Emerging Trends in Machine Learning, Data Science, and Internet of Things

Skin cancer, especially melanoma, is the most common cancer globally and in the United States. About 10% of melanoma cases have a family history. In 2023, an estimated 97,610 new melanoma cases are expected in the US, with 58,120 affecting males and 39,490 affecting females. To enhance diagnostic accuracy, the DeiT (DataEfficient Image Transformer) model was applied to the ISIC 2019 dataset, which comprises a diverse range of skin lesion images. Our Transformer model outperformed traditional CNN models, achieving a precision of 94%, a recall of 95%, and an accuracy of 93%, surpassing models such as EfficientNet and DenseNet169. This superior performance is attributed to the DeiT model's ability to capture complex image patterns efficiently. Its success indicates that Transformer-based models hold significant promise for improving skin cancer diagnosis, leading to earlier detection and better patient outcomes.

Recognition of a medical issue is a prerequisite for dealing with any healthrelated ailment. The execution of diagnostic tests is the first step taken by medical professionals to address the concern. As the nervous system is one of the major systems that control the functionality of a living body, abnormalities in the system lead to considerable medical disorders that must be recognized accurately and ultimately addressed appropriately. Epilepsy is one of the most common as well as complicated neurological disorders. Substantial effort is required to acknowledge the source of epileptic concern and treat it. As the patient with an epileptic disorder loses consciousness, detecting the occurrence of an epileptic fit and locating the source of the issue is the most challenging task in the process of treating the condition. The task of predicting the occurrence of an epileptic fit and identifying the root cause of concern is approached through various neurological tests. Several neurological tests, executed by medical professionals during the process of detecting, locating, and ultimately treating the debilitating syndrome of Epilepsy, are discussed in this paper. As research in the domain demands data, this paper presents an overview of datasets related to various neurological tests.

This study delves into the realm of Twitter sentiment analysis, employing an array of machine learning algorithms. Through meticulous preprocessing of a substantial Twitter dataset, the research meticulously examines algorithmic performance. Results are elucidated through visual representations, revealing comparative efficacy, strengths, and limitations. This exploration significantly advances the comprehension of sentiment analysis within the context of social media, providing valuable insights for subsequent research and the refinement of methodologies in this dynamic domain. The study's findings contribute to the evolving discourse on sentiment analysis and its applications in the digital landscape.

Sentiment analysis, a field within natural language processing, has become a focus of interest lately due to the rapid increase in content created by users on online platforms. Its main aim is to identify the sentiment or emotional tone conveyed in text. Within the modern e-commerce environment, online product reviews are essential tools for consumers who face numerous choices. Gleaning significant insights from these reviews is crucial, and sentiment analysis, a branch of Natural Language Processing (NLP), plays a vital role in interpreting and grasping consumer emotions. This document provides an in-depth examination of sentiment analysis techniques tailored explicitly to Amazon product reviews. With the rise of e-commerce, Amazon has become a key source of insightful consumer feedback, making it a prime subject for sentiment analysis studies. This analysis examines how sentiment analysis research has evolved in response to ongoing updates to Amazon's review platform, highlighting the importance of developing robust and adaptable models. By exploring a range of methods and strategies, this document aims to provide readers with a comprehensive understanding of the sophisticated models used in analyzing sentiments found in Amazon's extensive product feedback network.

This study presents a comparative analysis of the YOLO v8 deep learning model and image processing techniques employing blob detectors for detecting colonies in microscopic images. The objective is to assess the effectiveness of every approach in precisely recognizing colonies in the photos. The study aims to identify the advantages and disadvantages of each strategy through experimentation and data analysis from both methods. Birch clustering is also applied to analyze the discovered particles further based on their sizes, providing insights into the colony clustering patterns. The results of this study provide important new insights for future colony identification research and advance our understanding of the capabilities of deep learning and conventional image processing techniques.

