Browse Items (16488 total)
Sort by:
-
Leveraging Deep Learning for Early Detection of Alzheimer's Disease from MRI Scans
Alzheimer's disease (AD) remains shrouded in mystery, with its early detection posing a significant challenge. This research paper delves into the cutting-edge realm of deep learning, exploring its potential to explore the brain's secrets and revolutionize AD diagnostics using Magnetic Resonance Imaging (MRI) data. Upon comprehensively reviewing the performance of six state-of-the-art models and studying their strengths and limitations on MRI data, this paper proposes a novel deep-learning architecture based on the InceptionV3 model for Alzheimer's Disease prediction using MRI data. The proposed architecture leverages convolutional neural networks (CNNs) to extract subtle brain structure and function patterns, potentially identifying early AD signatures before noticeable cognitive decline. The proposed model is validated on a large-scale MRI dataset that comprises four stages of dementia, demonstrating more insights. Inception V3 base model yielded 82% accuracy, measured using the metric Area Under the Curve (AUC), on the dataset, and an improved AUC of 87% was achieved by performing data augmentation to remove the class imbalance in the dataset. The proposed deep learning model built on top of Inception V3 exhibited an improved performance with an AUC of 88% underlining the potential of deep learning models in early AD detection. The paper's findings will contribute to the ongoing effort to revolutionize AD diagnosis and accelerate the development of personalized treatment strategies. Grenze Scientific Society, 2025. -
Analysing the Effectiveness of Solana Blockchain Platform and PoH Consensus Algorithm in Providing a Solution for Blockchain Scalability Problem
Solana started its journey in April 2018 and is now a public blockchain - based platform which aspires better scalability than other existing blockchains while providing security and decentralization. It backs the development of decentralized applications and smart contracts (DApps). The goal of the study is to confirm several of its characteristics, like its transaction throughput, or the pace in which legitimate transactions are committed to a Solana network block over the course of a one-second period (TPS). A secondary dataset that was gathered over the course of 60 days and made available on GitHub was utilized. Our data analysis findings demonstrate that the transaction throughput on an average is about 3006 TPS at a much lower transaction fees than the fees users pay for many other blockchains that facilitate the same operations, such as use of smart contracts and the development of DApps. The document explains the workings of the Solana blockchain, which, in the words of its creators, claims to address the scalability issue without compromising security and decentralization. Grenze Scientific Society, 2025. -
CCIR: The Next Frontier in Mobile Network Evolution - Integrating Communication and Computing for Enhanced Services
The mobile RAN is the bridge and enabler between end users and application services, and fortunately, the computational capacity of base stations in RANs[1] has experienced tremendous growth accompanied by increasingly stringent service requirements from emerging applications such as AR/VR. These two trends imply an unprecedented opportunity that the abundant computing resources in base stations could be leveraged to host latency-sensitive applications if being managed properly, thereby giving rise to a new vision named Communication and Computing Integrated RAN (CCIR), where not only communication but also computing services are delivered by RANs in a coherent way[2]. In fact, CCIR implies an even more radical departure of designing future mobile networks-going beyond regarding the role of RANs as connecting links. This article provides an elaboration on the fundamental design philosophies and principles of CCIR, the logical architecture. We will explore further on key tech-nologies toward realizing our vision. Specifically, we provide a thorough view on what CCIR really means - it involves different integration granularities between communication and computing functions; meanwhile it keeps evolving until approaching real-time joint scheduling and holistic resource management based on one unified infrastructure. To assess the feasibility and advantages brought by CCIR, several field experiments were carried out where computing resources are migrated within different base stations. In conclusion, CCIR is a prospective evolution direction for future mobile networks dealing with communication-computation-converged requirements in new service use scenarios[3]. Using current infrastructure more intelligently and efficiently will facilitate better user experience, greener mobile networks as well as serve as better platform for future advancements in wireless technology. Grenze Scientific Society, 2025. -
Enhanced Spam Detection in Short Message Service using Hybrid Techniques
