Modern manufacturing enterprises have invested in a variety of sensors and IT infrastructure to increase plant floor information visibility. This offers an unprecedented opportunity to track performances of manufacturing systems from a dynamic, as opposed to static, sense. Conventional static models are inadequate to model manufacturing system performance variations in real-time from these large non-stationary data sources. This paper addresses a physics-based approach to model the performance outputs (e.g., throughputs, uptimes, and yield rates) from a multi-stage manufacturing system. Unlike previous methods, degradation and repair dynamics that influence downtime distributions in such manufacturing systems are explicitly considered. Sigmoid function theory is used to remove discontinuities in the models. The resulting model is validated using real-world datasets acquired from the General Motor's assembly lines, and it is found to capture dynamics of downtime better than traditional exponential distribution based simulation models. Index Terms-nonlinear stochastic differential equation (n-SDE) model, mean time between failure (MTBF), mean time to repair (MTTR), recurrence analysis, multi-stage manufacturing systems

In the realm of reservoir engineering, the application of machine learning has emerged as a transformative force, offering unprecedented insights into reservoir parameter characterization. In this study, we present a comprehensive analysis of four distinct machine learning models, namely Bagging, Extra Tree Regressor, XGBoost, and Ridge, to elucidate their efficacy in predicting permeability, a critical parameter for reservoir characterization. Our findings reveal an understanding of each model's performance. The Bagging model, while demonstrating an impressive trained accuracy of 0.99, exhibits some uncertainty in high permeability predictions, casting slight shadows on its applicability for reservoir characterization. In contrast, the Extra Tree Regressor model outshines the Bagging model with a trained accuracy of 100% and a prediction accuracy of 99.8%. It boasts lower absolute and absolute percentage errors, reinforcing its ability in permeability prediction. However, the XGBoost model takes a unique approach by emphasizing the density-corrected log over gamma-ray and sonic logs. Despite achieving remarkable trained and predicted data accuracy exceeding 99%, its reliance on the corrected density log introduces a mean absolute percentage error above 10, warranting closer scrutiny. In contrast, the Ridge model struggles, evident from its high AIC reading, signifying its limited compatibility with permeability prediction. Joint plots and LMplot analyses further showcases model behaviors. The Extra Tree model exhibits a 99% confidence interval, underscoring its reliability with minimal underpredictions. Conversely, the Bagging and Ridge models show susceptibility to high uncertainties in permeability predictions, particularly at extreme values. Our study concludes that the Extra Tree Regressor model excels in permeability prediction, with potential applications in reservoir interval assessments. The XGBoost model, while competent in sandstone reservoir prediction, bears a higher uncertainty burden. The Bagging and Ridge models, due to their uncertainty challenges, are less suitable for non-reservoir and sandstone reservoir interval predictions. High permeability correlations with elevated porosity, reduced water saturation, and lower gamma ray readings highlight the reservoir intervals' distinct characteristics. These observations underscore the reliability of our models and their potential contributions to reservoir engineering practices.

Key measures of socioeconomic indicators are essential for making informed policy decisions, but due to the high costs and operational difficulties of traditional data collection efforts, obtaining reliable socioeconomic data remains a challenge, particularly in developing countries. This work presents a deep learning methodology to estimate socioeconomic indicators using satellite imagery. The neural network model developed was trained at the Brazilian region of Vale do Ribeira with the goal of analyzing the socioeconomic indicator of income. The preliminary results showed that models using nightlight (NL) or multispectral daytime (MS) imagery performed better than models trained only on RGB bands and that models trained exclusively on NL or MS imagery performed similar to one another and nearly as well as the combined model MS+NL. Finally, the model yielded a low performance (R2 = 0.289), but it is still promising once the dataset employed was considerably smaller than the one used in the original study that attempted to replicate.

