Benchmarking Five Machine Learning Models for Accurate Steel Plate Defect Detection

The detection of surface and backside defects in steel plates is essential for safety maintenance. Current ac magnetic flux leakage (ACMFL) testing systems generally operate at a high frequency but cannot detect backside defects owing to the skin effect. Low-frequency ACMFL is limited by low sensitivity and low efficiency in detecting deep defects. This letter presents a multifrequency ACMFL testing method combined with the merits of low- and high-frequency excitation to detect surface and backside defects. In early work, the optimal frequency for detecting the surface defect with a depth of 4 mm or less was 400 Hz. Through simulation analysis, the optimal low frequency for detecting backside defects is determined. Additionally, the influence of the excitation intensity on the response signal is investigated. Relatively small excitation power should be used to obtain a reliable detection waveform. Finally, the experimental results of surface and backside defects detection validated the practicality of the method.

In the evolving landscape of cyber-physical systems (CPS), such as Electric Vehicle Charging Stations (EVCS) and the Industrial Internet of Things (IIoT), the convergence of cyber and physical domains introduces a myriad of opportunities for enhanced efficiency and connectivity. However, this integration also presents substantial security challenges, with vulnerabilities posing risks to both cyber operations and physical system functionality. Anomalies, indicative of cyber-attacks or system malfunctions, manifest as deviations from established operational norms, necessitating sophisticated detection mechanisms. This thesis presents a comprehensive investigation into the application of machine learning (ML) and deep learning (DL) algorithms for anomaly detection within these critical CPS frameworks, emphasizing IIoT and EVCS systems. Acknowledging the inherent complexity of these systems, the research initially applies a suite of ML algorithms—Support Vector Machines (SVM), Decision Trees (DT), and Random Forests (RF)—to IIoT systems, exploiting the relatively straightforward operational patterns to establish a foundational anomaly detection framework. This strategic application leverages the diverse strengths of each algorithm: SVM’s capacity for handling high-dimensional data, DT’s interpretability and ease of use, and RF’s robustness and accuracy in classification tasks. Subsequently, the thesis escalates the analytical depth by incorporating Long Short-Term Memory (LSTM) networks, a DL-based technique, to navigate the more intricate anomaly detection challenges encountered in both IIoT and EVCS systems. LSTM networks are selected for their proven efficacy in processing and making predictions based on long-term dependencies in time-series data, a common characteristic in the operational data of EVCS and IIoT systems. This transition underscores a methodological advancement towards models capable of capturing complex, temporal data relationships, essential for detecting sophisticated anomalies. We carried out ML-based anomaly detection using the WUSTL-IIoT datasets from 2018 and 2021. The ML algorithms underwent extensive training and evaluation, demonstrating substantial effectiveness. Specifically, the SVM and DT models attained an accuracy of 97.6%, with the RF model achieving a slightly superior accuracy of 98.8%. To enhance the detection capability, an LSTM model was implemented, which achieved a remarkable accuracy rate of 99.57%. This performance exemplifies the advanced potential of DL methodologies in navigating the complexities of anomaly detection within intricate data environments. We carried out LSTM-based anomaly detection in EVCS systems by utilizing the CICEVCS 2023 and 2024 datasets. These datasets, encompassing a wide range of attack scenarios along with normal operational data, provided a complex backdrop for the application of the LSTM model. The DL algorithm skillfully navigated these complexities, achieving an impressive accuracy rate of 99.589% in identifying anomalies. This experiment underscores the advanced capabilities of DL, specifically LSTM, in accurately analyzing and predicting anomalies across comprehensive time-series data streams within EVCS systems. The findings from these case studies highlight the pivotal role of ML and DL algorithms in advancing anomaly detection capabilities within IIoT and EVCS systems. By meticulously applying and evaluating SVM, DT, RF, and LSTM models against real-world operational and attack scenarios, this thesis demonstrates the efficacy of these computational techniques in identifying anomalies and enhances the strategic framework for securing CPS against emerging cyber threats.

