Browse Items (16488 total)
Sort by:
-
Deep Learning Driven Predictive Analytics Framework for Assessing Customer Satisfaction in Health Insurance Services
The paper aims to design a predictive analytics platform based on deep learning that can measure the satisfaction of health insurance policyholders with accuracy. One of the major challenges in predicting customer satisfaction is the use of multiple sources of data. These sources also include unstructured consumer sentiments and official demographic and policy measures. The system that is suggested integrates a 1-D Convolutional Neural Network with TabNet, a deep attention-based neural network that is specifically good at handling tabular data. The two are combined to deal with inputs that are highly sentiment-loaded. To enable joint feature learning while maintaining interpretability, the dual-path architecture leverages feature attribution. The suggested method surpasses standard machine learning and isolated deep learning baselines accuracy more than 97% through experimental evaluation on a real-world LIC health insurance dataset. The findings provide the basis for an interpretable and scalable customer-oriented decision-making framework in health insurance. 2025 IEEE. -
Indian Emotional Alignment Score (I-EAS): A Pilot Framework for Culturally-Aware AI Evaluation in Multilingual Contexts
The current Emotional Intelligence (EI) standards of large language models (LLMs) are predominantly Western, thereby excluding billions of non-Western users from being judged fairly by AI. I-EAS, a new culture-based evaluation system, is grounded in indigenous epistemology and a computational evaluation framework to address a critical gap in AI Fairness. I-EAS takes into account four weighted factors: Emotional Alignment, Cultural Resonance, Coherence, and Code-Switching. It prioritizes cultural appropriateness and maintains a rigorous methodology. The pilot study uses 150 systematically crafted prompts in Hindi, Tamil, and English in diverse Indian cultural scenarios to test three LLMs: Claude Sonnet 4, GPT-4o-mini, and Krutrim-Base. The results highlight the inherent drawbacks of automated cultural assessments, as the weak relationships between human and automated assessments indicate that current NLP models are unable to respond to cultural nuances. The human evaluation was more discriminative than the automated evaluations, as the automated code-switching evaluations were not successful. The limitations of the pilot study include the presence of only one annotator. However, they prove I-EAS as a preliminary methodological and theoretical framework to equitable AI evaluation in culturally diverse situations and the necessity to use hybrid human-computational measures of AI systems in different cultural contexts. 2025 IEEE. -
A Green Inventory Model for Growing Items with Mortality and Permissible Delay in Payment
With rapid industrial growth, environmental concerns have become increasingly important. The expanding market and rising greenhouse gas emissions are significantly contributing to environmental degradation, pushing the Air Quality Index to high levels. Simultaneously, the growing global population is driving up the demand for livestock, which heavily relies on natural resources. This paper proposes an inventory model that incorporates key environmental factors, including carbon emission reduction through optimal investment, payment delays and mortality. The model aims to determine the optimal solution while taking into account the environmental impacts. An analytical discussion on the concavity of the objective function in relation to the decision variable is included. The paper outlines a solution methodology to obtain the optimal result, supported by a numerical example. Sensitivity analysis reveals that the selling price and investment in carbon emission reduction are the most influential parameters. 2025 IEEE. -
Enhanced Digital Image Watermarking Using 3-Level Discrete Wavelet Transform (DWT)
This study compares the algorithm's performance to that of the DWT level 1 and level 2 techniques while proposing a digital picture watermarking technology using a 3-step Discrete Wavelet Transform (DWT). The suggested method uses alpha blending to overlay a multibit watermark into the frequency subband of the lower cover image. The watermark's appearance is controlled by the blending scale. For uniformity, watermark extraction uses the same scale factor. The 3-stage DWT approach is superior because the algorithm performs well for various scaling factors that are obtained in relation to statistical characteristics connected to the Mean Square Error (MSE) and Peak Signal-to-Noise Ratio (PSNR). 2025 IEEE. -
Relationship Between Interpersonal Conflict, Stress and Burnout in Nurses using ML Classification
