Twitter, a predominant platform for instantaneous communication and idea dissemination, is often exploited by cybercriminals for victim harassment through sexism, racism, hate speech, and trolling using pseudony-mous accounts. The propagation of racially charged online discourse poses significant threats to the social, political, and cultural fabric of many societies. Monitoring and prompt eradication of such content from social media, a breeding ground for racist ideologies, are imperative. This study introduces an advanced hybrid forecasting model, utilizing convolutional neural networks (CNNs) and long-short-term memory (LSTM) neural networks, for the efficient and accurate detection of racist and hate speech in English on Twitter. Unlabelled tweets, collated via the Twitter API, formed the basis of the initial investigation. Feature vectors were extracted from these tweets using the TF-IDF (Term Frequency-Inverse Document Frequency) feature extraction technique. This research contrasts the proposed model with existing intelligent classification algorithms in supervised learning. The HateMotiv corpus, a publicly available dataset annotated with types of hate crimes and ideological motivations, was employed, emphasizing Twitter as the primary social media context. A novel aspect of this study is the introduction of a revised artificial hummingbird algorithm (AHA), supplemented by quantum-based optimization (QBO). This quantum-based artificial hummingbird algorithm (QAHA) aims to augment exploration capabilities and reveal potential solution spaces. Employing QAHA resulted in a detection accuracy of approximately 98%, compared to 95.97% without its application. The study's principal contribution lies in the significant advancements achieved in the field of racism and hate speech detection in English through the application of hybrid deep learning methodologies.
The automation of railway signalling control table preparation, a task historically marked by labor-intensity and susceptibility to error, is critically examined in this study. Traditional manual methods of generating these tables not only demand extensive effort but also bear the risk of errors, potentially leading to severe consequences in subsequent project phases if overlooked. This research, therefore, underscores the imperative for automation in this domain. An extensive review of existing methodologies in the field forms the foundation of this investigation, culminating in the enhancement of a select approach with advanced automation capabilities. The outcome is a standardized procedure, adaptable with minimal modifications to the unique national signalling norms of various countries. This procedure promises to streamline project execution in railway signalling, reducing both time and error margins. Such a standardized, automated approach is particularly pertinent to the Republic of Serbia, where this study is situated, but its implications extend globally. Key technologies employed include AutoCAD and Mathematica, which facilitate the requirements-driven automation process. This research not only contributes to the academic discourse on railway signalling automation but also offers a practical blueprint for its implementation across diverse national contexts.
This study, rooted in extension theory and the principles of knowledge engineering, explores and formulates a novel method for generating sports protective gear designs. Given the critical role of sports protective gear in safeguarding athletes from injuries, coupled with escalating demands for product quality, the aim is to uncover a more effective approach to innovative design. This method involves formalizing modeling of various elements in the design process and representing this information in the elemental form of knowledge engineering. Through the related analysis, divergent analysis, as well as permutation and conduction transformations of these elements, innovative design schemes for sports protective gear are generated. This process not only optimizes design schemes in depth but also ventures into new design methods and processes. The objective is to offer a novel perspective in integrating extension theory and knowledge engineering in the design of sports protective gear, aspiring to provide more effective strategies to enhance existing design workflows. The goal of this new design method is to produce sports protective gear that is both practical and innovative, thereby enhancing the safety and enjoyment of athletes.
In the realm of managerial decision-making, particularly within the last few decades, the process has emerged as a formidable challenge. This paper focuses on strategic decision-making, crucial in determining organizational success or failure amidst prevailing uncertainties. To address this, the Matrix Approach to Robustness Analysis (MARA), a recent innovation, is integrated with the established Strengths-Weaknesses-Opportunities-Threats (SWOT) matrix. This integration aims to deliver robust outcomes in strategic planning for travel agencies. The methodology involves a comprehensive analysis of internal and external factors pertinent to a travel agency, applying the analytical rigor of the SWOT matrix. Subsequent to this analysis, a series of strategies are formulated. Central to this study is the identification of key environmental indicators, as perceived by stakeholders, which influence strategic outcomes. Through these indicators, various future scenarios are constructed, culminating in nineteen plausible scenarios. Each strategy, totalling twelve, is then evaluated against these scenarios to ascertain the conditions under which they are most effective, resulting in a performance matrix. The final phase involves calculating the robustness analysis scores for each strategy under two different assessment conditions: rigorous and lenient. These scores provide a basis for strategy prioritization in both scenarios. The analysis reveals that the strategy of expanding new pilgrimage tours holds the greatest promise, while the employment of relatives within the agency is deemed least effective. This study contributes to the field by offering a structured methodology for travel agencies to navigate uncertain environments, using a combination of MARA and SWOT. The findings underscore the importance of scenario-based strategic planning and robustness analysis in enhancing decision-making processes.