Thе usе of unsupеrvisеd machinе lеarning tеchniquеs,  spеcifically K-mеans clustеring algorithms,  in urban studiеs has gainеd significant attеntion in rеcеnt yеars.  Thеsе tеchniquеs havе provеn to bе valuablе in analyzing and undеrstanding various aspеcts of urban dеsign,  such as land usе pattеrns,  transportation systеms,  and population distribution.  This articlе aims to providе a comprеhеnsivе rеviеw of thе application of K-mеans clustеring algorithms in urban studiеs. Thе findings of this rеviеw dеmonstratе thе widе rangе of applications of K-mеans clustеring in urban studiеs,  from idеntifying distinct land usе catеgoriеs to undеrstanding thе spatial distribution of social amеnitiеs.  Furthеrmorе,  it is rеvеalеd that thе usе of K-mеans clustеring in urban studiеs allows for thе idеntification and charactеrization of hiddеn pattеrns and similaritiеs among urban arеas that might not bе immеdiatеly apparеnt through traditional analysis mеthods.  Ovеrall,  thе usе of K-mеans clustеring algorithms providеs a valuablе tool for urban plannеrs and rеsеarchеrs in gaining insights and making informеd dеcisions in urban dеsign.  keywords: k-means clustering,unsupervised Machine Learning ,urban studies  

This study tackles the limitations of traditional distance metrics in machine learning, particularly their inadequacy in managing missing data and varying feature importance. We introduce a novel custom distance metric that offers a more nuanced and robust measure of similarity, addressing the complexities and imperfections of real-world data. Our extensive evaluation of the CustomKNN algorithm, leveraging this custom distance metric, reveals exceptional performance and robustness across diverse datasets, consistently outperforming traditional metrics in various classification tasks. The custom distance metric shows superior robustness to missing data and scalability, despite its higher computational complexity and memory requirements. Our findings highlight the versatility and effectiveness of the CustomKNN algorithm, establishing it as a formidable option for classification tasks. This research advances the development of more adaptable and context-aware distance metrics, significantly enhancing the accuracy and reliability of machine learning models in practical applications.

The recognition of data as a natural resource has made headlines in the new era of industrialization. Companies now leverage this resource to enhance their services and products, often promising optimal outcomes. While data has always held significant value, recent advancements in AI and ML frameworks have brought this fact to the forefront. However, it was not until major scandals involving large corporations came to light that the critical issue of privacy was remembered. Consequently, the intersection of data, AI and ML frameworks, and privacy has emerged as a new area of research. Although numerous works have reviewed the development of various data protection techniques, it seems that most of them address the subject from a single perspective or attribute the entire concept of data privacy to a specific technique. This review aims to present an overarching view of the topic. It offers a systematic guideline that establishes a proper connection among the three elements: data, AI & ML frameworks, and privacy. The paper delves into each element from both an abstract and concrete standpoint, presenting the latest techniques to address data privacy concerns, including numerous lab simulations. It also recommends tools and resources for further study. Ultimately, it wraps up the central topic by outlining the challenges and prospective future research directions. Keywords: Data Privacy; AI & ML; Differential Privacy; Federated Learning;

As conversational AI systems such as ChatGPT have become more advanced, researchers are exploring ways to use them in education. However, we need effective ways to evaluate these systems before allowing them to help teach students. This study proposes a detailed framework for testing conversational AI across three important criteria as follow. First, specialized benchmarks that measure skills include giving clear explanations, adapting to context during long dialogues, and maintaining a consistent teaching personality. Second, adaptive standards check whether the systems meet the ethical requirements of privacy, fairness, and transparency. These standards are regularly updated to match societal expectations. Lastly, evaluations were conducted from three perspectives: technical accuracy on test datasets, performance during simulations with groups of virtual students, and feedback from real students and teachers using the system. This framework provides a robust methodology for identifying strengths and weaknesses of conversational AI before its deployment in schools. It emphasizes assessments tailored to the critical qualities of dialogic intelligence, user-centric metrics capturing real-world impact, and ethical alignment through participatory design. Responsible innovation by AI assistants requires evidence that they can enhance accessible, engaging, and personalized education without disrupting teaching effectiveness or student agency.

The reliable operation of transformers is critical to power system stability, and early fault detection in transformers can significantly reduce the risk of catastrophic failures. Dissolved gas analysis (DGA) is a commonly used method to assess transformer health, as gas concentrations correlate with fault conditions inside the transformer. Methane (CH4), a gas indicative of thermal faults, serves as a key parameter in detecting early signs of fault. This study proposes a linear epsilon support vector regression (SVR) model fine-tuned with Grid Search for predicting methane concentration based on transformer temperature, oil temperature, winding resistance, load factor, service life, and transformer age using grey relational analysis. By predicting methane concentration in ppm, this model supports proactive fault management, optimizing maintenance schedules, and effectively reducing transformer downtime. Comparative results suggest that the proposed SVR model is an effective baseline for improving dissolved gas analysis accuracy, contributing to enhanced transformer diagnostics and preventive maintenance strategies.