Practical Datasets Research Articles

XAIP: An eXplainable AI‐Based Pipeline for Identifying Key Factors of Surface Defects in Strip Steel

The surface defect has a direct impact on the performance and quality of the final product, so it is crucial to identify the key factors causing the defect. To achieve accurate key factor identification, an eXplainable AI‐based Pipeline (XAIP) is proposed herein. The proposed XAIP combines data mode clustering and local model construction to meet practical application requirements. Meanwhile, after analyzing the characteristics of practical data, a strategy for defect rate calculation is designed to provide more fine‐grained defect‐related information to supervise model training. The final identification results are presented by a two‐stage explainable method, which is designed to reveal the information of data modes and key variables and can give engineers more specific defect‐related information. The experiment results based on two practical manufacturing datasets show the effectiveness of the proposed method.

Machine‐Learning‐Based Accurate Prediction of Vanadium Redox Flow Battery Temperature Rise Under Different Charge–Discharge Conditions

ABSTRACTAccurate prediction of battery temperature rise is very essential for designing efficient thermal management scheme. In this paper, machine learning (ML)‐based prediction of vanadium redox flow battery (VRFB) thermal behavior during charge–discharge operation has been demonstrated for the first time. Considering different currents with a specified electrolyte flow rate, the temperature of a kW scale VRFB system is studied through experiments. Three different ML algorithms; linear regression (LR), support vector regression (SVR), and extreme gradient boost (XGBoost) have been used for prediction. The training and validation of ML algorithms have been done by the practical dataset of a 1 kW 6 kWh VRFB storage under 40 , 45, 50, and 60 A charge–discharge currents and 10 L min−1 of flow rate. A comparative analysis among ML algorithms is done by performance metrics such as correlation coefficient (R2), mean absolute error (MAE), and root mean square error (RMSE). XGBoost shows the highest R2 value of around 0.99, which indicates its higher prediction accuracy compared to other ML algorithms used. The ML‐based prediction results obtained in this work can be very useful for controlling the VRFB temperature rise during operation and act as an indicator toward further development of an optimized thermal management system.

Likelihood-Based Inference of Log-Logistic Accelerated Hazards Regression Models for Cross Hazards Data

ABSTRACT The proportional hazards model and accelerated failure time model are two important classes for analyzing survival data. However, the two models cannot capture time-scaled effects such as crossing hazard or the gradual effect of treatment over time. For the purpose of this detection, accelerated hazards model has been studied. In this paper, we propose the use of log-logistic distribution for the baseline hazard function of the accelerated hazards model which allows for time-dependent group effect. In particular, the log-logistic distribution gives various forms (e.g. non-monotone) of hazard function and also an easy implementation for likelihood-based model fitting due to an explicit form of hazard and survival functions. The likelihood-based estimation procedures of the accelerated hazards models are derived. Our method is demonstrated with two practical data sets, via the procedures of systematic model checking.

Development of a neural network with a layer of trainable activation functions for the second stage of the ensemble classifier with stacking

One of the promising directions for improving the quality of object recognition in images and parallelizing calculations is the use of ensemble classifiers with stacking. A neural network at the second level makes it possible to achieve the resulting quality of classification, which is significantly higher than each of the networks of the first level separately. The classification quality of the entire ensemble classifier with stacking depends on the efficiency of the neural networks at the first stage, their number, and the quality of the classification of the neural network of the second stage. This paper proposes a neural network architecture for the second stage of the ensemble classifier, which combines the approximating properties of traditional neurons and learning activation functions. Gaussian Radial Basis Functions (RBFs) were chosen to implement the learned activation functions, which are summed with the learned weights. The experimental studies showed that when working with the CIFAR-10 data set, the best results are obtained when six RBFs are used. A comparison with the use of multilayer perceptron (MLP) in the second stage showed a reduction in classification errors by 0.45–1.9 % depending on the number of neural networks in the first stage. At the same time, the proposed neural network architecture for the second degree had 1.69–3.7 times less learning coefficients than MLP. This result is explained by the fact that the use of an output layer with ordinary neurons allowed us not to enter into the architecture many learning activation functions for each output signal of the first stage, but to limit ourselves to only one. Since the results were obtained on the CIFAR-10 universal data set, a similar effect could be obtained on a large number of similar practical data sets.