This study explores dynamic simulation and integrated control in a space robotic arm system characterized by a fully-flexible arm and an elastic base. The elastic base is modeled as a lightweight spring, and the modal shapes of a simply-supported beam are selected via the assumed mode method to represent the bending vibrations of the flexible arm. Dynamic equations for the system are formulated by integrating Lagrangian mechanics with momentum conservation principles. The approach involves reducing the system into two lower-order subsystems using a dual-time-scale singular perturbation method. The first subsystem, exhibiting slow variation, accounts for the joint's rigid motion, while the second, fast-varying subsystem addresses the vibrations of the base and arm. Estimation of joint velocities is conducted through a Luenberger observer, complemented by the use of an Radial Basis Function (RBF) neural network to approximate parameter uncertainties within the system. This facilitates the control of rigid motion in the slow-varying subsystem. Subsequently, the fast-varying subsystem's vibration is actively controlled based on linear system optimal control theory. Numerical simulations validate the integrated control approach's effectiveness in managing both motion and vibration, demonstrating its potential in enhancing the operational precision and stability of space robot systems.
In the context of the global supply chain, the selection of Cold Chain Logistics Service Providers (CCLSPs) emerges as a paramount challenge, particularly for the transportation of temperature-sensitive goods. This study introduces a structured decision-making framework, addressing the need for efficient and reliable logistics services in this sector. Central to the framework is a hybrid Multi-Criteria Decision-Making (MCDM) model, which synergizes Fuzzy FActor RElationship (Fuzzy FARE) and Fuzzy Axial Distance based Aggregated Measurement (Fuzzy ADAM). This innovative approach is aimed at refining criteria weight determination and enhancing provider ranking accuracy. Special emphasis is placed on the integration of fuzzy logic to manage the inherent uncertainties present in subjective evaluations and decision data. The investigation underscores the criticality of factors such as stringent temperature control, robust infrastructure, and adherence to regulatory standards in the selection process. An application of this methodology is demonstrated through a case study involving the ranking of ten logistics providers in South-East Europe. The study's contributions are twofold: it advances the theoretical framework of supply chain management methodologies and offers a pragmatic tool for businesses operating within temperature-sensitive logistics networks. Prospective research directions include the adaptation of this framework to various regional contexts and the incorporation of emerging technologies, ensuring the framework's applicability and relevance in the dynamic domain of cold chain logistics.
In the realm of road safety management, the development of predictive models to estimate the severity of road accidents is paramount. This study focuses on the multifaceted nature of factors influencing accident severity, encompassing both vehicular attributes such as speed and size, and road characteristics like design and traffic volume. Additionally, the impact of variables, including driver demographics, experience, and external conditions such as weather, are considered. Recent advancements in data analysis and machine learning (ML) techniques have directed attention toward their application in predicting traffic accident severity. Unlike traditional statistical methods, ML models are adept at capturing complex variable interactions, thereby offering enhanced prediction accuracy. However, the efficacy of these models is intrinsically tied to the quality and comprehensiveness of the utilized data. This research underscores the imperative of uniform data collection and reporting methodologies. Through a meticulous analysis of existing literature, the paper delineates the foundational concepts, theoretical frameworks, and data sources prevalent in the field. The findings advocate for the continuous development and refinement of sophisticated models, aiming to diminish the frequency and gravity of road accidents. Such efforts contribute significantly to the enhancement of traffic control and public safety measures.
