Articles

This research is a collaboration involving a university team, a partnering corporation, and a hemodialysis clinic, which is a cross-disciplinary research initiative in the field of Artificial Intelligence of Things (AIoT) within the medical informatics domain. The research has two objectives: (1) The development of an <i>Internet of Things (IoT)</i><i>-based</i><i> Information System</i> customized for the hemodialysis machines at the clinic, including transmission bridges, clinical personnel dedicated web/app, and a backend server. The system has been deployed at the clinic and is now officially operational; (2) The research also utilized de-identified, anonymous data (collected by the officially operational system) to train, evaluate, and compare <i>Deep Learning-based Intradialytic Blood Pressure</i><i> (BP)</i><i>/Pulse </i><i>Rate (PR) </i><i>Predictive Models</i>, with subsequent suggestions provided. Both objectives were executed under the supervision of the Institutional Review Board (IRB) at Mackay Memorial Hospital in Taiwan. The system completed for objective one has introduced three significant services to the clinic, including automated hemodialysis data collection, digitized data storage, and an information-rich human-machine interface as well as graphical data displays, which replaces traditional paper-based clinical administrative operations, thereby enhancing healthcare efficiency. The graphical data presented through web and app interfaces aids in real-time, intuitive comprehension of the patients’ conditions during hemodialysis. Moreover, the data stored in the backend database is available for physicians to conduct relevant analyses, unearth insights into medical practices, and provide precise medical care for individual patients. The training and evaluation of the predictive models for objective two, along with related comparisons, analyses, and recommendations, suggest that in situations with limited computational resources and data, an Artificial Neural Network (ANN) model with six hidden layers, SELU activation function, and a focus on artery-related features can be employed for hourly intradialytic BP/PR prediction tasks. It is believed that this contributes to the collaborating clinic and relevant research communities.

Sentiment analysis, a vital aspect of natural language processing, involves the application of machine learning models to discern the emotional tone conveyed in textual data. The use case for this type of problem is where businesses can make informed decisions based on customer feedback, identify the sentiments of their employees, and make decisions on hiring or retention, or for that matter, classify a text based on its topic like whether it is about a particular subject like physics or chemistry as is useful in search engines. The model leverages a sequential architecture, transforms words into dense vectors using an Embedding layer, and captures intricate sequential patterns with two Long Short-Term Memory (LSTM) layers. This model aims to effectively classify sentiments in text data using a 50-dimensional embedding dimension and 20 % dropout layers. The use of rectified linear unit (ReLU) activations enhances non-linearity, while the SoftMax activation in the output layer aligns with the multi-class nature of sentiment analysis. Both training and test accuracy were well over 80%.

Time series data are increasingly prevalent across domains such as finance, healthcare, manufacturing, and IoT, making accurate forecasting and anomaly detection critical for decision-making and system reliability. Traditional statistical methods (e.g., ARIMA, Holt-Winters) often fail to capture complex temporal dependencies and high-dimensional interactions inherent in modern time series. Recent advances in artificial intelligence particularly deep learning architectures such as recurrent neural networks (RNNs), convolutional neural networks (CNNs), temporal convolutional networks (TCNs), graph neural networks (GNNs) and Transformers have demonstrated marked improvements in modeling both univariate and multivariate series, as well as in detecting anomalies that deviate from learned norms (Darban, Webb, Pan, Aggarwal, &#x26; Salehi, 2022; Chiranjeevi, Ramya, Balaji, Shashank, &#x26; Reddy, 2024) [1,2]. Moreover, ensemble techniques and hybrid signal-processing + deep-learning pipelines show enhanced sensitivity and adaptability in real-world anomaly detection scenarios (Iqbal, Amin, Alsubaei, &#x26; Alzahrani, 2024) [3]. In this work, we provide a unified survey and comparative analysis of AI-driven time series forecasting and anomaly detection methods, highlight key industrial application domains, evaluate performance trade-offs (e.g., accuracy vs. latency, supervised vs. unsupervised learning), and discuss emerging challenges including interpretability, data drift, real-time deployment on edge devices, and integration of causal reasoning. Our findings suggest that while AI approaches significantly outperform classical techniques in many settings, careful consideration of data characteristics, evaluation metrics and deployment environment remains essential for effective adoption.