A retrospective pilot study is conducted to demonstrate the utility of a novel support vector machine learning (SVML) algorithm in a small three-dimensional (3D) sample yielding sparse optical coherence tomography (spOCT) data for the automatic monitoring of neovascular (wet) age-related macular degeneration (wAMD). From the anti-vascular endothelial growth factor injection database, 588 consecutive pairs of OCT volumes (57.624 B-scans) were selected in 70 randomly chosen wAMD patients treated with ranibizumab. The SVML algorithm was applied to 183 OCT volume pairs (17.934 B-scans) in 30 patients. Four independent, diagnosis-blinded retina specialists indicated whether wAMD activity was present between 100 pairs of consecutive OCT volumes (9800 B-scans) in the remaining 40 patients for comparison with the SVML algorithm and a non-complex baseline algorithm using only retinal thickness. The SVML algorithm was assessed using inter-observer variability and receiver operating characteristic (ROC) analyses. The retina specialists showed an average Cohen's κ of 0.57±0.13 (minimum: 0.41, maximum: 0.83). The average κ between the proposed algorithm and the retina specialists was 0.62±0.05 and 0.43±0.14 between the baseline algorithm and the retina specialists. Using each of the four retina specialists as the reference, the proposed method showed a superior area under the ROC curve of 0.91±0.03 compared to the ROC 0.81±0.05 shown by the baseline algorithm. The SVML algorithm was as effective as the retina specialists were in detecting activity in wAMD. Support vector machine learning (SVML) may be a useful monitoring tool in wAMD suited for small samples that yield sparse OCT data possibly derived from self-measuring OCT-robots.

The Support Vector Machine (SVM) learning algorithm has emerged as a robust tool for predicting atmospheric air temperature from historical weather data, showcasing its potential in tackling complex meteorological challenges. Recent trends in environmental predictions emphasize the importance of advanced machine learning techniques over traditional methods, highlighting SVM's ability to handle dynamic atmospheric data and provide more accurate forecasts compared to conventional statistical approaches. By incorporating diverse weather features and real-time data, SVM significantly improves the reliability of temperature predictions, crucial for effective environmental management amidst urbanization-induced climate impacts. The efficacy of SVM extends beyond temperature forecasting to applications like predicting hazardous flight conditions and black ice events, demonstrating high accuracy rates in analyzing meteorological parameters. As SVM evolves and adapts, its role in temperature prediction is expected to expand, enhancing safety and operational efficiency in aviation and other sectors reliant on precise weather forecasting. The advancements in predictive accuracy through SVM represent a pivotal step towards understanding and mitigating the effects of climate change, underscoring its growing significance in environmental modeling and policy-making.

Software defects and quality assurance are crucial aspects of software development that should be considered during the software development cycle. To ensure high-quality software, it is essential to have a robust quality assurance process in place. System reliability and quality are very key components that must be considered during software development, and this can only be achieved when software undergoes a thorough test process for errors, anomalies, defects, omissions, and bugs. Early software defect prediction and detection play an essential role in ensuring the reliability and quality of software systems, ensuring that software companies discover errors or defects early enough and allocate more resources to defect-prone modules. This study proposes the development of an enhanced classifier model for software defect prediction and detection. The aim is to harness the collective intelligence of selected base classifiers like Support Vector Machine, Logistic regression, Decision Trees, Random Forest, AdaBoost, Gradient Boosting, K-Nearest Neighbor, GaussianNB, and Multi-Layer Perception to improve accuracy, robustness, and generalization in identifying potential defects using a soft voting ensemble technique. The ensemble model leveraged the confidence probability of the soft voting technique and the generalization advantage of cross-validation leading to a more robust and dynamic model. The performance of the model with existing classifiers was evaluated using accuracy, F1 score, Precision, and area under the ROC curve (ROC- AUC) as the evaluation metrics. The results of the experiment revealed that the Proposed Classifier produced an overall Accuracy rate of 93%, and ROC AUC of 98%. The results demonstrate the effectiveness of our enhanced ensemble classifier in software defect detection and prediction. By harnessing the strengths of diverse base classifiers, our approach provides a robust and adaptive solution to the challenges of early detection and mitigating defects in software systems. This research contributes to the advancement of reliable software development practices and lays the foundation for future enhancements in ensemble-based defect detection methodologies.