Nurses often face challenges due to heavy workloads, ambiguous roles, and hierarchical pressures. This makes them particularly susceptible to stress and burnout. These factors impair their mental health, productivity, and interpersonal dynamics, resulting in emotional exhaustion, dissatisfaction, and high turnover rates. Therefore, the association between interpersonal conflict, stress, and burnout, among nurses is examined in this study. Data were collected from 636 registered nurses in Karnataka. An Artificial Neural Network was used to classify nurses on low/high interpersonal conflict levels based on their stress/burnout levels. It was found that stress and burnout strongly predict interpersonal conflicts. Factors like impaired judgment and altered behavior contribute heavily to these challenges. AI and IoT can be used to manage stress and burnout. Predictive AI analytics can monitor physiological and behavioral indicators, which can help provide an early intervention. 2025 IEEE. -
Revolutionizing Legal Intelligence: Advances in Neural Networks and Language Models for Legal NLP
As the legal field continues to generate vast amounts of complex text, from contracts to court rulings, machine learning and natural language processing (NLP) techniques have emerged as valuable tools to help analyze and organize this data. In this paper, a number of state-of-the-art models will be reviewed and evaluated, including transformer models like BERT, GPT, and T5, and neural network models such as LSTM and CNN-RNN hybrids. These were then tested for the legal tasks of document classification, text summarization, and entity recognition. Some of the metrics used for evaluation include Accuracy, F1-Score, ROUGE, and BLEU. Advanced models, in particular large language models (LLMs), outperform the traditional methods by a large margin since they capture the niceties of legal language and structure much more completely. Meanwhile, high-quality legal datasets remain scarce, legalese remains incomprehensible to most, and the models are still relatively unexplainable. In sum, these challenges clearly call for future research in terms of data augmentation, explainable AI techniques, and more robust training methods that would allow AI-powered tools to be integrated much more effectively within lawyers' workflows to support them in their decision-making processes. 2025 IEEE. -
IoT Based Bus Identification and Distance Notifier for the Visually Impaired
Public transport is a major obstacle for the visually impaired, and it tends to limit their independence and mobility. To overcome this problem, the current project presents an IoT-based bus identification and distance notification system that is meant to offer real-time support and improve the traveling experience of the visually impaired commuters. The system uses a Raspberry Pi controller in combination with an RFID reader and several RFID cards, each one of which is designated for a particular bus, to detect oncoming vehicles. A GPS unit monitors the position of the bus at all times, so it is possible to calculate exact distances to any of the stops. For additional convenience, the system includes three push-button switches programmed to three predetermined bus stops. Users may pick their destination stop, and the system will offer auditory feedback in terms of distance to the selected destination. Feedback is offered by an earphone, providing hands-free use and receiving instructions without visual interaction. The integration of RFID-based bus identification, GPS location, and voice guidance provides smooth real-time information, less reliant on outside help. The system increases mobility confidence through accurate, timely, and convenient information, enabling blind travelers to make use of public transport independently. Through the combination of IoT technology and assistive technologies, the project enables enhanced accessibility and inclusion in urban mobility systems. 2025 IEEE. -
Advanced Machine Learning Framework for Precision Rainfall Prediction for Jharkhand, India
Jharkhand, characterized by a substantial agricultural population predominantly reliant on rain-fed agriculture, faces significant challenges due to the erratic nature of precipitation. The study uses meteorological variables and historical rainfall data from the Jharkhand Space Agency Centre (JSAC) to predict rainfall with precision and resilience. Three supervised machine learning algorithms, Random Forest, K-Nearest Neighbour (KNN), and Ridge Regression, are employed to evaluate their performance across monsoon and non-monsoon periods. A novel algorithm is proposed for Jharkhand, offering better modularity and accuracy to predict the Rain Index. The results show the efficacy of these algorithms in capturing the temporal variability of rainfall in Jharkhand. The ensemble modeling model obtained an MSE score of 0.457, providing valuable insights into the viability and competence of machine learning algorithms for rainfall estimation. This research offers a valuable scope for policymakers, researchers, and stakeholders to formulate sustainable strategies to address climate variability and its impact on rain-fed agriculture. The study contributes significantly to meteorological research and offers valuable insights for policymakers, researchers, and stakeholders. 2025 IEEE. -