The main goal of the research paper is to contribute to augmenting the efficiency level of camera-monitoring systems of security and surveillance through economically efficient vision-detection methods. It will especially focus on the early detection of fire-related accidents and mishaps, causing significant accidents. Moreover, the paper emphasizes the need for modern technologies in the system of fire detection, paying much attention to the complexity of the task because of factors like deformity, varying camera flection, and variation in the environment. For these challenges, in this paper, a new approach to transfer learning using the convolutional neural network with dilated convolutions is proposed by employing the InceptionV3 model. From the results of the experiment on a dataset custom-built from fire images, the performance of the method proposed here is very efficient for generalization with new data and has a very low false alarm rate. It also presents the basic principles of fire detection systems involving control panels, fire detectors, and alarm output devices, and their use for prompt notice to the occupants and responding emergency organizations in the case of emergency fire events. 

This paper presents Smart Eat, an intelligent Mess Management System designed to improve food management practices in college environments through the use of machine learning. The system prioritizes user preferences by employing collaborative filtering algorithms to generate personalized menu recommendations, thereby enhancing the dining experience for students. By leveraging real-time data, the system enables food service managers to dynamically adjust menus based on attendance patterns and customer feedback, thereby reducing food waste and improving resource efficiency. The application integrates Firebase, Android Studio, and Flutter to ensure seamless performance of both the backend and frontend, featuring a userfriendly interface for both customers and business owners. Business owners gain access to comprehensive data analytics and automated notifications, allowing them to make informed decisions about menu planning and customer engagement. A limited survey was conducted to assess the prototype's usability, which indicated high satisfaction levels in areas such as menu customization, feedback quality, and order placement speed. The system fosters sustainable practices while ensuring economic scalability and operational efficiency. Through Smart Eat, the paper demonstrates how data-driven decision-making and personalization can redefine the food service ecosystem around educational institutions and potentially scale to broader food management scenarios. 

The rise in cancer cases worldwide, particularly in India, necessitates the development of quick and accurate diagnosis methods. A new initiative aims to utilize machine learning to enhance cancer gene detection, thereby saving time and reducing manual labor in conventional diagnostic procedures. The MERN stack tool and a machine learning algorithm are utilized to create a full-stack website that rapidly identifies probable malignant mutations. The approach combines Naive Bayes, Linear Regression, and Random Forest Classifier models for data processing and analysis. This cutting-edge strategy aims to revolutionize cancer diagnosis, enabling prompt and potentially life-saving therapies. This initiative represents a technological achievement and a crucial step towards reducing cancer prevalence. Additionally, our proposed model achieves an accuracy of 88 percent. The core contribution of this work lies in the integration of multiple machine learning algorithms into a unified, full-stack cancer gene detection system—capable of handling various cancer types, deployed in realtime via cloud infrastructure, and designed for practical clinical use through high performance, feature explanation ability, and accessibility. 

Cursor movement simplifies the user interaction with the system; it is a motive and visible pointer that is controlled by the user by using hardware (i.e., mouse and touchpad) to interact with the system. This technology can be simplified by using hand gestures instead of a physical device to manipulate the cursor. It can be useful to make the user experience simple and systematic. Cursor movement by hand gesture can be provided as an energy and time-efficient technology where the user interacts with the software by using gesture recognition. The gestures that manipulate the screen first will be stored in the system database. Then the hand present in front of the device will be recognized and stored as an object in the system, and then the gestures performed in front of the system with the hand will be matched with the gesture present in the memory of the system, which will lead to cursor operations and movements. This technology will be helpful to operate a system from a distance according to the user’s ease and without using a hardware component. This technology can be applied in all the sectors where there is a need for a Computer System, as it is an adaptable technology for users of any age and field.

The paper focuses on handwritten digit detection using the MNIST 0-9 dataset to develop an automated grading system. Four distinct machine learning models are employed: basic and latest Neural Network Models, such as Multi-Layer Perceptron (MLP) and Convolutional Neural Network (CNN), as well as traditional models like K-Nearest Neighbors (KNN) and Gradient Boosting (XGBoost). The objective is to accurately identify handwritten digits, which is crucial for grading purposes. By evaluating the models' performance metrics, the aim is to determine the most suitable model for efficient digit detection. The ultimate goal is to streamline the grading process, ensuring timely result declaration while maintaining proper record maintenance.

