The rapid proliferation of Industry 4.0 technologies has created an urgent need for intelligent and reliable predictive maintenance (PdM) systems. While multi-criteria decision-making (MCDM) frameworks like the Best–Worst Method (BWM) offer structured approaches for prioritizing maintenance tasks, their traditional reliance on subjective expert opinion limits their scalability and adaptability in dynamic industrial settings. This study addresses these limitations by introducing a robust, data-driven framework that integrates machine learning (ML) with BWM. This study presents a framework integrating ML models with BWM, an MCDM technique. While prior work has explored ML for fault detection/classification and hybrid MCDM + ML approaches, our innovation lies in automating BWM weight calculation via ML-derived feature importances, transforming tacit expert knowledge (traditionally subjective) into explicit, data-driven criteria weights aligned with Knowledge Management (KM) principles. The proposed methodology moves beyond a single-model proof-of-concept to present a comprehensive validation blueprint for industrial deployment. The framework’s efficacy is demonstrated using the standard Case Western Reserve University (CWRU) dataset, where rigorous cross-validation and statistical significance testing identified the optimal model, offering a compelling balance of high stability and efficiency for adaptive systems. Furthermore, simulations demonstrated the framework’s real-time viability, with low processing latency, and its resilience to concept drift through an adaptive retraining strategy. By integrating the empirically validated model’s feature importances into the BWM, this work establishes an objective, data-driven, and adaptive system for prioritizing maintenance, thereby advancing the transition toward autonomous and self-optimizing industrial ecosystems.

In this study, we investigate the role of Generative Pre-trained Transformer 4 (GPT-4) in enhancing interpretability of sequential predictions in Natural Language Processing (NLP). Our study introduces a hybrid model that integrates traditional sequential prediction models with GPT-4, aiming to generate detailed, context-sensitive explanations for model outputs. This approach is rooted in the use of advanced transformer architectures and a specialized tokenization method that maintains semantic coherence, allowing for deep contextual analysis by GPT-4. Additionally, we devise a rationale generation algorithm that achieves a balance between succinctness and informativeness. Our experimental validation spans across various high-dimensional datasets, including financial time-series and multilingual texts, employing both qualitative and quantitative metrics to evaluate the model’s performance. These metrics focus on the plausibility and consistency of the rationales, as well as the model’s predictive accuracy. Preliminary results demonstrate that our approach not only enhances the accuracy of sequential predictions but also significantly improves their interpretability. This finding highlights the potential of generative AI to bridge the gap between complex AI decision-making processes. This research underscores the viability of employing generative AI to elucidate the underlying mechanisms of sequential prediction models, paving the way for more transparent AI systems.

In biomedical text mining, it is imperative to be able to get accurate clinical insights from large, varied datasets. Transformer-based models like BERT and BioBERT have gotten top-notch results. However, the effectiveness of these models largely depends on the proper hyperparameter settings, which are typically determined by hand or through grid search, both of which are computationally expensive. The authors of this study recommend employing a Genetic Algorithm (GA) to autonomously modify transformer architecture hyperparameters such as learning rate, batch size, dropout rate, and attention heads when handling biomedical datasets. The GA uses selection, crossover, and mutation to change hyperparameters that are stored as chromosomes. It uses the validation F1-score fitness as a guide. The experimental results on benchmark biomedical datasets show a 6.8% improvement in F1-score and a 23% reduction in training time compared to baseline tuning methods. The results show that GA can efficiently search large search spaces and find high-performance configurations. This makes it possible to deploy transformer models in clinical text mining applications more quickly.

The explosive growth of IoT devices has radically expanded the attack surface across modern networks, introducing security risks that traditional defenses are ill-equipped to address. At the cutting edge, ensemble learning approaches have taken center stage for intrusion detection, however real-world implementation in IoT environments is hindered by many practical and technical challenges. In this review, we critically examine how bagging, boosting, stacking, and hybrid ensemble techniques are being adapted for IoT attack detection. Patterns emerge: most reported accuracy gains are achieved on benchmark datasets, yet practical details, such as inference speed, memory usage, and long-term reliability under changing traffic, are too often overlooked. Hardware and protocol diversity within IoT ecosystems present additional challenges, leading to the transfer of lab results to live deployments, which is uncertain at best. Methods for interpreting model decisions are still in the early stages of development, which means that system operators don't have clear instructions for what to do in high-stakes situations. Transparent reporting, validation under real conditions, and building solutions that balance adaptability with operational simplicity are imperative for meaningful progress in reliably intuitive interpretation. Although promising, ensemble methods are stuck in a gulf between academic results and genuine, robust, scalable security. Bridging this divide will require ongoing, honest engagement with these persistent challenges.