Over the past few decades, cardiovascular disease and related complications have surpassed all others as the important causes of death on a universal scale. At the moment, they are the important cause of mortality universal, including in India. It is important to know how to find cardiovascular problems early so that patients get better care and prices go down. This project utilizes the UCI Heart Disease Dataset to develop ML and DL models capable of detecting cardiac diseases. Heart illness was categorized using Convolutional Neural Network (CNN) models, which are able to detect intricate patterns in supplied data. A confusion matrix rating, an F1-score, a ROC curve, accuracy, precision, and recall were some of the measures used to grade the model. It did much better than the Neural Network, Deep Neural Network (DNN), and Gradient Boosted Trees (GBT) models, with 91.71% accuracy, 88.88% precision, 82.75% memory, and 85.70% F1-score. Comparative study showed that CNN was the most accurate model. Other models had different balances between accuracy and recall. The experiment results show that the optional CNN model is a decent way to identify cardiovascular disease. This means that it could be used in healthcare systems to find diseases earlier and treat patients better.

With the rapid growth of online communication platforms, the identification and management of toxic comments have become crucial in maintaining a healthy online environment. Various machine learning approaches have been employed to tackle this problem, ranging from traditional models to more recent attention-based transformer networks. This paper aims to compare the performance of attention-based transformer networks with several traditional machine learning methods for toxic comments classification. We present an in-depth analysis and evaluation of these methods using a common benchmark dataset. The experimental results demonstrate the strengths and limitations of each approach, shedding light on the suitability and efficacy of attention-based transformers in this domain.

Generative AI refers to deep learning technology that can automatically produce original text, images, audio, video, and other outputs. With its emerging capabilities, Generative AI can radically change the dynamics of key operational processes in most industries. In this document, we illustrate how it is possible to integrate Generative AI technologies into the Financial Reporting System (FRS) of a corporation. The integration will allow the FRS to deliver on demand concise and lucid insights to its associated users on what is happening in the corporation and different aspects of the corporation performance assessment, such as its liquidity, solvency, profitability, organizational structure, and share buy back decision. The integration will also facilitate the delivery of what-if analyses associated with different strategic and tactical decisions taken by the corporation management, such as capital budgeting and profit distribution decisions. The unique added value of several attributes of these insightful analytics is automating the responses to ongoing requests of the FRS users on the corporation. Generative AI capabilities are rapidly expanding. The integration can be applied not only for the corporate FRS but any FRS at the national or global levels delivered by a central bank or an accounting standards setter. Any of these FRS can be made into a unique “hub” for the integrated Generative AI technologies. An equally innovative possible generalized integration could put any corporate process at the center and its supporting FRS tasks and deliverables in its periphery.

Financial markets have become more and more complex, so has been the number of data sources. Stock price prediction has hence become a tough but important task. The time dependencies in stock price movements tend to escape from traditional models. In this work, a hybrid ARIMA-LSTM model is suggested to enhance accuracy of stock price forecasts. Based on time series indicators like adjusted closing prices of S&#x26;P 500 stocks over a decade (2010–2019), the ARIMA-LSTM model combines influences of both autoregressive time series forecasting with the substantial sequence learning property of LSTM. Data preprocessing in all aspects including missing values interpolation, outlier’s detection and data scaling – Min-Max guarantees data quality. The model is trained on 90/10 training/testing split and met with main performance metrics: MaE, MSE &#x26; RMSE. As indicated in the results, the proposed ARIMA-LSTM model gives a MAE value and MSE value of 0.248 and 0.101 respectively and RMSE of 0.319, a measure high accuracy on stock price prediction. Coupled comparative analysis with other Artificial Neural Networks (ANN) and BP Neural Networks (BPNN) are examples of machine learning reference models, further illustrates the suitability and superiority of ARIMA-LSTM approach as compared to the underlying models with the least MAE and strong predictive capability. This work demonstrates the efficiency of integrating the classical time series models with deep learning methods for financial forecasting.

Crime is a serious and widespread problem in their society, thus preventing it is essential. Assignment. A significant number of crimes are committed every day. One tool for dealing with model crime is data mining. Crimes are costly to society in many ways, and they are also a major source of frustration for its members. A major area of machine learning research is crime detection. This paper analyzes crime prediction and classification using data mining techniques on a crime dataset spanning 2006 to 2016. This approach begins with cleaning and extracting features from raw data for data preparation. Then, machine learning and deep learning models, including RNN-LSTM, ARIMA, and Linear Regression, are applied. The performance of these models is evaluated using metrics like Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE). The RNN-LSTM model achieved the lowest RMSE of 18.42, demonstrating superior predictive accuracy among the evaluated models. Data visualization techniques further unveiled crime patterns, offering actionable insights to prevent crime.