Bivariate cubic normal distribution for non-Gaussian problems

Probabilistic models play critical role in various engineering fields. Numerous critical issues exist in probabilistic modeling, especially for non-Gaussian correlated random variables. However, conventional parameter-based bivariate distribution models are generally developed for specific types of random variables, limiting their flexibility and applicability. A flexible bivariate distribution model is proposed, in which the joint cumulative distribution function is derived with the probability equivalently expressed as the summation of three basic probabilities corresponding to simple functions. These three basic probabilities are calculated with the aid of univariate cubic normal distribution, and thus the proposed model is named as bivariate cubic normal (BCN) distribution. The proposed BCN distribution has been applied in modeling several common bivariate distributions and actual engineering datasets. Results show that the BCN model accurately models both typical theoretical distributions and the JCDF of practical datasets, offering a significant improvement over existing models. Furthermore, the proposed BCN distribution has been utilized in seismic reliability assessment and the calculation of the mean recurrence interval and hazard curve of hurricane wind speed and storm size. Results demonstrate that the BCN distribution excels in modeling and matching capabilities, proving its accuracy and effectiveness in civil engineering applications.

An inverse lomax-uniform poisson distribution and joint modeling of repeatedly measured and time-to-event data in the health sectors.

Methodological developments in different sectors like health, biomedical and biological areas are the recent burning issue in the statistical literature. The approach of implementing declining hazard function which is obtained by compounding truncated Poisson distribution and a lifetime distribution is a special concern in a few studies. In this paper we proposed a newly introduced distribution called inverse Lomax-Uniform Poisson distribution mostly applied in the health, biomedical, biological, and related sectors. Some basic statistical properties of the distribution are discussed. The capability of the model is checked by comparing it with six potential models by using a practical real data set. Based on the comparison techniques, the newly proposed model out performs all its counterparts. The simulation study is also conducted. Furthermore, the joint modelling of repeatedly measured and time-to-vent processes is discussed in detail with the real data set in the health sector.

Creating a Cleaned Dataset for Ease of Use

ObjectiveThe Welsh Longitudinal General Practice (WLGP) dataset contains over 4 billion records. Due to the size and the need to filter the dataset to get the necessary general practice interaction results, query performance is poor. To overcome this, a ‘cleaned’ version of the dataset needed to be created. ApproachAn R package was created that would be run when a new refresh of the WLGP data is provided. Two new tables would be created based on the original WLGP Events table. ResultsA WLGP Cleaned Dataset R Package was produced and creates a reformatted events table and events look up table. The reformatted GP events table reduces the original events table from 14 columns to 8 key columns. The events look up table takes distinct events from the original WLGP events table and produces a new table including columns such as the event description, type and hierarchy levels. Both tables are then linked using an event code ID. ConclusionThe R package now creates a new reformatted events table as well as an events look up table which is run after a WLGP refresh is provided. The new tables are then provisioned to any projects that have access to the dataset. ImplicationsThe new tables has improved query performance for our researchers, providing them with a table with all the necessary information and an easy way to decipher events codes. This has improved the amount of time researchers have spent manipulating and querying the tables.