Financing Green Startups: A Blockchain-Powered Approach
Green startups routinely encounter difficulty in obtaining financing owing to the high funding needs to launch their business, and the imprecise market acceptance and future returns of those businesses, rendering them unacceptable to traditional latter-day funding methods. The funding modality available today fails to provide the needed transparency, flexibility and accessibility to encourage ventures based on green projects. This paper develops a funding model based in blockchain for green startups, that employs tokenization, decentralised finance (DeFi) and smart contracts with automaticity, as an efficient way of funding to provide secure, traceable funding structures that make funds available related to the performance of the venture. We present a conceptual model showing the responsibilities of startups, funders and smart contracts within a de-centralised funding ecosystem. Various case studies such as Power Ledger and WePower are investigated in order to validate the practical relevance of the model. Our research indicates how blockchain mechanisms can heighten trust, enhance liquidity, and automate funding that is linked to impact. This paper contributes to future work on scalable platforms that are both regulation compliant and also provide a fit between Blockchain infrastructure and the unique requirements of sustainable innovation. 2025 IEEE. -
Analyzing Uzbekistan's Export Trends: A Data-Driven Study on Annual Export Volume by Country
Uzbekistan's export sector plays a crucial role in economic development, contributing to GDP growth and international trade partnerships. This study analyzes annual export volume by country using the SDMX dataset, highlighting key trends, regional trading partners, and sectoral contributions. The research employs time-series analysis, trade flow modeling, and export diversification measures to provide insights into Uzbekistan's evolving trade structure. The results indicate shifting export dynamics influenced by geopolitical factors, regional economic policies, and trade agreements. Policy recommendations focus on enhancing market diversification, improving export efficiency, and fostering trade competitiveness. 2025 IEEE. -
Analysis of Market Communication and Informatization Services: A Data-Driven Study Based on SDMX Statistics
The market communication and informatization services sector plays a crucial role in modern economic development, facilitating digital transformation and connectivity. This study leverages official statistical data to analyze trends, growth patterns, and the economic impact of communication services in Uzbekistan. Using the SDMX dataset, we evaluate sectoral contributions, regional disparities, and the role of technological advancements. The findings provide insights into investment efficiency and policy recommendations for sustainable sectoral growth. 2025 IEEE. -
Analyzing Wholesale Trade Volume in Uzbekistan: A Data-Driven Study of Internal Trade Dynamics
Wholesale trade as a segment of internal trade possesses a great potential for shaping supply chain management, and market conditions in Uzbekistan. This paper discusses monthly changes of the wholesale trade volume regarding its main issues, seasonal fluctuations, and regional diversification using the SDMX dataset. Using timeseries analysis, trend decomposition, and correlation modeling, this paper aims at determining the effects of the economic policy, consumer demand factors, and trade restrictions in the wholesale trade sector in Uzbekistan. It also reveals seasonality in the trade, trade heterogeneity across regions, and impact of a shock in the market, which is very relevant and useful for policymakers and key players in the trade business. 2025 IEEE. -
From Producer to Consumer: AI-Blockchain Integration for Sustainable Supply Chain Tracking and Optimization
The blockchain is transforming the way supply chains operate. It is a decentralized peer-to-peer system that ensures safe and transparent data interchange. Because there is no central authority, blockchain technology can't be hacked; the data are dispersed throughout numerous nodes. Therefore, more openness, safety, and resistance to tampering are assured as opposed to conventional centralized database systems that depend on a single authority to manage all the data. This would ensure the immutability of the transaction log, which accurately traces the products from origin to destination. The supply chain refers to the transportation and distribution of goods when transactions are verified in real time and integrated into a secure, cryptographic ledger. This decentralized framework provides better tracking of various manufacturers, reduced chances of fraud, and greater trust among the participants. The early understanding based on tracking will definitely helps the stakeholders to improve the blockchain activities, reliability, it will minimize risks, and enhances efficiency. This will parallelly help those who are involved within supply chain management giving high accountability and seamlessness in the movement of details. 2025 IEEE. -