Receiving unwanted text messages, or SMS spam, costs consumers time and money and poses a security concern. To address this issue, we can deploy a system that recognizes and automatically filters out undesirable messages. This method, a testament to the advancement in technology, employs machine learning algorithms that gain knowledge from a pool of communications classified as spam or not. Managing various message contents and languages is one of the system's unique challenges. Notwithstanding these challenges, the approach may be effective in reducing unsolicited communications, improving the security of people's mobile devices and saving them time and money. To address this issue, a variety of machine learning approaches have been employed, ranging from more modern deep learning methods like Convolutional Neural Networks (CNNs) to more traditional ones like Naive Bayes. It is common practice to assess the effectiveness of SMS spam classifiers using measures like as F1-score, precision, and recall. All things considered, SMS spam classification is crucial for protecting the security and privacy of mobile phone users and has useful applications in everyday situations. Grenze Scientific Society, 2025. -
Automated Waste Segregation using Raspberry Pi and Deep Learning
With rapid urbanization and increasing waste generation, efficient waste segregation has become a critical challenge for sustainable waste management. Traditional waste disposal methods rely heavily on manual sorting, which is inefficient, labor-intensive, and prone to errors, leading to improper recycling and environmental hazards. To address these problems, a clever waste segregation method is presented in this research. It automatically sorts waste into four categoriesglass, metal, plastic, and paper/cardboardusing computer vision and machine learning. A 720p webcam is used to collect images in real time, and the system is powered by a Raspberry Pi 4B with 4GB of RAM. A Convolutional Neural Network (CNN) model that was developed using the TrashNet dataset forms its basis. The model can correctly identify the waste in the photos due to an optimized training method that incorporates data augmentation, regularization strategies, and early stopping to prevent overfitting. An SG90 servo motor controls the lid, ensuring the garbage is placed in the appropriate compartment, while an MG996R servo motor swings the bin into place after the waste has been classified. The bin and lid go back to their initial places once the garbage has been dumped, preparing the system for usage again. Here, we are able to combine automated mobility, automated categorization, and real-time waste detection with embedded technologies, machine learning, and automation to separate waste with the least amount of human intervention. Furthermore, the system's scalability and adaptability make it appropriate for smart city initiatives, urban trash management, and wider industrial application. Consequently, this technology helps to tackle intelligent waste management problems, which facilitates the emergence of a sustainable and eco-friendly future. The system achieved a testing accuracy of 88.1%, showcasing its effectiveness and reliability. Grenze Scientific Society, 2025. -
Leveraging Hybrid Dual-Level Contextual Attention and Spiking Neural Networks for Effective Hepatic Malignancy Diagnosis
s Abstract Liver cancer remains the leading cause of cancer-related mortality, prompt-ing advanced diagnostic techniques for early detection and accurate classification in the health sector worldwide. Multifaceted deep learning methods have shown significant potential in medical imaging, but challenges exist in capturing intricate contextual information. In our research, we propose a novel hybrid framework that integrates Dual-Level Contextual Attention (DLCA) with Spiking Neural Networks (SNNs) to enhance the diagnosis of liver cancer. The proposed framework uses a DLCA mechanism that effectively extracts both local and global contextual features within the medical images and aids in precise lesion differentiation. The SNNs module supports computational efficiency and robust pattern recognition, enabling precise identification of subtle cancerous patterns by reducing redundant activations while preserving critical diagnostic information. Experimental evaluations on publically available datasets demonstrate the effectiveness of our work, showcasing its reliability in clinical applications. Moreover, the model offers a direction for future AI-assisted diagnostic tools in medical imaging and oncology. Grenze Scientific Society, 2025. -
Real-Time Implementation of Deep Learning Model for Polyp Classification and Segmentation in Medical Imaging
Real-time deep learning models for polyp identification and segmentation in medical imaging. Recognising the limits of current database systems for real-time applications, the research focusses on creating a deep learning model capable of recognising crucial picture components to aid in precise polyp categorisation. The suggested methodology is intended for realtime, practical healthcare and diagnostic applications that need quick polyp detection via preliminary colonoscopy testing. Performance investigation demonstrates that ResNet50 and EfficientNet B2 outperform other models, implying that they are suitable for real-world application and optimal outcomes. 2025 Bharati Vidyapeeth, New Delhi. -
Leveraging ML Based Technique for Mobile Sales Forecasting