Recent advances in data analysis and wearable sensors for human movement monitoring can promote objective and accurate clinical evaluation of neurological symptoms and improve outcome measures in clinical trials [1–3]. The aim of this study was to combine modern technique of data analysis and wearable sensors to determine which supervised machine learning (ML) algorithm can most accurately classify people with Parkinson’s disease (pwPD) from speed-matched healthy subjects (HS) based on a selected minimum set of IMU-derived gait features. Twenty-two gait features were extrapolated from the trunk acceleration patterns of 81 pwPD and 80 HS, including spatiotemporal, pelvic kinematics, and acceleration-derived gait stability indexes. After a three-level feature selection procedure, seven gait features were considered for implementing five ML algorithms: support vector machine (SVM), artificial neural network, decision trees (DT), random forest (RF), and K-nearest neighbors. Accuracy, precision, recall, F1 score, AUC and generalization error were calculated. SVM outperformed the other ML algorithms in terms of classification metrics (test accuracy = 0.86; F1 score = 0.85; AUC = 0.85) and generalizability (generalization error = 2.95%) in classifying the gait impairment of pwPD compared with speed-matched healthy subjects, using a selected dataset of gait features based on lower trunk acceleration data. Although significantly lower than SVM, tree-based algorithms revealed good classification performances with low generalization errors (RF: test accuracy= 0.86; F1 score = 0.85; AUC = 0.85), and lower computational demand than SVM. ANN was similar to DT in terms of classification metrics but showed significantly higher generalization error (7.26%) than tree-based algorithms and SVM and higher computational demand than the other ML algorithms. Even though KNN showed the fastest time performance, its classification metrics were the lowest. We proposed a feature selection procedure based on the combination of filter, wrapper, embedded, and domain-specific methods that was effective in lowering the risk of overrepresenting multicollinear gait features in the model, resulting in a lower risk of overfitting in the test performances by increasing the explainability of the results at the same time. Because of their accurate results, their simplicity of understanding, and explanability, DT and RF algorithms could represent useful tools for the comprehension of gait disorders by making clinicians participate in the decision process. This is the first time that the accuracy and generalizability of the most performed ML algorithms in classifying pwPD gait abnormalities based on gait data from a single lumbar-mounted IMU have been compared. The findings of this study could be used to incorporate machine learning algorithms into software that processes gait data from lumbar-mounted IMUs. Future research could focus on finding the best tree- based model for classification and prediction problems in gait analysis.

Railway track defects in Malaysia pose significant risks of train derailments and accidents, underscoring the urgency for early and accurate defect detection and classification. This study presents a novel approach utilizing deep learning models, VGG16 and YOLOv5, for detecting and classifying railway track defects, explicitly focusing on corrugation and squat defects. The research's uniqueness lies in its application of these specific models and the composition of a dataset collected from extensive field measurements and inspections across various railway tracks within the Track Network Maintenance Ampang Line in Malaysia. The results demonstrate that these models achieve high precision in defect classification and detection of defects by more than 80%. The proposed methodology provides the railway industry with a powerful tool to streamline maintenance planning and prioritize defect remediation efficiently. Early defect detection can prevent potential accidents and improve safety and operational efficiency. Future studies can expand on these findings by exploring the extension of the proposed techniques to address other types of rail defects. Incorporating a diverse range of scenarios and operating conditions in the dataset could further enhance the models' performance and generalization. Real-time deployment and integration with existing maintenance systems are crucial for practical adoption. This research has strengths but acknowledges limitations. Additional evaluation metrics and a diverse dataset are essential for model performance. Leveraging deep learning models offers a reliable solution for railway maintenance, enhancing safety and efficiency. Addressing these limitations will drive proactive defect management, ensuring safe and reliable railway networks.