The European Union Settlement Scheme Data Linkage Project in Wales: Findings relating to mental health and education

Objective Following the UK’s exit from the European Union (EU) a settlement scheme was introduced for EU nationals currently living in the UK to apply to continue living in the UK. The European Union (EU) Settlement Scheme project compares the living experiences of EU and UK citizens in Wales. ApproachWe conducted a retrospective population cohort study within the Secure Anonymised information Linkage (SAIL) Databank to explore the mental health, educational experiences and involvement in the Welsh labour market of EU citizens living in Wales in comparison to British citizens. UK Census data relating to country of birth, sex, week of birth, date of arrival in the UK and socio-economic status have been linked to variables of relevance in the Education for Wales and Welsh General Practice datasets respectively and analysed. ResultsFindings indicate statistically significant differences in secondary school attendance and attainment between Welsh born pupils and pupils born in EU countries. We will present findings on our exploration if differences exist between Welsh born and EU country born individuals in relation to diagnosis of depression and anxiety. Additional findings from this linking and analysis reveal other interesting differences between EU and UK born citizens. ConclusionsLinking data in this way has gained a better understanding of the experiences and outcomes of EU citizens in Wales, generating better evidence to inform policies and services that address the needs of this population and a dataset of interest to academics. Reproduceable analytical pipelines have been created for other researchers to use.

A novel model for count data: zero-inflated Probit Bell model with applications

A novel model for count data is proposed in this work. This new model is briefly called the Zero-Inflated Probit Bell (ZIPBell) model. The ZIPBell model can be used to simulate and analyze traditional count data and can be used well in the simulation and analysis of count data with the phenomenon that the frequency for zero occupies a very high percentage. In the scope of this study, we first provide the general formula, functions or related equations for the ZIPBell model. The results of asymptotic inferences for the ZIPBell model are also given in this work. Maximum likelihood estimator (MLE) is used to check its performance of the proposed model. According to the numerical illustrations via several simulation researches and the analysis of a practical data set, it has been seen that the MLE method has offered correct and reliable inferences. Finally, the inferences and conclusions of this study are mentioned.

Research on Molten Iron Quality Prediction Based on Machine Learning

The quality of molten iron not only has a significant impact on the strength, toughness, smelting cost and service life of cast iron but also directly affects the satisfaction of users. The establishment of timely and accurate blast furnace molten iron quality prediction models is of great significance for the improvement of the production efficiency of blast furnace. In this paper, Si, S and P content in molten iron is taken as the important index to measure the quality of molten iron, and the 989 sets of production data from a No.1 blast furnace from August to October 2020 are selected as the experimental data source, predicting the quality of molten iron by the I-GWO-CNN-BiLSTM model. First of all, on the basis of the traditional data processing method, the missing data values are classified into correlation data, temporal data, periodic data and manual input data, and random forest, the Lagrangian interpolation method, the KNN algorithm and the SVD algorithm are used to complete them, so as to obtain a more practical data set. Secondly, CNN and BiLSTM models are integrated and I-GWO optimized hyperparameters are used to form the I-GWO-CNN-BiLSTM model, which is used to predict Si, S and P content in molten iron. Then, it is concluded that using the I-GWO-CNN-BiLSTM model to predict the molten iron quality can obtain high prediction accuracy, which can provide data support for the regulation of blast furnace parameters. Finally, the MCMC algorithm is used to analyze the influence of the input variables on the Si, S and P content in molten iron, which helps the steel staff control the quality of molten iron in a timely manner, which is conducive to the smooth running of blast furnace production.

A channel resilient RFF extraction scheme for cyclic prefix contained systems

Radio frequency fingerprint (RFF) based identification technique has been proved efficient for ensuring the validity of devices connected to network. However, it is still a tough task to extract robust RFF in the scenarios with multi-path channel and moving terminals. To solve this problem, this paper proposes a channel-resilient RFF extraction scheme which can effectively reduce the influences from complex channel condition and retain robust device fingerprint. In the proposed system, blind synchronization and symbol-scale carrier frequency offset (CFO) estimation are designed for signal preprocessing for preparations of the following RFF extraction. A cyclic-prefix based de-channel algorithm (CPDCA) which can effectively weaken channel interference is proposed to meet the channel robustness of our system. Additionally, symbol-scale feature stacking algorithm (SFSA) is applied for RFF denoising, which can further enhance the performance of proposed system. Experiments using practical dataset collected from Long Term Evolution (LTE)-V2X communication system has been carried out under different signal-to-noise ratio (SNR). The results demonstrate that the proposed scheme has the ability to extract channel-robust RFF and to achieve reliable classification performance under complex channel conditions.