The mobile phone industry is very competitive, so mobile sales forecasting is now imperative for businesses to forecast demand and order inventory in advance to plan strategically. This research focuses on the higher accuracy of mobile sales prediction and studies several machine learning models like Brand, Ratings, RAM, ROM, Battery- Power, pixel- height- and width, and targets alongside Camera Details as an alternate set to association rule mining. A real-time dataset that covers real-world mobile phone sales data has been collected and had its features pre-processed to fill in missing values and do the definite column encoding. Dataset were tested to understand the model performance of several predictive models, such as Decision Trees, Support Vector Machine (SVM), and ensemble methods (Random Forest and Gradient Boosting). The performance of each model was measured by accuracy, precision, recall, and F1-score. To address the issue of class in the sales categories (Low, Medium, High), stratified sampling and Synthetic Minority Over-sampling Technique (SMOTE) techniques were used. The results showed the predictive solid abilities of all the models in forecasting sales for different segments, with ensemble models performing better than individual classifiers in terms of prediction accuracy and robustness. This approach was further strengthened by applying hyperparameter tuning and cross-validation to improve the model's performance. The results are predicted to drive mobile retailers in the direction of improving demand forecasting and making data-driven decisions towards operational efficiency. 2025 Bharati Vidyapeeth, New Delhi. -
Sustainable X-Band Microwave Absorber Using Tea Waste and Recycled Carbon Composites
This paper introduces an innovative eco-friendly flat microwave absorber designed for X-band applications, utilizing plant waste and recycled materials, thereby contributing to a circular economy. The composite material is composed of 75 % used tea powder and 25 % carbon sourced from discarded batteries. The electromagnetic (EM) absorption performance of the proposed composite has been characterized, revealing enhanced absorption capabilities, averaging around 26.2 % for a thickness of 5 mm, compared to a 100 % tea waste configuration across the X-band frequency range. Furthermore, the results indicate that EM absorption improves with increasing frequency. The environmental advantages of this composite, including waste reduction and a minimal carbon footprint, position it as an attractive alternative to conventional RF absorbers. This sustainable solution shows promise for applications in EM interference shielding and microwave energy harvesting. 2025 European Association on Antennas and Propagation. -
MVTamperBench: Evaluating Robustness of Vision-Language Models
Multimodal Large Language Models (MLLMs), are recent advancement of Vision-Language Models (VLMs) that have driven major advances in video understanding. However, their vulnerability to adversarial tampering and manipulations remains under-explored. To address this gap, we introduce MVTamperBench, a benchmark that systematically evaluates MLLM robustness against five prevalent tampering techniques: rotation, masking, substitution, repetition, and dropping; based on real-world visual tampering scenarios such as surveillance interference, social media content edits, and misinformation injection. MVTamperBench comprises 3.4K original videos, expanded into over 17K tampered clips covering 19 distinct video manipulation tasks. This benchmark challenges models to detect manipulations in spatial and temporal coherence. We evaluate 45 recent MLLMs from 15+ model families. We reveal substantial variability in resilience across tampering types and show that larger parameter counts do not necessarily guarantee robustness. MVTamperBench sets a new benchmark for developing tamper-resilient MLLM in safety-critical applications, including detecting clickbait, preventing harmful content distribution, and enforcing policies on media platforms. We release all code, data, and benchmark to foster open research in trustworthy video understanding. 2025 Association for Computational Linguistics. -
Online Fake News Detection using Machine Learning and Natural Language Processing Algorithms
Fake news in the digital platforms plays a vital threat to many of the influencing factors of the society. The research focuses on the challenges of online fake news detection from the performance achieved by the machine learning classifiers using the natural processing techniques. Logistic Regression and Random Forest are taken to test the dataset containing labelled fake and real news for the study. The models are evaluated using the key metrices as accuracy, precision, recall, F1-score, confusion matrix and from the ROC AUC score. The research demonstrates which model is more reliable and more consistent for finding the fake and real news. For semantic text analysis BERT embeddings are used in the research as it will help analyze the article with more accuracy and precision. The metrics evaluation and the ROC curve of the both models helps in knowing even the slight deviations projected in the metrics. The differences in the valuations and the metrics values showed the capabilities of both the models in detecting the online news as fake or real. The comparisons made from the two models helps in evaluating the models and to understand the limitations of it as the analyze and detection of the text is more complicated. The research aims to deliver a strong foundation for real-time fake news detection in this new era. 2025 IEEE. -
Efficient Multilingual Language Detection Using Machine Learning Algorithms
