-feldspar minerals have attracted great interest recently as they have been identified as one of the most important ice nucleating particles under mixed-phase cloud conditions. The mechanism by which feldspar minerals facilitate ice nucleation remains, however, elusive. Here, we present atomic force microscopy (AFM) experiments on microcline (001) performed in an ultrahigh vacuum and at the solid-water interface together with density functional theory (DFT) and molecular dynamics (MD) calculations. Our ultrahigh vacuum data reveal features consistent with a hydroxyl-terminated surface. This finding suggests that water in the residual gas readily reacts with the surface. Indeed, the corresponding DFT calculations confirm a dissociative water adsorption. Three-dimensional AFM measurements performed at the mineral-water interface unravel a layered hydration structure with two features per surface unit cell. A comparison with MD calculations suggests that the structure observed in AFM corresponds to the second hydration layer rather than the first water layer. In agreement with previous computation results, no ice-like structure is seen, questioning an explanation of the ice nucleation ability by lattice match. Our results provide an atomic-scale benchmark for the clean and water-covered microcline (001) plane, which is mandatory for understanding the ice nucleation mechanism on feldspar minerals.

A FRACTAL-FRACTIONAL TSUNAMI MODEL CONSIDERING NEAR-SHORE FRACTAL BOUNDARY

Every fluid problem is greatly affected by its boundary conditions, especially the near-shore seabed could produce an irrevocable harm when a tsunami wave is approaching, and a real-life mathematical model could stave off the worst effect. This paper assumes that the unsmooth seabed is a fractal surface, and fractal-fractional governing equations are established according to physical laws in the fractal space. The geometrical potential theory is used to explain the force produced by the wave surface, and Kong-He friction law is applied to further figuring out the local and memory properties of the friction along the fractal boundary. This paper aims at studying tsunami waves in a fractal space, rendering a reliable mathematical model for both prediction of the tsunami motion and the coastal protection.

A Reliable and Robust Deep Learning Model for Effective Recyclable Waste Classification

A Reliable and Robust Deep Learning Model for Effective Recyclable Waste Classification

ACCURATE PREDICTION OF DRILL BIT PENETRATION RATE IN ROCK USING SUPERVISED MACHINE LEARNING TECHNIQUES BASE ON LABORATORY TEST DATA

Knowing the rate of penetration of a drill bit in rocks is among the most important parameters in their behaviour measurement. However, the direct measurement of ROP in rocks is a high-cost and time-intensive process. Therefore, obtaining the ROP parameter through a method other than direct measurement can be very useful and effective. Predictive machine learning methods are among the strong and precise techniques for the indirect measurement of ROP. To this end, 492 samples were tested under different UCS, µ, WOB, and ω conditions to obtain the corresponding ROP. To achieve an accurate model, three methods of linear regression analysis, lasso regression, and ridge regression were compared in terms of prediction accuracy. These models were compared through performance criteria of the prediction process and error-based charts. The performance criteria were measured using three measures: mean absolute percentage error, D-squared pinball score, and mean Poisson deviance error. For the MAPE index, the Lasso and Ridge models performed the best with values of 0.2557. Concerning the D2PS index, the linear regression model and Ridge performed better with values of 0.4083 and 0.4025, respectively. Finally, for the MPDE index, the Ridge model provided a more accurate performance with a value of 0.0105. For a better comparison, an objective function was created and calculated by combining these three indicators. The results showed the best rank for the Ridge model with an estimated value of 659.475. Finally, it was concluded that the Ridge model is a reliable and accurate model for predicting the ROP.

Solar photovoltaic cell model optimal parameter identification by using an improved chimp optimization algorithm

Identifying the parameters of solar photovoltaic (PV) cell models accurately and reliably is crucial for simulating, evaluating, and controlling PV systems. For this reason, we present an improved chimp optimization algorithm (IChOA) for the generation of precise and reliable solar PV cell models. As a new and improved version of the standard chimp optimization algorithm (ChOA), IChOA embeds two mutation rules in ChOA that include the elite opposition-based learning and visual search mechanism. The first rule is applied to strengthen global exploration capacity of ChOA, and the second one is utilized to enhance ChOA’s local exploitation ability (convergence accuracy). Based on the six benchmark test functions with different characteristics, the effectiveness of IChOA is evaluated by comparing to other five well-known optimization algorithms. The results suggest that IChOA offers superior performance over other competing algorithms. Finally, IChOA’s performance is confirmed through optimizing parameters for three widely employed mathematical models, specifically the single diode model, the double diode model, and the multi-cell PV modules. The findings prove the excellent performance of the suggested approach.