Due to the variety of chatbot types and classifications, students and advisers may struggle to select a trusted and effective chatbot. Chatbot classification depends on various factors, including task complexity, response style, and domain specificity. To explore the most suitable option for high school advising, this study employed semi-structured interviews with eight high school students to assess their perspectives on seven generative responses from a domain-specific chatbot, HSGAdviser, in comparison with ChatGPT-3.5. The advising-related questions covered topics such as university applications, admission tests, and academic majors. Transcribed data were analyzed using thematic analysis. Findings indicate that most students preferred HSGAdviser for its ease, brevity, and speed, especially for Yes/No questions. However, when dealing with complex or high-impact decisions, some students favored ChatGPT-3.5 for its detailed responses. A key limitation of the study is the small sample size.

The introduction of neural machine translation (NMT) into the teaching of philology is an opportunity that can radically change the process of decoding and teaching ancient texts in a more accurate and accessible way. This paper discusses a new strategy that will use high-level methods of translation to facilitate language knowledge and the conservation of heritage. The current approaches are mostly inadequate to retain the semantic richness and contextual depth of ancient languages, mainly because of low linguistic data and obsolete systems of translations. The consequences of these concerns are disjointed interpretations and hinder learning. To overcome these issues, it creates a proposal of a Transformer-Based Neural Machine Translation (T-NMT) architecture that is aimed at translating and maintaining the semantic structure of ancient texts. Through attention mechanisms and contextual embeddings, T-NMT is able to guarantee further linguistic and meaning alignment of language pairs. The suggested approach can be used in the curriculum of philology to enable comparative study of languages, digital preservation, and to make students interested in historically significant content. It is also used to aid endangered or extinct languages through automation of translation and reduces experts reliance. We have shown that T-NMT drastically increases the level of translation coherence, preserves semantic status, and increases the pedagogic quality of ancient textual information. This practice demonstrates great prospects in the development of AI-based philological education and cultural conservation.

The rapid expansion of the Internet of Things (IoT) has revolutionized smart home environments, enhancing automation and efficiency through devices like smart thermostats, security cameras, and voice assistants. However, this technological advancement has introduced significant cybersecurity challenges, as many IoT devices enter the market with inadequate security measures. This paper addresses the critical issue of malware detection in smart home networks, focusing on the limitations of traditional security mechanisms such as firewalls and signature-based intrusion detection systems. We propose an automated machine learning-based solution that employs Decision Tree and Random Forest classifiers to detect and mitigate malware threats effectively. Our approach includes comprehensive data preprocessing, feature selection using Chi-Square and MRMR techniques, and model evaluation through accuracy, precision, recall, F1-score, and ROC-AUC metrics. The study utilizes the CTU-IoT-Malware-Capture dataset, demonstrating that both classifiers achieve high accuracy in distinguishing between benign and malicious network traffic. The Random Forest classifier shows superior performance, highlighting its potential for real-world smart home security applications. This research contributes to the development of adaptive and scalable malware detection systems, offering practical insights for network administrators, security professionals, and IoT manufacturers. By integrating machine learning with IoT security, we aim to establish a robust framework for protecting smart home ecosystems from evolving cyber threats.

The isolation between product designers and potential customers is an upskilling and reskilling obstacle that has been researched using different tools and techniques. One of the tools linking the customer expectations to the product design is the house of quality and the quality function deployment sciences. Another field of closing the gap between the customer expectations and the product design is the research on 6 sigma and the minimization of the variations and the standard deviations besides the quality control statistics and charts. 6 stigmas led to intensive training for upskilling workers to be able to use this methodology. On the other hand, a research contribution is made in the field of designing services to close the gap between customer expectations and the service design like, for example, the SERVQUAL tool. SERVQUAL added gap-based upskilling to minimize gaps. Because of the dynamic nature of the market and the rapidly growing level of customer expectations and the continuous need to find exciters in the product design according to KANO model and because of the trending joint venture between technology and product design as a new trend in upskilling, there is a need for new methodologies to increase the potential level of customer satisfaction globally in different countries, cultures, and demographics. There is also a need to find a way to orient the skill of using the cognitive effort of the product designer toward the right tools that are indirectly impacting customer satisfaction positively. In this research paper, the researchers are trying to design a new methodology for upskilling the young product designer and measure its impact on potential customers’ satisfaction to find evidence of its validity. The research is applied on beginners’ level at early stages of designing new products because of the global orientation toward encouraging school students to start the upskilling process for innovative product design.The research is a quasi-experiment to measure the difference between the controlling group and the experimented one before and after using the STEM tool, the change in the cognitive balance full functioning skill, and the machine learning testing of the changes in the design complexity, and finally the impact of this change on the difference between the customer satisfaction before and after applying this methodology are studied. The results reveal the significant difference and increase in the level of customer satisfaction after using this new methodology. This contributes to customer-oriented upskilling to empower young designers to penetrate the dynamic market rapidly and successfully.