Technique for Kernel Matching Pursuit Based on Intuitionistic Fuzzy c-Means Clustering

Kernel matching pursuit (KMP) requires every step of the searching process to be global optimal searching in the redundant dictionary of functions in order to select the best matching signal structure. Namely, the dictionary learning time of KMP is too long. To solve the above drawbacks, a rough dataset was divided into some small-sized dictionaries to substitute local searching for global searching by using the property superiority of dynamic clustering performance, which is also superior in the intuitionistic fuzzy c-means (IFCM) algorithm. Then, we proposed a novel technique for KMP based on IFCM (IFCM-KMP). Subsequently, three tests including classification, effectiveness, and time complexity were carried out on four practical sample datasets, the conclusions of which fully demonstrate that the IFCM-KMP algorithm is superior to FCM and KMP.

Micro-expression recognition based on a novel GCN-transformer cooperation model for IoT-eHealth

If a person is truly healthy, his/her well-being encompasses both physical and psychological health. However, the existing IoT-eHealth system typically focus only on monitoring the user’s physical data through various sensors, neglecting their mental state. To enhance the intelligence level of IoT-eHealth system and enable it to have the psychological monitoring ability, a novel collaborative model based on Graph Convolutional Network (GCN) and Transformer is designed for Micro-Expression (ME) recognition in this paper. Firstly, facial information within each frame is transformed into a Spatial Topological Relationship Graph (STRG) by using facial landmarks detection and psychological relationship of local patches. Then, in order to automatically aggregate the key information on facial patches that contribute to ME recognition from the structured graph data, a Hierarchical Adaptive Graph Pooling (HAGP) module is designed for obtaining discriminative frame-level feature based on GCN utilizing graph structure and vertex global dependencies. Finally, in order to model the long-term dependencies among frames and capture the key frame-level features that are beneficial for ME recognition, a Temporal Sensitive Self-Attention (TSSA) mechanism is designed, and a novel Temporal Sensitive Transformer (TST) encoder is constructed based on TSSA to explore the evolution law of facial patterns and obtain discriminative video-level features that are helpful for ME recognition. In the comparative experiments of standard dataset verification and practical dataset testing, designed collaborative model is superior to other methods and can achieve the highest recognition accuracy, which almost can meet the application requirements of IoT-eHealth system.

Fusion of multi-resolution data for estimating speed-density relationships

Estimating traffic flow models, such as speed-density relationships, using data from multiple sources with different temporal resolutions is a prevalent challenge encountered in real-world scenarios. The resolution incompatibility is often intuitively addressed by averaging the high-resolution (HR) data to synchronize with the low-resolution (LR) data. This paper shows that ignoring the variability of HR data within the LR interval during the averaging process could lead to systematic data point distortions, resulting in biased model estimations. The average absolute biases of models estimated from the average data increase with the lost variability of HR data within the LR intervals. Subsequently, it proves that for any given complete average data dataset, there must exist an optimal dataset that minimizes the average absolute bias in model estimations introduced by the averaging process. A novel procedure for determining the practical optimal dataset is proposed. To test the proposed method, real-world HR data from four sites in Hong Kong and Nanjing, China were collected to mimic situations with multi-resolution data. Results demonstrated that the proposed method can significantly reduce the average absolute biases of models estimated from the determined practical optimal dataset, as compared to models estimated from the complete average dataset.

Matrix Separation and Poisson Multi-Bernoulli Mixture Filtering for Extended Multi-Target Tracking with Infrared Images

Multi-target tracking using infrared images is receiving more and more attention. There are many state-of-the-art methods, and the deep learning network and low-rank and sparse matrix separation are two kinds of methods with high accuracy. However, the former suffers from heavy training samples, and the latter requires high-dimensional processing, meaning its computing cost is huge. In this work, a united detection and tracking method with matrix separation and PMBM filtering is proposed. In the detection process, a low-rank and sparse matrix separation algorithm with a differentiable form based on a single image is constructed. In the filtering process, the multi-target state is modeled as a PMBM distribution, which is conjugate in the Bayesian framework. The two processes interact mutually in that the detection provides measurements, and the filtering offers prior information for the next detection to improve accuracy. The computational complexity is given by a theoretical analysis, which shows a significant reduction. The numerical analysis, carried out on a practical dataset, verifies an enhancement in the BSF and SCRG metrics and ROC curves.