Natural Language Processing (NLP) is one of the important technologies in recent days, because language detection this NLP is play a vital role. This research focuses on detecting languages using various machine learning algorithms. FastText, Recurrent Neural Networks (RNN), Support Vector Machines (SVM) algorithms are used for this experiment. The following datasets are used to take this result that is Europarl and Tatoeba. The proposed method is to preprocess, train, and test these models. Evaluation is done by measuring precision, recall, and F1 score of the three algorithms. Results show that RNN provides precision close perfect or near-perfect results in both bilingual and multilingual datasets. SVM performs with high precision and recall, but less than RNN. Its performance slightly decreases as the dataset increases. On the other hand, FastText, although fast and efficient, drops significantly in performance as the dataset grows, especially with the inclusion of a third language. It provides an all-inclusive methodology that has pinned the strengths and weaknesses of each algorithm, providing valuable insight into which one best fit real-world language detection task: RNN with their ability to handle complex sequences, SVM for large-scale high-dimensional sparse features, and FastText for simpler, smaller dataset. 2025 IEEE. -
An Extensive Analysis of Artificial Intelligence Integration in Management Approaches
Artificial Intelligence (AI) is now a strategic enabler across a large number of management domains in the digital transformation era. We conducted this review which analyzes AI and its integration into management by 990 peer reviewed publications from 2015 to 2025 from the Scopus database. It filtered studies on AI's role in strategy, HR, finance, operations and decision-making in a systematic manner. Latent Dirichlet Allocation (LDA) formed five key themes of predictive analytics, AI in HR, financial planning, intelligent decision systems, and explainable AI. The findings suggest that digital resilience needs drive the surge of AI related management research after 2019. This review points out emerging trends, difficulties in integrations, as well as critical insights which can orient the future research over such specific studies. 2025 IEEE. -
Thematic Mutual Funds: Performance and Benchmark Analysis Leveraging AI in Futuristic New-Age Investment Themes
Young investors love thematic mutual funds. By 2022, AUM will reach 1.73 lakh crore. These funds have a great possibility of long-term growth since they are timing the market and investing wisely. This research uses AI-driven analytics to examine a select thematic mutual fund that has done well despite risk, diversity, and market volatility concerns. Five actively managed themes were chosen from various fund firms. Infrastructure, ESG, Technology, Banking and Financial Services, and Pharma & Healthcare. We evaluated the best strategies in each topic using qualitative and AI-enhanced quantitative success metrics, including Alpha, Beta, Standard Deviation, Rolling Return, and Sharpe Ratio. We compared themes to key metrics to evaluate their performance over five years (2018-2022). Themed mutual funds perform well for future savings. In particular, AI-driven asset selection algorithms that were market-appropriate outperformed other purchasers. The findings demonstrate the importance of economic cycle-based themes and clever distribution strategies to maximize profits. This research illuminates investment choices and illustrates how AI-enabled active investing might improve market forecasts, volatility, and returns in themed mutual funds. 2025 IEEE. -
Automated Detection of Deepfakes using Integrated AI and Computer Vision Strategies
Deepfakes, or artificial intelligence-generated fake videos, are becoming a greater concern for online information trust, personal privacy, and digital content security. This paper presents a straightforward and understandable technique for automatically identifying deepfakes in order to address this significant problem. The approach makes use of conventional computer vision and machine learning methods. The model examines manually produced visual cues such as eye distance, mouth movement, and head tilt in video footage. To increase accuracy, it employs a variety of classifier types, including Random Forest, Gradient Boosting, and a soft Voting Classifier. A method known as SMOTE was used to clean and balance the data, and categorical data was transformed into a format suitable for machine learning models. With an F1-score of 0.9802 and 98% accuracy, the results demonstrate that the Voting Classifier, which combines several models, works admirably while being straightforward and effective. This method makes detection successful and simple to comprehend while offering a helpful tool for swiftly identifying deepfakes, especially on systems with constrained resources. 2025 IEEE. -
Improving Financial Audits and Management of Compliance using Artificial Intelligence and Secure Cloud Technology
Modern financial ecosystem requires highly complex audit trails and more stringent compliance issues therefore require highly advanced secure and intelligent systems. This research outlines a hybrid framework which juxtaposes Artificial Intelligence (AI) and Secure Cloud Technology to improve financial audit process and establish strong compliance management. Taking advantage of the strengths of AI, the strengths in question including Natural Language Processing (NLP), anomaly detection and machine learning classifiers, this system is used to enhance data accuracy, and detect irregularities in real time and automate regulatory reporting. At the same time implementation of Zero Trust Cloud Architectures, along with homomorphic encryption, provides data integrity, privacy, and end to end security. The proposed methodology is centred around the integration of intelligent document processing and blockchain-verified logs in the federated learning framework - where both transparency and decentralization are fostered. In addition, predictive analytics are used for the prediction of possible risks and non-compliance incidents to facilitate proactive decision making. Extensive simulations are used to reveal enhanced performance relative to traditional systems, with increased accuracy of anomaly detections, audit traceability, and validation speed-up of compliance. This integration is not only focused on streamlining audit workflows, but can also cut on operational cost and human error as well. The results emphasize the importance of employing AI-enabled secure cloud infrastructures as a primary strategy for financial institutions in a growing regulated digital economy while trying to sustain compliance. The new system achieves a 96.2% rate of accuracy while auditing and only consumes 91.3% the time in compliance to encourage efficiency. 2025 IEEE. -