Even though the process of integrating AI into the educational domain continues to increase, the literary analysis remains a difficult problem to automate because of the intricate and subtle character of literary works. The classic approaches do not necessarily represent the complicated literary devices, themes, and character development. In order to address the concerns, the paper uses AI based Automated Literary Analysis (AI-ALA) model which uses Generative Pretrained Transformer (GPT-4), a state of the art generative language model in order to automate literary analysis in high school literature courses. The fact that this method is interactive at the time of lesson and that the teacher works with pupils to assess literary tools such as metaphors and sarcasm as well as discuss the character and theme in a setting is revolutionary. The successful adjustment of the algorithm to a set of books with high school students is the guarantee of their adequacy and applicability in the field of education. The experiments indicate that GPT-4 is more precise and context-responsive as compared to rule-based. It is an in-depth examination of characters, themes, and literary strategies. The strategy, according to the statistics, can present teachers with an effective and scalable tool to enhance student learning. The concept simplifies the process of literary analysis through interactive interaction combined with questions and answers sessions. Literary analysis automation and personalization will break the existing boundaries and enhance literature classes teaching.

Due to the fast growth of smart city infrastructures, intelligent, multilingual conversational systems that can provide real-time services to various populations are in high demand. The rise of smart city infrastructures has intensified this desire. Transformer-based solutions can be challenging to implement in chatbots like BERT due to their high processing and time requirements, particularly in metropolitan areas with poor infrastructure and limited IoT connectivity. BERT with Knowledge Distillation (BERT-KD) is the research goal. The program helps smart cities use low-latency multilingual chatbots. This strategy reduces huge BERT models to smaller student networks without compromising semantic richness or cross-lingual understanding using knowledge distillation. Adaptive distillation and multilingual embeddings improve inference speed, memory use, and conversational accuracy across languages. Experimental assessments on benchmark multilingual conversation datasets show that the BERT-KD can reduce model size by 42%, speed up inference by 35%, and keep correctness at 93.6% compared to the full-scale BERT baseline. Due to its reaction time dropping from 412 to 265 milliseconds, the chatbot might have real-time conversations for innovative city installations. Reducing Latency enabled this. These findings suggest that smart city ecosystem-based chatbot services can be efficient, scalable, and real-time. This would allow people to communicate with care providers regardless of culture or language.

Advancements in state-of-the-art models and algorithms for chatbots have significantly driven the growth and evolution of human–computer interaction (HCI) in recent years. As a result, numerous authors are inspired to study the most effective interactive chatbot design that can maximize students’ experiences. Despite the fact that high school is one of the most critical stage in a student’s life, there is a lack of studies that focused on developing effective interactive high school advising chatbots. To address this current gap, this study aims to elicit the main effective features for high school advising chatbot. The authors in this study conducted a semi-structured qualitative interview with six high school students in UAE and the MAXQDA Analytics software—ver. 22.6.0 is implemented by processing the thematic analysis to study the findings. The results revealed that high school students recommended the same general interaction design characteristics that was derived from the previous systematic review with emphasizing more into accurate, reliable, and trustworthy short answers with having minimal conversational issues. In addition to that, emotions and empathic factors are less important for high school students.

Behind the academic institution, it is a competitive environment in which there is a question of how to best represent the quality of education produced, what system of evaluation of students’ productivity best suited, and how to best predict future education needs. With the recent developments in the computerization of along with other methods data management, the rapidly expanding educational field is becoming more and more interested in finding the novel machine learning (ML) applications into educational Data Analysis. The purpose of this review is to in-depth critically assess the existing literature on the ML application to the educational field. The target is to search for the trend in this field and to define which way tends to give better results. The review focuses on the ML algorithms, examining their effectiveness in predicting academic performance, creating space for adaptive learning formats, and impacting other educational variables. The major objectives include better understanding of the existing gaps in research and opportunities to suggest relevant questions on the subject in the directions of future ML implementation development in education. Four main research questions include different ML approaches and algorithms used in educational data analyses, the role of features in ML modeling, types of educational data often considered in the study, the educational outcomes and changes associated with ML integration. By providing the thorough examination of those areas, the review is designed to present valuable implications for academic, student, and administrative practice, offering guidance for possible further studies in the field.

