Search

The tandem metal inert gas (MIG) process uses two wires that are continuously fed through a special welding torch and disbursed to form a single molten pool. Within the contact tip of the modern approach, the wires are electrically insulated from one another. This study identified the effect of welding electrode spacing on the distortion of AA5052 aluminum plates and different mechanical properties including hardness and thermal cycle using grey relational analysis. Plate distortion was subsequently predicted using the grey prediction model GM (1, 6). This research used a pair of 400 mm × 75 mm × 5 mm of AA5052 plates and electrode distances of 18, 27, and 36 mm. The welding current, voltage, welding speed, and argon flow rate were 130 A, 23 V, 7 mm/s, and 17 L/min, respectively. The temperature was measured using a type-K thermocouple at 10, 20, 30, and 40 mm from the center of the weld bead. The smallest distortion at an electrode distance of 27 mm was 1.4 mm. At an electrode distance of 27 mm, the plate may reach a proper peak temperature where the amount of heat input and dissipation rate are similar to those for electrode distances of 18 mm and 36 mm. The highest relative VHN of 57 was found in the BM, while the lowest, 46, was found in the WM, showing good agreement with their respective grain sizes. Six parameters were designed using grey relational analysis (GRA) and subsequently employed in the grey prediction model GM (1, 6). Process evaluation results show that predictions for welding distortions are consistent with actual results, thus, the GM (1, 6) model can be used as a predictive model for welding distortions of 5052 aluminum plates. [ABSTRACT FROM AUTHOR]

Frequent pattern mining (FIM) identifies the most important patterns in data sets. However, due to the huge and high-dimensional nature of transactional data, classical pattern mining techniques suffer from the limitations of dimensions and data annotations. Recently, data mining while preserving privacy is considered as an important research area. Information privacy is a tradeoff that must be considered when using data. Through many years, privacy-preserving data mining (PPDM) made use of methods that are mostly based on heuristics. The operation of deletion was used to hide the sensitive information in PPDM. In this study, we used deep active learning to protect private and sensitive information. This paper combines entropy-based active learning with an attention-based approach to effectively hide sensitive patterns. The constructed models are then validated using high-dimensional transactional data with attention-based and active learning methods in a reinforcement environment. The results show that the proposed model can support and improve the effectiveness of decision-making by increasing the number of training instances through the use of a pooling technique and an entropy uncertainty measure. The proposed paradigm can achieve data sanitization by the hiding sensitive items and avoiding to hide the non-sensitive items. The model outperforms greedy, genetic, and particle swarm optimization approaches. [ABSTRACT FROM AUTHOR]

Early prediction of students' learning performance and analysis of student behavior in a virtual learning environment (VLE) are crucial to minimize the high failure rate in online courses during the COVID-19 pandemic. Nevertheless, traditional machine learning models fail to predict student performance in the early weeks due to the lack of students' activities' data in a week-wise timely manner (i.e., spatiotemporal feature issues). Furthermore, the imbalanced data distribution in the VLE impacts the prediction model performance. Thus, there are severe challenges in handling spatiotemporal features, imbalanced data sets, and a lack of explainability for enhancing the confidence of the prediction system. Therefore, an intelligent framework for explainable student performance prediction (ESPP) is proposed in this study in order to provide the interpretability of the prediction results. First, this framework utilized a time-series weekly student activity data set and dealt with the VLE imbalanced data distribution using a hybrid data sampling method. Then, a combination of convolutional neural network (CNN) and long short-term memory (LSTM) was employed to extract the spatiotemporal features and develop the early prediction deep learning (DL) model. Finally, the DL model was explained by visualizing and analyzing typical predictions, students' activities' maps, and feature importance. The numerical results of cross-validation showed that the proposed new DL model (i.e., the combined CNN-LSTM and ConvLSTM), in the early prediction cases, performed better than the baseline models of LSTM, support vector machine (SVM), and logistic regression (LR) models. [ABSTRACT FROM AUTHOR]

