Daily Archives: January 27, 2026

Authors: Sangeeta Rani, Vikram Singh, Tanisha Mittal

Abstract: Customer churn poses a significant challenge to businesses, necessitating robust predictive solutions. We propose a novel hybrid stacking framework that integrates four diverse base classifiers—logistic regression (LR), random forest (RF), artificial neural network (ANN), and XGBoost—with a meta-learner to enhance churn prediction performance. In the first stage (Level 0), the base models independently learn from preprocessed customer behaviour and demographic features, capturing both linear and non-linear patterns. Their predicted class probabilities subsequently serve as input features to a deep feedforward neural network at Level 1, which functions as the meta-learner. This architecture is trained using categorical cross-entropy loss with the Adam optimiser, incorporating dropout to mitigate overfitting. The stacking ensemble leverages the complementary strengths of the base models (e.g., interpretability from LR, decision-boundary flexibility from RF, complex pattern recognition from ANN, and from XGBoost to achieve superior predictive accuracy and generalisation compared to any individual classifier. Experimental results on a real-world churn dataset demonstrate that the hybrid model consistently outperforms traditional baselines, achieving statistically significant improvements in AUC and F1-score. The findings suggest that stacking heterogeneous learners with a deep meta-model provides a powerful methodology for addressing the complexities of churn prediction.

DOI: http://doi.org/10.5281/zenodo.18385968

Published by: vikaspatanker

Authors: Neh Sharma

Abstract: GenAI has altered how businesses think about data and how they use it to make decisions. After 2020, better large language models (LLMs), retrieval-augmented generation (RAG), and agentic pipelines have made it possible for analytics systems to go from only reporting on data to coming up with fresh insights. But this change makes people very worried about fairness, openness, and data privacy, especially since models affect how businesses make decisions and how people from different backgrounds work together. This paper looks at new developments in architecture and talks about the ongoing ethical and evaluative problems that come up in generative analytics. A single Generative Analytics System Framework (GASF) is proposed, integrating architectural, evaluative, and ethical dimensions to achieve a balance between analytical efficacy and accountability. A simulation demonstrates that various departments and demographic user groups utilise LLM-based analytics in distinct manners. The findings indicate that user skill and contextual diversity influence factual accuracy, delay, and trust in distinct ways. This means we need to make systems that are fair and keep people's information safe. The report concludes with a proposal for research aimed at developing generative analytics ecosystems that are ethical, comprehensible, and adaptable to diverse populations.

Published by: vikaspatanker

Authors: Narendra Reddy Burramukku

Abstract: The increasing adoption of multi-vendor network architectures in enterprise, data center, cloud, and hybrid environments has significantly enhanced flexibility, cost efficiency, and technological innovation. However, the heterogeneity of hardware, software, firmware, and management interfaces across vendors introduces substantial challenges in maintaining a consistent and resilient security posture. Security vulnerability management in such environments is particularly complex due to interoperability limitations, asynchronous patch cycles, fragmented monitoring systems, and inconsistent policy enforcement. This paper presents a comprehensive review of security vulnerability management strategies in multi-vendor network environments, focusing on vulnerability identification, classification, prioritization, and remediation. It examines traditional and modern vulnerability assessment techniques, including automated scanning, penetration testing, threat modeling, and standardized vulnerability databases. The study further analyzes vulnerability management frameworks encompassing patch management, policy-based security, integration with SIEM and threat intelligence platforms, and automation through orchestration. Key challenges related to scalability, real-time monitoring, compliance, and governance are critically discussed. Performance and effectiveness metrics such as remediation time, detection accuracy, operational efficiency, and risk reduction are evaluated to assess practical deployment feasibility. Emerging approaches, including AI- and ML-driven vulnerability management, zero-trust architectures, micro-segmentation, blockchain-based security mechanisms, and cloud-native platforms, are explored as potential solutions to existing limitations. By synthesizing current research, identifying literature gaps, and outlining future research directions, this review provides a structured reference for researchers, network architects, and security practitioners seeking to enhance vulnerability management in complex, heterogeneous network infrastructures.

DOI: https://doi.org/10.5281/zenodo.18383985

 

Published by: vikaspatanker

Authors: Narendra Reddy Burramukku

Abstract: The rapid expansion of large-scale computer networks has introduced significant complexity in managing diverse network configurations. Manual classification and analysis of configurations are time-consuming, error-prone, and increasingly infeasible in dynamic environments. This paper presents a novel framework for automated classification of large-scale network configurations using machine learning combined with semantic vectorization. Network configuration files are first pre-processed and transformed into high-dimensional vector representations that capture both semantic and hierarchical relationships among configuration commands, protocols, and policies. These embeddings serve as input to supervised machine learning models, including Random Forest, Support Vector Machines, and Neural Networks, enabling accurate classification of network devices, roles, and compliance profiles. Experiments are conducted on real-world enterprise, cloud, and synthetic network datasets, comprising thousands of configuration files with diverse structures and device types. Results demonstrate that the proposed framework significantly outperforms traditional rule-based and feature-based approaches, achieving up to 94.5% F1-score with graph-based embeddings. Scalability analysis indicates the method can efficiently handle large volumes of configurations while maintaining high accuracy. The study highlights the effectiveness of semantic vectorization in capturing complex configuration semantics and facilitating robust automated classification. This framework provides a foundation for intelligent, scalable network management, supporting proactive policy enforcement, misconfiguration detection, and operational efficiency. Future work explores real-time classification, integration with network orchestration systems, and transformer-based embeddings for richer semantic representation.

DOI: https://doi.org/10.5281/zenodo.18383730

 

Published by: vikaspatanker

Authors: M. Gowsalya,, N. Devapriya, K. Abinaya

Abstract: In the modern digital era, the increasing demand for intelligent monitoring systems has become a critical concern across domains such as healthcare, surveillance, and smart environments. Conventional monitoring approaches primarily rely on single- modality data sources, which often limit their accuracy, reliability, and adaptability in real-world conditions. To address these limitations, this paper proposes a Smart Multimodal Analysis System (SMAS) that integrates multiple data modalities, including visual, audio, sensor, and textual information, into a unified intelligent framework. The proposed system leverages advanced machine learning and deep learning techniques to perform real-time data acquisition, preprocessing, feature extraction, and multimodal fusion. By combining information at both feature and decision levels, SMAS enhances detection accuracy and robustness, even in the presence of noisy or incomplete data. The system supports intelligent classification, anomaly detection, and predictive analysis, enabling timely alerts and informed decision-making. Experimental evaluation demonstrates that the multimodal approach outperforms traditional single-modality systems in terms of accuracy and reliability. The results highlight the potential of SMAS as an effective and scalable solution for next-generation smart monitoring applications.

DOI: https://doi.org/10.5281/zenodo.18410791

 

Published by: vikaspatanker