Prediction of steam/water stratified flow characteristics in NPPs transients using SVM learning algorithm with combination of thermal-hydraulic model and new data mapping technique

In recent years, there has been a massive increase in the amount of Internet of Things (IoT) devices as well as the data generated by such devices. The participating devices in IoT networks can be problematic due to their resource-constrained nature, and integrating security on these devices is often overlooked. This has resulted in attackers having an increased incentive to target IoT devices. As the number of attacks possible on a network increases, it becomes more difficult for traditional intrusion detection systems (IDS) to cope with these attacks efficiently. In this paper, we highlight several machine learning (ML) methods such as k-nearest neighbour (KNN), support vector machine (SVM), decision tree (DT), naive Bayes (NB), random forest (RF), artificial neural network (ANN), and logistic regression (LR) that can be used in IDS. In this work, ML algorithms are compared for both binary and multi-class classification on Bot-IoT dataset. Based on several parameters such as accuracy, precision, recall, F1 score, and log loss, we experimentally compared the aforementioned ML algorithms. In the case of HTTP distributed denial-of-service (DDoS) attack, the accuracy of RF is 99%. Furthermore, other simulation results-based precision, recall, F1 score, and log loss metric reveal that RF outperforms on all types of attacks in binary classification. However, in multi-class classification, KNN outperforms other ML algorithms with an accuracy of 99%, which is 4% higher than RF.

Machine learning for the prediction of problems in steel tube bending process

Machine learning algorithms are crucial for crop identification and mapping. However, many works only focus on the identification results of these algorithms, but pay less attention to their classification performance and mechanism. In this paper, based on Google Earth Engine (GEE), Sentinel-2 10 m resolution images during a specific phenological period of winter wheat were obtained. Then, support vector machine (SVM), random forest (RF), and classification and regression tree (CART) machine learning algorithms were employed to identify and map winter wheat in a large-scale area. The hyperparameters of the three machine learning algorithms were tuned by grid search and the 5-fold cross-validation method. The classification performance of the three machine learning algorithms were compared, the results of which demonstrate that SVM achieves best performance in identifying winter wheat, and its overall accuracy (OA), user’s accuracy (UA), producer’s accuracy (PA), and kappa coefficient (Kappa) are 0.94, 0.95, 0.95, and 0.92, respectively. Moreover, 50 various combinations of training and validation sets were used to analyze the generalization ability of the algorithms, and the results show that the average OA of SVM, RF, and CART are 0.93, 0.92, and 0.88, respectively, thus indicating that SVM and RF are more robust than CART. To further explore the sensitivity of SVM, RF, and CART to variations of the algorithm parameters—namely, (C and gamma), (tree and split), and (maxD and minSP)—we employed the grid search method to iterate these parameters, respectively, and to analyze the effect of these parameters on the accuracy scores and classification residuals. It was found that with the change of (C and gamma) in (0.01~1000), SVM’s maximum variation of accuracy score is up to 0.63, and the maximum variation of residuals is 76,215 km2. We concluded that SVM is sensitive to the parameters (C and gamma) and presents a positive correlation. When the parameters (tree and split) change between (100~600) and (1~6), respectively, the RF’s maximum variation of accuracy score is 0.08, and the maximum variation of residuals is 1157 km2, indicating that RF is low in sensitivity toward the parameters (tree and split). When the parameters (maxD and minSP) are between (10~60), the maximum accuracy change value is 0.06, and the maximum variation of residuals is 6943 km2. Therefore, compared to RF, CART is sensitive to the parameters (maxD and minSP) and has poor robustness. In general, under the conditions of the hyperparameters, SVM and RF exhibit optimal classification performance, while CART has relatively inferior performance. Meanwhile, SVM, RF, and CART have different sensitivities toward the algorithm parameters; that is, SVM and CART are more sensitive to the algorithm parameters, while RF has low sensitivity toward changes in the algorithm parameters. The different parameters cause great changes in the accuracy scores and residuals, so it is necessary to determine the algorithm hyperparameters. Generally, default parameters can be used to achieve crop classification, but we recommend the enumeration method, similar to grid search, as a practical way to improve the classification performance of the algorithm if the best classification effect is expected.