This investigation delineates the impacts of mining on karst systems, with a focus on specific karst zones, namely the epikarst, the vadose zone, and the phreatic zone, which includes the epiphreatic zone. Mining activities, regardless of the karst area type, predominantly affect these zones. When mining occurs at the surface or within the epikarst, it results in the destruction of surface features and the disruption of the epikarst, thereby locally halting karstification processes. The extraction in the vadose zone can lead to surface alterations, characterized by collapses, the formation of depressions, and the modification of epikarst activity, ultimately impacting surface karstification and inducing atectonic changes on the surface. The exploitation of the phreatic zone is associated with the artificial lowering of the karst water table and the removal of materials from cavities and depressions. This study emphasizes the importance of understanding the zone-specific impacts of mining on karst systems, highlighting the need for tailored conservation and management strategies to mitigate these effects. The findings contribute to the broader understanding of karst dynamics and provide a foundation for future research on the sustainable management of karst environments in the context of mining activities.
In the rapidly evolving domain of digital finance, the interplay between cryptocurrencies and external variables such as financial and social media indicators warrants thorough examination. This investigation employs a novel, entropy-weighted Multiple Attribute Decision Making (MADM) model to decipher these intricate relationships. The study's foundation is an expansive dataset, meticulously compiled to encompass a broad spectrum of financial data alongside diverse social media indicators. Central to this analysis is the employment of the Stepwise Weight Assessment Ratio Analysis (SWARA) method, meticulously applied to ascertain the relative importance of various social media indicators. Complementing this, the Complex Proportional Assessment (COPRAS) methodology is adeptly utilized to derive utility functions for each cryptocurrency under scrutiny. The analytical prowess of neural network regressions is harnessed to delineate the influence exerted by a multitude of financial indicators on these utility functions. The findings of this research are pivotal in understanding the dynamics within the cryptocurrency market. Bitcoin and Ripple emerge as pivotal entities, primarily functioning as primary conduits for market shocks. In contrast, Ethereum is identified as a stabilizing force, predominantly absorbing such fluctuations. A nuanced aspect of this study is the differential impact of social media indicators on various cryptocurrencies. Bitcoin and Ethereum display a negative correlation with these indicators, suggesting a complex, possibly inverse relationship with social media dynamics. Conversely, Litecoin, Dogecoin, and Ripple exhibit a positive responsiveness, indicating a heightened susceptibility to social media attention, sentiment, and prevailing uncertainty.
In the rapidly evolving landscape of digital healthcare, the integration of cloud computing, Internet of Things (IoT), and advanced computational methodologies such as machine learning and artificial intelligence (AI) has significantly enhanced early disease detection, accessibility, and diagnostic scope. However, this progression has concurrently elevated concerns regarding the safeguarding of sensitive patient data. Addressing this challenge, a novel secure healthcare system employing a blockchain-based IoT framework, augmented by deep learning and biomimetic algorithms, is presented. The initial phase encompasses a blockchain-facilitated mechanism for secure data storage, authentication of users, and prognostication of health status. Subsequently, the modified Jellyfish Search Optimization (JSO) algorithm is employed for optimal feature selection from datasets. A unique health status prediction model is introduced, leveraging a Deep Convolutional Gated Recurrent Unit (DCGRU) approach. This model ingeniously combines Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) processes, where the GRU network extracts pivotal directional characteristics, and the CNN architecture discerns complex interrelationships within the data. Security of the data management system is fortified through the implementation of the twofish encryption algorithm. The efficacy of the proposed model is rigorously evaluated using standard medical datasets, including Diabetes and EEG Eyestate, employing diverse performance metrics. Experimental results demonstrate the model's superiority over existing best practices, achieving a notable accuracy of 0.884. Furthermore, comparative analyses with the Advanced Encryption Standard (AES) and Elliptic Curve Cryptography (ECC) models reveal enhanced performance metrics, with the proposed model achieving a processing time and throughput of 40 and 45.42, respectively.