Strategic Integration of AI in Modern Data Management
The exponential growth of data from sources such as social media, IoT, and enterprise systems has catalyzed a transformative shift in data management practices. This paper explores the integration of artificial intelligence (AI), edge computing, cloud-native frameworks, and graph-based techniques to support intelligent, low-latency, and scalable data processing across complex ecosystems. It presents a comparative analysis of classical versus modern data architectures, highlighting how technologies like Graph Neural Networks (GNN4TS), reinforcement learning, and large language models (LLMs) enable more adaptive, interpretable, and automated pipelines. The study also addresses challenges in legacy system modernization, time-series modeling, and cyber threat detection while underscoring the role of AI in autonomous database management and metadata enrichment. Further, it examines critical risks - including explainability, adversarial vulnerabilities, concept drift, and privacy preservation - associated with AI-integrated data workflows. A structured overview of emerging paradigms such as neuro-symbolic AI, adaptive governance in multi-agent systems, and the potential of quantum computing provides a future-focused lens on intelligent data ecosystems. The insights presented aim to assist researchers, data engineers, and decision-makers in navigating the evolving landscape of AI-driven data management. 2025 IEEE. -
Navigating the Ripple Effect: A Risk-Resilient Framework for Supply Chain Optimization
Supply chains around worldwide are growing increasingly vulnerable to intricate and interrelated hazards, such as localized disruptions, ie. the COVID-19 pandemic - have repercussions that affect manufacturers, suppliers, and buying habits in other geographical areas. This study offers a risk-resilient paradigm for supply chain optimization that combines cutting-edge risk mitigation and recovery techniques with conventional efficiency-driven tactics. The framework tackles both short-term operational issues and long-term sustainability objectives by fusing lean management, scenario-based stochastic demonstrating, Bayesian network analysis, and digital technologies like blockchain and artificial intelligence. In order to reduce the spread of disruptions and enhance decision-making in the face of uncertainty, this work highlights the significance of supplier collaboration, decentralized planning, and predictive analytics. This paper includes a thorough strategy for managing the ripple effect, improving supply chain adaptation, and guaranteeing continuous value delivery in unstable circumstances using comparative analysis and data from recent literature. 2025 IEEE. -
IT Strategies for Effective Marketing in Globally Diverse Corporate Environments
Today, organizations discern multicultural teams, dynamic consumer tastes and shrinking landscapes of competition bordering on the internet-centric global economy. This paper investigates the role of IT strategies in improving the strategic marketing within different corporate environments. An examination of IT's role in addressing marketing and management complexity in different cultural context is made. Theoretical models are reviewed and responsible global marketing practices are promoted through digital transformation to reshapes the business operations. We also illustrate IT based solutions for dealing with cross cultural communication barriers, resistance to change as well as team dynamics. In this case study and trend analysis with trends, we show how own best practices of market segments, digitalization as well as cross cultural management work together with IT to encourage agility, customer focus, and continuous learning in the organization. 2025 IEEE. -
Deep Learning in Project Planning and Scheduling
Controlling construction projects requires careful planning, and the most popular modelling techniques are the discrete-event simulator (DES), linear schedule (LS), and the critical path method (CPM). DES techniques, however, may become laborious and struggle to appropriately represent decision possibilities as complexity and restrictions increase. Through the reinforcement learning methods, deep learning-based artificial intelligence (AI) may be a viable substitute, enabling a quicker evaluation and suggestion of planning solutions for intricate building projects. This study investigates if artificial intelligence (AI) can replace DES in Insight, an illustrated constraint-based procedure planning tool for production and building. In the study, the difficulties of integrating AI into planning for building are discussed, along with the process modifications required to support deep learning techniques. Enhanced schedule, expenses, and efficiency in operation result from early planning of projects, which also balances conflicting project requirements. The planning of modern building projects is suggested to use a new conceptual methodology. 2025 IEEE.