Abstract: This research examines howfocusing on the evaluation of the efficacy of five machine learning models: Maxor Min, AdaBoost, Bagging, Random Forest and Decision Tree. The Adjoint Ratio Assessment (ARAS) method was applied to establish the models are analyzed based on six evaluation parameters: precision, accuracy, recall, log loss, MCC (Matthews’s correlation coefficient), and model complexity. Research Significance: The significance of this research is the systematic assessment of machine learning models intended for supply chain optimization. It provides insights into model selection, ensuring the adoption of algorithms that align with operational goals such as forecasting demand, optimizing inventory, and risk management. Methodology: The ARAS method is used to rank models based on their performance across evaluation parameters. This approach ensures a thorough evaluation of the advantages and disadvantages of each model, aiding supply chain practitioners in making informed decisions. Alternative: Adaboost Adaptive Boosting, commonly known as AdaBoost, is a technique of ensemble learning that merges several ineffective classifiers to produce a robust predictive model. In the field of supply chain management, it is often utilized for activities like demand forecasting and anomaly detection. the one of AdaBoost ability to focus on misclassified events during training is particularly useful in noisy environments. However, its sensitivity to outliers and high computational demands for large datasets can limit its scalability in some applications. Random Forest: Random Forest is a robust ensemble method that builds multiple decision trees and combines their outcomes to. Its ability to handle complex datasets makes Random Forest a reliable choice in a variety of situations. Decision Tree: Decision tree models are simple but powerful tools To make decisions. Their application in supply chain is extensive scenarios such as supplier selection, inventory categorization, and shipment prioritization. Packing: Bootstrap aggregation, or packing, improves model consistency and accuracy by training multiple versions of the same model on different data subsets and combining their predictions. In supply chain applications, packing is particularly useful for reducing variance and improving forecast reliability. Its ability to use multiple data sources ensures robust performance in Jobs like demand forecasting and stock control. Evalution parameter: Accuracy measures the overall correctness A model To make decisions. Their application in supply chain is extensiveIn supply chain situations, accuracy is important for tasks such as demand forecasting and product classification. While high accuracy indicates strong general performance, it does not fully reflect a model’s ability to handle unbalanced datasets. Precision: Accuracy evaluates the trustworthiness of affirmative forecasts, In supply chain applications such as fraud detection or quality control, high accuracy ensures that false positives are minimized, saving time and resources in decision making. Recall: Recall (or sensitivity) evaluates how effectively a model can recognize true positives from all true positives. This parameter essential for detecting rare but critical events in supply chain operations such as stockouts or supplier disruptions.Log Loss: Log Loss estimates the uncertainty in probabilistic predictions by imposing a higher penalty for incorrect predictions. In supply chain optimization, it is most useful for models that output probabilities, such as those predicting shipment delays or demand spikes. Lower log loss values ​​indicate higher confidence in the predictions. Matthews Correlation Coefficient (MCC): The MCC offers a measure of parity model performance, one is especially valuable in asymmetric datasets, such as those found in supply chain risk management, where accurate classification across all classes is important. Model Complexity: Model complexity reflects the complexity of the machine learning model, including the number of parameters, features, and computational resources required. In supply chain management, this is essential balance model complexity with performance.. Result: Random Forest emerges as the best performer due to its balance across metrics, followed by Decision Tree and Packing, which offer strengths in recall and consistency. Maxor Min shows competitive performance, while AdaBoost stands out in performance but requires further improvement for broader applications. Keywords: Machine Learning, Supply Chain Optimization, ARAS Method, Random Forest, Decision Tree, Predictive Analytics.