Using Machine Learning to Predict Heart Failure: A Comparative Analysis of Various Classification Algorithms

This study presents a comprehensive analysis of machine learning algorithms for predicting heart failure, a significant cause of morbidity and mortality worldwide. Employing a robust dataset of patient records with diverse clinical features, the performance of several widely-used classification algorithms, including K-Nearest Neighbours (KNN), Decision Trees, Support Vector Machines (SVM), Random Forests (RF), and Gaussian Naive Bayes etc. are systematically evaluated and compared. The methodology encompasses loading the dataset, data pre-processing, feature selection, model training, and validation. The performance of each algorithm is rigorously assessed using metrics such as accuracy, precision, recall, F1-score, and the area under the ROC curve with Random Forest outperforming the other classification algorithms. The study further investigates the influence of hyper-parameter tuning and cross-validation on model efficacy. Additionally, an interpretative analysis of the models, offering insights into feature importance and the clinical relevance of the prediction outcomes were provided. The findings aim to contribute to the field of medical informatics by identifying the most effective machine learning strategies for heart failure prediction, thereby facilitating early intervention and personalized patient care. This research not only underscores the potential of machine learning in healthcare but also highlights the challenges and considerations in developing reliable predictive models for complex medical conditions.

Analytical model of reliability of the freight rail transportation forecasting system

The purpose of the study: to develop an analytical model of the reliability of the freight transportation forecasting system by railway transport. Methods: economic and mathematical methods and models of transportation forecasting adapted to the peculiarities of railway freight transportation –models based on “Bottom/Top” approaches, methods of predictive adjustment by types of cargo, methods of scenario forecasting. Results: an analytical model of the reliability of the freight rail transportation forecasting system has been developed, the distinctive feature of which is the integration of cargo transportation evaluation indicators, characteristics of scenario methods and parameters of the forecasting process, a forecast has been formed for the volume and structure of freight traffic following through the Oktyabrskaya railway landfill for the future up to 2030 in the category of main types of cargo.

A New Early Warning Method of Train Tracking Interval Based on CTC

For railway operations worldwide, train rear-end collision accidents occasionally occur, resulting in heavy casualties and property loss. To avoid the occurrence of such incidents, a novel early warning method of train tracking interval based on centralized traffic control (CTC), or simply CTC interval prewarning, is proposed. When combined with the existing automatic train protection (ATP), the monitoring of train tracking interval is doubled. The proposed method calculates the minimum safety interval in real time and compares it with the actual spatial distance. When the actual spatial distance is less than the minimum safety interval, particularly when the ATP is at fault, the proposed method is capable of raising the alert and taking control measures to effectively avoid rear-end collision accidents. At the same time, this method displaces the manual supervision of train operation, thereby relieving labor intensity, and improves the effectiveness of supervision. This paper analyzes the necessity and feasibility of the proposed CTC interval prewarning method and explains the basic architecture of the system, data acquisition, tracking interval calculation, prewarning rules, and implementation. The CTC tracking interval calculation model has been established, and the table of warning distance of the different train speeds has been calculated. The reliability model of the train control system (TCS) has been established based on the CTC interval prewarning method, which is used to analyze the reliability and safety of the TCS.

System reliability calculation based on the run-time analysis of ladder program

Programmable logic controller (PLC) system, a typical member in the embedded family, is now widely applied in industry. For safety critical PLC systems, reliability is of top significance. However, due to subcomponents’ temporal correlations caused by the run-time execution of embedded ladder programs, the complexity of reliability analysis is greatly increased. In this paper, we propose a novel probabilistic model to analyze reliability of PLC systems, called run-time reliability model (RRM). RRM is automatically constructed based on the structure and run-time execution of the embedded ladder program. Moreover, it is also a dynamic bayesian network (DBN) capturing full dependencies in a PLC system. Then, according to execution semantics of RRM nodes, we present customized conditional probability distribution (CPD) tables to calculate final reliability of the system, with failure probability of every referenced component as refinement. The strength of this model is that not only does it explicitly specify the correlations between run-time execution of embedded software and system components, but also it serves as a computational mechanism for probabilistic inference. Besides, the proposed approach is superior to previous works in both accuracy and efficiency. Compared to monte carlo based simulation, the average error rate of reliability values inferred from RRM model is small.

FeFET Reliability Modeling for In-Memory Computing: Challenges, Perspective, and Emerging Trends

FeFET Reliability Modeling for In-Memory Computing: Challenges, Perspective, and Emerging Trends