The incorporation of cloud-based systems is a novel use of Information Technology (IT) in the modern healthcare environment. To understand the far-reaching consequences of cloud computing adoption on data management, collaboration, and ultimately, patient care, this paper examines the profound implications of this practice. Cloud-based systems emerge as a promising way to overcome traditional limitations as the demand for effective and accessible healthcare solutions grows. They offer scalable, affordable, and interoperable solutions that have the potential to completely change the way healthcare information is stored, accessed, and used. In light of this, the article aims to provide a comprehensive analysis of the existing research, examine realworld case studies, and incorporate insights from medical experts to present a complete picture of the benefits and challenges of cloud-based healthcare systems. Through an exploration of specific use cases, including Electronic Health Records (EHR), telemedicine, and data analytics, the study aims to clarify how cloud computing might revolutionise healthcare delivery and influence the industry in the future.

Developments in communication technology have facilitated more data sharing in geographically dispersed settings, but they have also enlarged the attack surface, raising questions about network security. Research focuses on AI-based anomaly detection systems to improve Network Intrusion Detection Systems (NIDSs) in order to address this. However, data imbalance makes it more difficult for AI models to learn and effectively identify threats when legitimate traffic outnumbers malicious traffic. To balance data classes and enhance NIDS performance, suggested options include employing Generative Adversarial Networks (GANs) to produce artificial data that mimics small attack traffic. In order to guarantee that AI models acquire enough training data for both legitimate and malicious network traffic, the method includes GANs. Methods such as “voting classifier” (RF+AB+DT) and “stacking classifier” (LightGBM+DT) are used to improve the system's accuracy and resilience in detecting threats. 

Self-driving cars are the latest sensation in the 21st century. At the same time, in the present scenario, passengers lack trust and confidence in the system that operates independently. Therefore, it is crucial to have an interactive environment that allows users to interact with the system in real-time and take appropriate action. Uncertain situations may arise from these interactions. If the passenger and the Decision-making model differ in the instructions, the system can get perplexed. As a result, the purpose of this study is to create an environment for a self-driving car that can manage such unpredictable scenarios while facilitating user-system interaction. This paper proposes a VUR (Visual Uncertainty Resolver) algorithm to address the stochastic environment in which self-driving cars operate. This three-step algorithm addresses the uncertainties associated with the car's vision. The proposed algorithm utilizes YOLOv5 for object detection and classification, along with Mono-Depth Estimation and speech synthesis, to extract features from the environment for uncertainty detection and resolution. Our supervised learning model, the Decisionmaking model, considers various parameters to compute the best decision, which is efficient and effective for the self-driving car.

The Web of Clinical Things (WCT) coordinates the medical care industry with the Web of Things (WoT) environment. It enables the creation, collection, transmission, and analysis of clinical data through IoT integration. These IoT networks comprise various medical services, IT frameworks, clinical sensors, and management software. The IoMT enhances the medical care framework by creating a seamless, open, integrated, and reliable system. Nonetheless, ensuring secure verification in IoMT is challenging due to the diverse communication environment of various IoMT devices. Although various examinations have investigated potential IoMT device authentication techniques, more work is required, especially in client validation, to develop long-term IoMT solutions. Secret key administration remains one of the significant challenges in this specific circumstance. This paper presents an IoMTrelated electronic passwordless verification strategy that is quick, powerful, and secure. To achieve cross-stage similarity, it integrates a version of FIDO2/WebAuthn, one of the latest standards for passwordless authentication. This streamlines the process of obtaining and further developing client accreditations while maintaining compliance with regulations. The paper also assesses the delays in IoMT gadget verification and enrollment processes. Results and reproductions demonstrate how the proposed framework can achieve rapid validation on cloud servers with minimal enrollment and approval processes, regardless of device types.