Now days, there is a huge increase in number of peoples who have been accessing many social networking sites and microblogging websites which open a new door to the impression of today’s generation. Various user reviews for a specific product, company, brand, individual, forums and movies etc. have been much helpful in judging the perception of people. Thus the analysts took the initiative to develop algorithms to automate the classification of distinctive reviews on the basis of their polarities particularly: Positive, and Negative. This automated classification mechanism is referred as Sentiment Analysis. The primary aim of this paper is to apply Support Vector Machine (SVM) machine learning algorithm to classify the sentiments and texts for product reviews that analyses different datasets used for classification of sentiments and texts. Furthermore, various data sets have been utilized for training as well as testing and implementing the Support Vector Machine learning algorithm to find the polarity of the ambiguous sentiments. Objectives: The primary objective of this paper is to implement Support Vector Machine (SVM) machine learning algorithm to classify the sentiments and texts for product reviews that analyses different datasets used for classification of sentiments and texts. Methods/Statistical Analysis: In this research study, several datasets have been applied for training as well as testing and simulating the support vector machine learning algorithm to compute the polarity of the ambiguous sentiments or reviews. Findings: At the outset, Support Vector Machine (SVM) classification algorithm gives higher accuracy 89.98% than the other ones. The obtained accuracy would be enhanced further by including more sentence forms. Finally, it concludes that the Support Vector Machine (SVM) algorithm behaves well. Application/Improvements: The performance resulting models are tested to measure accuracy of Support Vector Machine learning algorithm. Finally, the SVM classification algorithm has been achieved high accuracy and found better classification algorithm than others. Keywords: Clustering, Product Reviews, SVM, Sentiment Analysis

Supervised machine learning uses the label dataset to classify data. In this paper supervised machine learning algorithm Random Forest Classifier and Support Vector Machine learning algorithm are used to classify gender. Google Colaboratory is a Cloud based service which is also known as Colab. Google Colab is based on Jupyter Notebook where machine learning and deep learning concepts can be implemented. Google Colab provides free access to GPU which is very much required to disseminate Random Forest and Support Vector Machine Learning concepts. This paper made an attempt to classify gender based on the person’s height, weight, and shoe size using Random Forest and Support Vector Machine. The result shows that Random Forest Classifier accuracy is relatively better when compared with Support Vector Machine learning algorithm.

Pharming scams are carried out by exploiting the DNS as the main weapon while phishing attacks employ spoofed websites that appear to be legitimate to internet users. Phishing makes use of baits such as fake links but pharming leverages and negotiates on the DNS server to move and redirect internet users to a fake and simulated website.Having seen several challenges through pharming resulting into vulnerable websites, personal emails and accounts on social media, the usage and reliability on internet calls for caution. Against this backdrop, this work aims at enhancing pharming detection strategies by adopting machine learning classification algorithms. To further obtain the best classification results, an ensemble learning approach was adopted. The algorithms used include K-Nearest Neighbors (KNN), Decision Tree, Random Forest, Gaussian Naive Bayes, Logistic Regression, Support Vector Machine, Adaptive Boosting, Gradient Boosting, and Extra Trees Classifier. During the testing process, the classifiers were tested against four popular metrics: accuracy, recall, precision, F1 score, and Log loss. The results demonstrate the performance of all algorithms used, as well as their relationships. The ensemble model that included Logistic Regression, K-Nearest Neighbors, Decision Tree, Support Vector Machine, Gradient Boosting Classifier, AdaBoost Classifier, Extra Trees Classifier, and Random Forest produced the best results after evaluating them on the two datasets. Random Forest Classifiers showed a better performance of the classifiers, with mean accuracies of 0.932 and 0.939, respectively for each of the datasets when compared to 0.476 and 0.519 obtained for Naive Bayes.