Recent advances in deep learning have shown many successful stories in smart healthcare applications with data-driven insight into improving clinical institutions' quality of care. Excellent deep learning models are heavily data-driven. The more data trained, the more robust and more generalizable the performance of the deep learning model. However, pooling the medical data into centralized storage to train a robust deep learning model faces privacy, ownership, and strict regulation challenges. Federated learning resolves the previous challenges with a shared global deep learning model using a central aggregator server. At the same time, patient data remain with the local party, maintaining data anonymity and security. In this study, first, we provide a comprehensive, up-to-date review of research employing federated learning in healthcare applications. Second, we evaluate a set of recent challenges from a data-centric perspective in federated learning, such as data partitioning characteristics, data distributions, data protection mechanisms, and benchmark datasets. Finally, we point out several potential challenges and future research directions in healthcare applications. [ABSTRACT FROM AUTHOR]

As the number of Internet of massive Things (IoT) applications grows in vehicles, factory machinery, smart buildings, and city infrastructure, a secure and automated solution of enabling a mesh network for transactional processes is a significant demand. However, the trust evaluation among those unknown IoT devices which communicate to each other are still the high demand in distributed massive IoT network. In this paper, an online cognitive authentication and trust evaluation decision application programming interface (API) for Cognitive Security Gateway (CSG) based on distributed massive IoT network is proposed. The online cognitive security APIs consisting of cognitive authentication sub‐API, cognitive trust evaluation sub‐API and cognitive authorization sub‐API are designed for distributed massive IoT applications. The modified electronic product code (EPC) Class1Gen2 tags generated by CSG server is and proposed to be the online identity for the system used. Moreover, both the security analyses and trust evaluations for the CSG are discussed. Finally, the proposed system is shown that it is an appropriate solution for solving the online authentication problem in distributed cognitive IoT network. [ABSTRACT FROM AUTHOR]

Abstract Background The symptom of tongue deviation is observed in a stroke or transient ischemic attack. Nevertheless, there is much room for the interpretation of the tongue deviation test. The crucial factor is the lack of an effective quantification method of tongue deviation. If we can quantify the features of the tongue deviation and scientifically verify the relationship between the deviation angle and a stroke, the information provided by the tongue will be helpful in recognizing a warning of a stroke. Methods In this study, a quantification method of the tongue deviation angle was proposed for the first time to characterize stroke patients. We captured the tongue images of stroke patients (15 males and 10 females, ranging between 55 and 82 years of age); transient ischemic attack (TIA) patients (16 males and 9 females, ranging between 53 and 79 years of age); and normal subjects (14 males and 11 females, ranging between 52 and 80 years of age) to analyze whether the method is effective. In addition, we used the receiver operating characteristic curve (ROC) for the sensitivity analysis, and determined the threshold value of the tongue deviation angle for the warning sign of a stroke. Results The means and standard deviations of the tongue deviation angles of the stroke, TIA, and normal groups were: 6.9 ± 3.1, 4.9 ± 2.1 and 1.4 ± 0.8 degrees, respectively. Analyzed by the unpaired Student’s t-test, the p-value between the stroke group and the TIA group was 0.015 (>0.01), indicating no significant difference in the tongue deviation angle. The p-values between the stroke group and the normal group, as well as between the TIA group and the normal group were both less than 0.01. These results show the significant differences in the tongue deviation angle between the patient groups (stroke and TIA patients) and the normal group. These results also imply that the tongue deviation angle can effectively identify the patient group (stroke and TIA patients) and the normal group. With respect to the visual examination, 40% and 32% of stroke patients, 24% and 16% of TIA patients, and 4% and 0% of normal subjects were found to have tongue deviations when physicians “A” and “B” examined them. The variation showed the essentiality of the quantification method in a clinical setting. In the receiver operating characteristic curve (ROC), the Area Under Curve (AUC, = 0.96) indicates good discrimination. The tongue deviation angle more than the optimum threshold value (= 3.2°) predicts a risk of stroke. Conclusions In summary, we developed an effective quantification method to characterize the tongue deviation angle, and we confirmed the feasibility of recognizing the tongue deviation angle as an early warning sign of an impending stroke.