Towards Smarter Warehouse Layouts: Simulation-Driven Insights on Congestion and Forklift Flow Patterns
In high-mix, make-to-order warehouse environments, slotting decisions in constrained warehouse settings affect flow dynamics, yet their behavioral implications remain underexplored. This study employs discrete event simulation (DES) using the software FlexSim to evaluate three slotting strategies: baseline reflecting current operations, a benchmark informed by heuristic frequency-based clustering, and a proposed layout based on a learned configuration within a forklift-operated warehouse characterized by narrow aisles, unidirectional traffic, and spatial contention. Instead of emphasizing throughput alone, the study takes a closer look at how forklifts operate on a day-to-day basis, specifically how their time is divided between active movement, waiting due to blockages, and idling with no task assigned. Each strategy was tested over 20 simulation replications. Notably, the proposed layout cut blocked time by more than 30% and allowed forklifts to remain idle (and ready) more often, without reducing overall utilization. These patterns held consistently across runs. Statistical analysis confirmed that the differences in forklift behavior were significant, and Levenes tests showed that performance didnt become more erratic. These findings demonstrate that the improvements are systematic, not random. The work presents a simulation-based method for diagnosing layout effectiveness by looking at behavior, not just outputs, connecting slotting choices to real operational flow and system stability. This approach supports more resilient warehouse designs in settings with limited space and high product mix. 2025 IEEE. -
Co Adaptation Vs Signal Altering Regularization Layer in Deep Learning: A Trade Off Analysis via Node Redundancy and Transfer Learning
This study investigates the trade-off between incorporating regularization layers-such as Dropout, R-Drop, and Gaussian Dropout-and the phenomenon of co-adaptation in deep learning models. While regularization is designed to enhance generalization by disrupting hidden layer activations and reducing overfitting, it may also introduce node redundancy, potentially diminishing the model's capacity to learn efficiently. Conversely, co-adaptation, though often considered undesirable, may help preserve beneficial internal representations that contribute to learning generalizable data patterns-particularly in transfer learning scenarios-where regularization may inadvertently hinder such learning. Using the CIFAR-10 dataset, this study conducts an empirical analysis of how various regularization strategies influence neuron redundancy and downstream transfer performance. The results indicate that, although regularization effectively controls overfitting, excessive distortion in hidden representations can impair the model's ability to generalize across tasks. These findings provide insights into the need for balanced regularization strategies that maintain useful structure while minimizing detrimental redundancy. 2025 IEEE. -
Blockchain and Social Networking in the Age of Privacy: A New Approach to Data Protection and Transparency
Online social networks (OSNs) have become a significant area of application due to the rapid increase in online interactions. However, the unauthorized exposure of users' private information can lead to severe repercussions, including risks to users' personal safety. Privacy concerns in OSNs have garnered widespread attention. While some research efforts have attempted to address these privacy challenges in recent years, they often overlook the need to maintain essential social network functionalities including data access, retrieval and sharing. Consequently, ensuring the protection of sensitive information for delivering privacy-preserving and efficient services for social network remains a complex challenge. 2025 IEEE. -
Dynamic Load Balancing in Cloud Computing Using Memetic Algorithm for Better Response Time
Load balancing is a critical aspect of distributed computing systems, involving the distribution of workload across multiple resources to ensure optimal utilization and performance. In today's digital landscape, where the demand for online services continues to grow exponentially, efficient load balancing mechanisms are indispensable for maintaining high availability, scalability, and reliability of web applications, cloud services, and network infrastructures. The paper implements a load balancing solution using a Memetic Algorithm, an amalgamated optimization method that integrates elements of genetic algorithms with local search methods. By leveraging the principles of evolutionary computation and problem-specific knowledge, memetic algorithms offer a promising approach to addressing the complex optimization and the hurdles connected with the load distribution across the systems. The proposed algorithm is compared with the existing algorithms and proved to be slightly better than the traditional algorithm in terms of response time. Experimental results show that the proposed approach reduces response time by 0.26 seconds. 2025 IEEE. -