Managing supply chains efficiently has become a major concern for organizations. One of the important factors to optimize in supply chain management is logistics. The advent of technology and the increase in data availability allow for the enhancement of the efficiency of logistics in a supply chain. This discussion focuses on the blending of analytics with innovation in logistics to improve the operations of a supply chain. An approach is presented on how predictive analytics can be used to improve logistics operations. In order to analyze big data in logistics effectively, an artificial intelligence computational technique, specifically deep learning, is employed. Two case studies are illustrated to demonstrate the practical employability of the proposed technique. This reveals the power and potential of using predictive analytics in logistics to project various KPI values ahead in the future based on the contemporary data from the logistics operations; sheds light on the innovative technique of employing deep learning through deep learning-based predictive analytics in logistics; suggests incorporating innovative techniques like deep learning with predictive analytics to develop an accurate forecasting technique in logistics and optimize operations and prevent disruption in the supply chain. The network of supply chains has become more complex, necessitating the need for the latest technological advancements. The sectors that have gained a fair amount of attention for the application of technology to optimize their operations are manufacturing, healthcare, aerospace, and the automotive industry. A little attention has been diverted to the logistics sector; many describe how analytics and artificial intelligence can be used in the logistics sector to achieve higher optimization. Currently, significant research has been done in optimizing logistics operations. Nevertheless, with the explosive volume of historical data being produced by the logistics operations of an organization, there is a great opportunity to learn valuable insights from the data accumulated over time for more long-term strategic planning. To develop the logistics operations in an organization, the use of historical data is essential to understand the trends in the operations. For example, regular maintenance planning and resource allocation based on trends are long-term activities that will not affect logistics operations immediately but can affect the business’s strategic planning in the long run. A predictive analysis technique employed on historical data of logistics can narrow down conclusions based on the future trends of logistics operations. Thus, the technique can be used to prevent the disruption of the supply chain.

The major automated plants have produced large volumes of high-quality products at low cost by introducing various technologies, including robotics and artificial intelligence. The code of many defects on the surface of products is embedded with economic loss and sometimes functionality loss because products are rarely found with defects. Therefore, most items’ production is done based on prediction and has an invisible fluctuation in production. The detection process for hidden defect images requires a lot of costs and needs to be supported for better progress and quality enhancement. Paint shop defects should be analyzed from color changes to detect defects effectively by preventing the variability of product demand over time. It is not easy to take visible light images without noise because the paint surfaces are glossy. A few parts of illumination and shadows remain in images, even in larger size and high-resolution images. The various painted surfaces are also needed to reflect both color and texture information in computer vision models to classify defects precisely. Several automated detection systems have been applied to paint shop inspections using lasers, infrared, x-ray, electrical, magnetic, and acoustic sensors. The chance of paint shop defects can be low, unnecessarily low, compared to clouds in the sky, but those chances impact defect functionalities. Thus, they are called as “lessons learned.” Lately, artificial intelligence has been introduced to the field of factory automation, and many defect detection feeds have found footsteps in machine learning and deep learning. Recent attempts at deep learning-based defect detection are proposing simple techniques using specific neural network architectures with big data. However, big data is still in its early stages, and significant challenges exist in normalizing and annotating such data. To get cost-efficient and timely solutions tailored to automotive paint shops, it might be a better consideration to combine deep learning solutions with traditional computer vision and more elaborate machine learning methods.

Recent technologies have made life smarter. vehicles are vital components in daily life that are getting smarter for a safer environment. Advanced Driving Assistance Systems (ADAS) are widely used in today's vehicles. It has been a revolutionary approach to make roads safer by assisting the driver in difficult situations like collusion, or assistance in respecting road rules. ADAS is composed of a huge number of sensors and processing units to provide a complete overview of the surrounding objects to the driver. In this paper, we introduce a road signs classifier for an ADAS to recognize and understand traffic signs. This classifier is based on a deep learning technique, and, in particular, it uses Convolutional Neural Networks (CNN). The proposed approach is composed of two stages. The first stage is a data preprocessing technique to filter and enhance the quality of the input images to reduce the processing time and improve the recognition accuracy. The second stage is a convolutional CNN model with a skip connection that allows passing semantic features to the top of the network in order to allow for better recognition of traffic signs. Experiments have proved the performance of the CNN model for traffic sign classification with a correct recognition rate of 99.75% on the German traffic sign recognition benchmark GTSRB dataset.