Our First Experience of Using Blue Light Laser for Endoscopic Endonasal Removal of Juvenile Angiofibromas

Background. Juvenile nasal angiofibroma (JNA) is a very rare benign lesion originating from the pterygopalatine fossa with distinctive epidemiologic features and growth patterns. The typical patient is an adolescent male with a clinical history of recurrent severe nasal bleedings and blockage the nose. Currently, even though surgery often caused severe operative bleeding, it is considered the ideal treatment for JNA. Refinement in preoperative embolization, which provides significant reduction of intraoperative bleeding. In spite of breakthrough of Endoscopic techniques which minimize the risk of residual disease, the search of new devises which can help to management of JNA is still continue in order to minimize the surgical complications. Laser methods as surgical treatment have been actively used in rhinology during the past decades. We purposed to summarize information about current lasers and their use in rhinology. Aim. The purpose of the study is the literature review and to describe our experience of using new blue laser during JNA removal. Evaluation of advantages and disadvantages, determination of indications and limits of its use in rhinosurgery Methods- A literature review from 2000 to 2022 using the PubMed database was employed. Keywords used included “laser surgery”, “blue light laser”, “photoangiolytic laser”, “laser in rhinosurgery”. The most up to date studies published for each rhinology condition that was treated with laser surgery was reviewed. Then endoscopic removal through the nose was performed. During the surgery TrueBlue laser was used to cut the tissues and coagulate injured vessels. Results. Rhinological conditions appropriate for laser applications are discussed. There are related papers to a number of applications including hereditary hemorrhagic telangiectasia, rhinitis, turbinate surgery, dacryocystorhinostomy, septoplasty, choanal atresia, and sphenopalatine artery ligation, paying attention to the outcomes of the studies. It is the first experience of using blue laser in rhinosurgery and particulary ib removing JNA. Conclusion. Intranasal laser surgery, despite the fact that interventions are performed almost bloodlessly, and often do not require nasal tamponade, indications for their performance are limited, due to deep burns of the nasal mucosa, and alteration of mucociliary clearance therefore lasers are not so often used in rhinology. The short-wave blue laser with a wavelength of 445 nm, used in Europe in oto- and laryngosurgery, may have some potential advantages in rhinosurgery, but the practical dataset are limited yet.

Validating Benfordness on contaminated data

Benford’s law is a mathematical model, very recurrent in practice for a wide variety of datasets, used to represent the frequencies of digits. A well-established usage of Benfordness statistical testing lies within investigations aimed to ascertain if balance sheet and income statement data are genuine. A typical, frustrating problem of Benfordness statistical tests on big, practical datasets is that they often provide p-valuessmaller than expected when the Benfordness null hypothesis is very realistic. A possible reason is that data are contaminated by some kind of noise. In this paper we propose the deconvolution approach to alleviate this issue, using both simulated and real data.

Semi-supervised prototype network based on compact-uniform-sparse representation for rotating machinery few-shot class incremental fault diagnosis