A Hybrid Clustering Approach for Enhanced Classification Efficiency in Data Analytics
Clustering is a fundamental technique in data analytics that groups data points with similar characteristics into clusters. It is crucial for uncovering hidden patterns, trends, and structures in datasets. Clustering reduces the complexity of large datasets by summarizing data into representative clusters. This simplification makes it easier to analyze and interpret data, especially when dealing with high-dimensional datasets. By identifying meaningful groups, clustering provides actionable insights that supports decision-making. For instance, businesses can make concrete decisions about product recommendations, pricing strategies, or resource allocation based on cluster analysis. The approach described in the paper offers an efficient method for combining K-means and Gaussian Mixture Model (GMM) clustering techniques. The method combines two wellknown clustering techniques, K-means and GMM, to leverage their respective strengths. K-means is known for its simplicity and efficiency, while GMM can model complex data distributions with varying covariance structures. Instead of directly integrating the results of K-means and GMM, the approach uses a simplified averaging technique to converge the cluster labels obtained independently from both methods. This suggests that the method may involve assigning weights to the cluster labels obtained from K-means and GMM and then averaging them to obtain final cluster assignments. Overall, this approach presents a promising direction for combining K-means and GMM clustering techniques, offering a streamlined integration process that simplifies the consideration of varying covariance types in GMM. The effectiveness of the method is evaluated through empirical studies and comparisons with existing clustering approaches. 2025 IEEE. -
Analyzing Technology Ecosystem Business Models: A Predictive Modelling Approach
In the rapidly changing landscape of technology, companies are devoting an increasing amount of their resources to developing product ecosystems that collaborate to deliver enhanced consumer experiences and strengthen their business models. As opposed to traditional standalone solutions, these ecosystems are intended to facilitate everyday tasks, increase user engagement, and provide seamless integration, all of which ensure a steady stream of revenue and dedicated customer base. This analysis provides an overview of the many ecosystem models that are now transforming the technology industry. An examination of ecosystems that help businesses maintain long-term revenue sustainability and high customer retention rates is provided by the model analysis, along with insights into how ecosystems may enhance user experience by being more connected, straightforward, and user-friendly. Technology ecosystems' quantitative effects are lacking, which makes it difficult to comprehend how they affect long-term revenue sustainability and customer retention. It is challenging to understand how technological ecosystems impact long-term revenue sustainability and customer retention due to the lack of measurable consequences. Through the use of multiple linear regression, this study illustrates the ecosystem business models' long-term revenue and customer retention. The study visualized the relationships of the technology ecosystem with an accuracy of 90-99%. This shows how to measure ecosystem impact and gives firms data-driven insights to improve their ecosystem initiatives. 2025 IEEE. -
Machine Learning Approaches for Predicting Player Position in Football
This paper presents a comparative study of different machine learning algorithms, including K- Nearest Neighbor (KNN), Random Forest, Gradient Boosting, XG Boost, Support Vector Machine, Voting Classifier and Logistic regression to develop a Player Position Prediction System in football. Initially, the study utilized a modified dataset containing 18434 records, focusing on simplicity for ease of analysis. Through experimentation, it was found that Logistic regression provided a strong balance between efficiency and scalability, making them ideal for rapid decision-making in environments with limited features. In contrast, Support Vector Machine, XGboost and voting classifier excelled in offering more detailed, feature-rich analyses, which are particularly beneficial when handling complex data. Building on these findings, the plan is to apply the same algorithms to improve the system's overall accuracy and efficiency. By leveraging the strengths of each approach, the aim is to create a scalable, effective recommendation system tailored for real-world applications in the car industry. This study highlights the importance of choosing the right algorithm based on the tradeoffs between computational efficiency and the depth of analysis required in recommendation systems. 2025 IEEE.