Purpose – This paper aims to review several studies published between 2018 to 2022 about advising chatbots in schools and universities as well as evaluating the state-of-the-art machine learning models that are deployed into these models. Methodology – This paper follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), it demonstrated the main phases of the systematic review, it starts with screening 128 articles and then including 11 articles for systematic review which focused on the current services of the advising chatbots in schools and universities, as well the artificial models that are embedded into the chatbots. Findings – Two main dimensions with other sub-dimensions are extracted from the 11 included studies as it shows the following: 1- Advising chatbots AI Architecture which includes other sub-dimensions on identifying the deep learning based chatbots, hybrid chatbots and other open-resources for customizing chatbots; 2- The goals of the advising chatbot as it includes both the admission advising and academic advising. Conclusion – Most of studies shows that advising chatbots are developed for admission and academic advising. Few researchers who study the chatbots in high schools, there is a lack of studies in developing chatbots for students advising in high schools. Limitations and future work – This study is constrained to review the studies from 2018–2022, and it is not exposed to the chatbots artifacts, even though, the human-chatbot interaction has an essential impact on students’ experiences. Future research should include the impact of chatbots interactive design and students’ experiences.

Purpose - This paper aims to develop a novel chatbot to improve student services in high school by transferring students’ enquiries to a particular agent, based on the enquiry type. In accordance to that, comparison between machine learning and neural network is conducted in order to identify the most accurate model to classify students’ requests. Methodology - In this study we selected the data from high school students, since high school is one of the most essential stages in students’ lives, as in this stage, students have the option to select their academic streams and advanced courses that can shape their careers according to their passions and interests. A new corpus is created with (1004) enquiries. The data is annotated manually based on the type of request. The label high-school-courses is assigned to the requests that are related to elective courses and standardized tests during high school. On the other hand, the label majors universities is assigned to the questions that are related to applying to universities along with selecting the majors. Two novel classifier chatbots are developed and evaluated, where the first chatbot is developed by using a Naive Bayes Machine Learning Algorithm, while the other is developed by using Recurrent Neural Networks (RNN)-LSTM. Findings - Some features and techniques are used in both models in order to improve the performance. However, both models have conveyed a high accuracy score which exceeds (91

Secure knowledge management is crucial for enterprises that have access to a lot of data in today's data-driven world. They can be more productive and innovative because it aids in decision-making. To protect sensitive information and stop cyberattacks, it is essential to make sure that knowledge management is done securely. The development of AI-enabled analytics approaches has the potential to significantly improve numerous facets of secure knowledge management and cybersecurity. These include areas such as Cyber Threat Intelligence, Disinformation and Computational Propaganda, Security Operations Centers, and Adversarial Machine Learning. Even though these topics have been widely researched, recent developments in AI present encouraging potential for advancement. The paper discusses the potential advantages of cutting-edge AI technologies like ChatGPT in a variety of contexts, including improving cybersecurity in various businesses. Further study, according to the article, is required to determine the appropriate contexts for implementation in cybersecurity, investigate the ethical and legal issues surrounding its use, explore biases in training datasets and processes, and explore the skills and resources needed to handle generative AI. Although perspectives on the need for limitations or regulations on ChatGPTs use vary, more research is required to answer these crucial topics.

Microgrid control is complex due to its need to accommodate the intermittence of renewables, balance generation with load, transit between grid-connected and islanded modes, and maintain reliable power supply to customers. Much research has addressed microgrid control complexity in both centralized and decentralized settings. This paper presents an intelligent software agent control with advanced autonomous capabilities to address the intermittent nature of renewables and their integration in real-world scenarios. Such capabilities include data acquisition, load and renewable generation forecasting, energy management, scheduling, optimal power flow, and real-time control to maintain generation-load balance in a secure and reliable microgrid environment. Accurate AI predictive models, machine learning algorithms, and non-linear optimization will be at the core function of the control agents.

With the speedy advancement of the web, a consistently expanding number of people that utilize online social media. Subsequently, hate speech becomes uncontrolled in social media, and it is critical to group the hate speech and control it before it spread. With the presentation and the advancement of deep learning, hate speech recognition becomes practice. Many examinations use information from social platforms, for example, Twitter and Facebook along with machine learning or deep learning advances to identify and perceive hate speech. In any case, there are insufficient surveys about this area. After studying various article's and research papers, no such review is available to see assortment of feature extraction/engineering methods (FEM) and ML-Algorithms that assess, which feature extraction/engineering procedure and ML-Algorithms can perform better on open source or openly accessible dataset. Thus, the purpose of this research paper is to look at 3-FET or strategies and 8-ML-Algorithums to assess their performance on an openly accessible dataset and this dataset has 3 classes. The research outcomes exhibited that SVM-Algorithm with BIGRAM features performed better with almost 80% accuracy. This research paper shows viable ramifications what's more, can be used as a gauge concentrate on to identifying hate speech communications/messages. Also, the result of various correlations will be utilized as condition of- workmanship strategies to look at future explores for existing mechanized text classification procedures.