Deep learning approaches are very useful to enhance cybersecurity protocols for industry-integrated big data enterprise resource planning systems. This research study develops deep learning architectures of variational autoencoder, sparse autoencoder, and deep belief network for detecting anomalies, fraud, and preventing cybersecurity attacks. These cybersecurity issues occur in finance, human resources, supply chain, and marketing in the big data integrated ERP systems or cloud-based ERP systems. The main objectives of this creative research work are to identify the vulnerabilities in various ERP systems, databases, and the interconnected domains; to introduce a conceptual cybersecurity network model that incorporates variational autoencoders, sparse autoencoders, and deep belief networks; to evaluate the performance of the proposed cybersecurity model by employing the appropriate parameters with real-time and synthetic databases and simulated scenarios; and to validate the model performance by comparing it with traditional algorithms. A big data platform with an integrated business management system is known as an integrated ERP system, which plays an instrumental role in conducting business for various organizations in society. In recent times, as uncertainty and disparity increase, the cyber ecosystem becomes more complex, volatile, dynamic, and unpredictable. In particular, the number of cyber-attacks is increasing at an alarming rate; the resultant security breaches have a disruptive and disturbing effect on businesses around the world, with a loss of billions of dollars. To combat these threats, it is essential to develop a conceptual cybersecurity network model to secure systems by functioning as a mutually supporting and strengthening network model rather than working in isolation. In this dynamic and fluid environment, introducing a deep learning approach helps to support and prevent fraud and other illicit activities related to human resources and the supply chain, among others. Some cybersecurity vulnerabilities include, for example, database vulnerabilities, service level vulnerabilities, and system vulnerabilities, among others. The proposed methodology focuses only on database vulnerabilities, with the main aim of detecting and mitigating new potential vulnerabilities in other dependent domains as a future initiative.

The growing complexity of the operating environment urges pharmaceutical innovation. This essay addresses the need for the integration of advanced technologies in the pharmaceutical supply chain. It justifies the value proposition and presents a concrete use case for the integration of deep learning insights to make data-driven decisions. The supply chain has always been a priority for the pharmaceutical industry; research and development recognizes companies' increasing investment in big data strategies, with plans for a CAGR in big data tool adoption. The work presented herein has a preliminary explorative character to recuperate and integrate evidence from partly overlooked practical experience and know-how. The practical relevance of the essay is directed toward practitioners in pharmaceutical production, supply chain management, logistics, and regulatory agencies. The literature has shown a long-term concern for enhanced performance in the pharmaceutical supply chain network. This essay demonstrates the application of deep learning-driven insights to reveal non-evident flow dependencies. The main aim is to present a comprehensive insight into deep learning-driven decision support. The supply chain is portrayed in a holistic manner, seeking end-to-end visibility. Implications for public policy are discussed, such as data equity: many countries are protecting their populations and economic growth by building resilience and efficiency to ensure the capacity to move goods across supply chains. The implementation strategy is covered. The combined reduction of variability, efficiency as matured richness, reliability (on stochastic flows and their understanding through deep learning and data), and system noise (increased dampening through the inclusiveness of all stakeholders) results in increased responsiveness of supply chains for pharmaceutical products. Future work involves the integration of external data, closing the loop between planning and its application in reality.

Neural networks are bringing a transformation in wearable healthcare technology analytics. These networks are able to analyze a vast amount of data to help in making decisions concerning patient care. Advancements in deep learning have brought neural networks to the forefront, making data analytics a straightforward process. This study will help in unveiling the use of ICT and AI in medical healthcare technology, crawling through some industry giants. Wearable Healthcare Technologies are becoming more popular every day. These technologies facilitate collecting, monitoring, and sharing every vital aspect of the human body necessary for diagnosing and treating an ailment. At the advent of global digitization, health data storage and systematic analysis are taking shape to ensure better diagnostics, preventive, and predictive healthcare. Healthcare analytics powered by neural networks can significantly improve health outcomes, maximizing individuals' potential and quality of life. The breadth and possibilities of connected devices are getting wider. From personal activity monitoring to quantifying every bit of health statistics, connected devices are making an impact in measurement, management, and manipulation. In healthcare, early diagnosis could be a lifesaver. Data analytics can help in a big way to make moves and predictions to save lives. We are in another phase of the digitization era, "Neural Network and Wearable Healthcare Technology Analytics." A neural network could be conceived as an adaptive system made up of a large number of neurons connected in multiple layers. A neural network processes data in a similar way as the human brain does. Using a collection of algorithms, for many neural networks, objects are composed of 'input' and 'output' layers along with the layers of the neural network.