Epilepsy, the most prevalent neurological disorder, is marked by spontaneous, recurrent seizures due to widespread neuronal discharges in the brain. This condition afflicts approximately 1% of the global population, with only two-thirds responding to antiepileptic drugs and a smaller fraction benefiting from surgical interventions. The social stigma and emotional distress associated with epilepsy underscore the importance of timely and accurate seizure detection, which can significantly enhance patient outcomes and quality of life. This research introduces a novel convolutional neural network (CNN) architecture for epileptic seizure detection, leveraging electroencephalogram (EEG) signals. Contrasted with traditional machine-learning methodologies, this innovative approach demonstrates superior performance in seizure prediction. The accuracy of the proposed CNN model is established at 97.52%, outperforming the highest accuracy of 93.65% achieved by the Discriminant Analysis classifier among the various classifiers evaluated. The findings of this study not only present a groundbreaking method in the realm of epileptic seizure recognition but also reinforce the potential of deep learning techniques in medical diagnostics.
The swift global spread of Corona Virus Disease 2019 (COVID-19), identified merely four months prior, necessitates rapid and precise diagnostic methods. Currently, the diagnosis largely depends on computed tomography (CT) image interpretation by medical professionals, a process susceptible to human error. This research delves into the utility of Convolutional Neural Networks (CNNs) in automating the classification of COVID-19 from medical images. An exhaustive evaluation and comparison of prominent CNN architectures, namely Visual Geometry Group (VGG), Residual Network (ResNet), MobileNet, Inception, and Xception, are conducted. Furthermore, the study investigates ensemble approaches to harness the combined strengths of these models. Findings demonstrate the distinct advantage of ensemble models, with the novel deep learning (DL)+ ensemble technique notably surpassing the accuracy, precision, recall, and F-score of individual CNNs, achieving an exceptional rate of 99.5%. This remarkable performance accentuates the transformative potential of CNNs in COVID-19 diagnostics. The significance of this advancement lies not only in its reliability and automated nature, surpassing traditional, subjective human interpretation but also in its contribution to accelerating the diagnostic process. This acceleration is pivotal for the effective implementation of containment and mitigation strategies against the pandemic. The abstract delineates the methodological choices, highlights the unparalleled efficacy of the DL+ ensemble technique, and underscores the far-reaching implications of employing CNNs for COVID-19 detection.
In the field of computer-aided diagnostics, the segmentation and classification of biomedical images play a pivotal role. This study introduces a novel approach employing a Self-Augmented Multistage Deep Learning Network (SAMNetwork) and Deep Belief Networks (DBNs) optimized by Coot Optimization Algorithms (COAs) for the analysis of dermoscopy images. The unique challenges posed by dermoscopy images, including complex detection backgrounds and lesion characteristics, necessitate advanced techniques for accurate lesion recognition. Traditional methods have predominantly focused on utilizing larger, more complex models to increase detection accuracy, yet have often neglected the significant intraclass variability and inter-class similarity of lesion traits. This oversight has led to challenges in algorithmic application to larger models. The current research addresses these limitations by leveraging SAM, which, although not yielding immediate high-quality segmentation for medical image data, provides valuable masks, features, and stability scores for developing and training enhanced medical images. Subsequently, DBNs, aided by COAs to fine-tune their hyper-parameters, perform the classification task. The effectiveness of this methodology was assessed through comprehensive experimental comparisons and feature visualization analyses. The results demonstrated the superiority of the proposed approach over the current state-of-the-art deep learning-based methods across three datasets: ISBI 2017, ISBI 2018, and the PH2 dataset. In the experimental evaluations, the Multi-class Dilated D-Net (MD2N) model achieved a Matthew’s Correlation Coefficient (MCC) of 0.86201, the Deep convolutional neural networks (DCNN) model 0.84111, the standalone DBN 0.91157, the autoencoder (AE) model 0.88662, and the DBN-COA model 0.93291, respectively. These findings highlight the enhanced performance and potential of integrating SAM with optimized DBNs in the detection and classification of skin cancer in dermoscopy images, marking a significant advancement in the field of medical image analysis.