Distinct from the ideal ‘closed-set’ assumption of comprehensive fault types and arbitrary data access, the entangled challenges of labeled instance paucity and fault increasing with operation are difficult to circumvent in industrial diagnostics. However, conventional deep learning-based diagnostic models lack the capacity for dynamic and incremental acceptance of new faults under limited labeled instances. When confronted with this more industrially realistic diagnostic task, their performance is constrained by overfitting, imbalance, and catastrophic forgetting, thus diminish the feasibility of engineering deployment. In this research, we define this task as a few-shot class incremental fault diagnosis (FS-CIFD) problem and develop a semi-supervised prototype network based on compact-uniform-sparse representation (CUS-SSPN) for this purpose. Structurally, CUS-SSPN incorporates the designed compact-uniform-sparse (CUS) composite loss, optimal transport (OT)-based semi-supervised strategy, and multi-level knowledge distillation (KD) strategy into the prototype network. With the CUS composite loss and OT-based semi-supervised strategy, CUS-SSPN can assign representative prototypes for each incremental fault and form an intra-class compact and inter-class uniform and sparse representation space. This allows the model to incrementally learn new faults while effectively suppressing the overfitting triggered by instance paucity and the biased diagnosis triggered by unbalanced labeling. Combining the designed multi-level KD strategy to convey category centroid information and classification decision knowledge from both prototype and logits levels, catastrophic forgetting of known faults can be effectively mitigated. Extensive FS-CIFD tasks are constructed on both publicly available dataset and practical wind turbine dataset to validate the effectiveness of CUS-SSPN and the feasibility of engineering diagnostics.

Telemetry Monitoring System with Features Explaining Anomalies Based on Mahalanobis Distance

Because satellites cannot be repaired once launched, operators must detect anomalies early and prevent failures before they occur. Thus, satellite telemetry monitoring systems need to alert operators of anomalies and provide them with useful information to deal with these anomalies. However, traditional knowledge-based monitoring systems have the challenges of difficulty in building comprehensive models and a high dependency on experts. In recent years, data-driven approaches have been actively studied with the development of machine learning algorithms. These approaches solve the challenges of knowledge-based methods; however, they are often less capable of explaining anomalies than knowledge-based methods. In this study, we propose the new telemetry monitoring system with feature engineering to explain anomalies. The proposed method realizes identifiability of anomaly types and unusual telemetry by designing features based on moving averages, telemetry periods, waveform differences, and the Mahalanobis distance. We applied the proposed features to artificial and practical abnormal datasets and evaluated their usefulness. The results showed that the proposed method is capable of identifying trend, periodic, and waveform anomalies, specifying the telemetry in which the anomaly occurred and providing the information to operators.

Improved sports image classification using deep neural network and novel tuna swarm optimization

Sports image classification is a complex undertaking that necessitates the utilization of precise and robust techniques to differentiate between various sports activities. This study introduces a novel approach that combines the deep neural network (DNN) with a modified metaheuristic algorithm known as novel tuna swarm optimization (NTSO) for the purpose of sports image classification. The DNN is a potent technique capable of extracting high-level features from raw images, while the NTSO algorithm optimizes the hyperparameters of the DNN, including the number of layers, neurons, and activation functions. Through the application of NTSO to the DNN, a finely-tuned network is developed, exhibiting exceptional performance in sports image classification. Rigorous experiments have been conducted on an extensive dataset of sports images, and the obtained results have been compared against other state-of-the-art methods, including Attention-based graph convolution-guided third-order hourglass network (AGTH-Net), particle swarm optimization algorithm (PSO), YOLOv5 backbone and SPD-Conv, and Depth Learning (DL). According to a fivefold cross-validation technique, the DNN/NTSO model provided remarkable precision, recall, and F1-score results: 97.665 ± 0.352%, 95.400 ± 0.374%, and 0.8787 ± 0.0031, respectively. Detailed comparisons reveal the DNN/NTSO model's superiority toward various performance metrics, solidifying its standing as a top choice for sports image classification tasks. Based on the practical dataset, the DNN/NTSO model has been successfully evaluated in real-world scenarios, showcasing its resilience and flexibility in various sports categories. Its capacity to uphold precision in dynamic settings, where elements like lighting, backdrop, and motion blur are prominent, highlights its utility. The model's scalability and efficiency in analyzing images from live sports competitions additionally validate its suitability for integration into real-time sports analytics and media platforms. This research not only confirms the theoretical superiority of the DNN/NTSO model but also its pragmatic effectiveness in a wide array of demanding sports image classification assignments.