Authors: Rayapudi Gautam Kumar

Abstract: MO-NAS: Multi-Objective Neural Architecture Search Using NSGA-IIThis paper presents MO-NAS, a production-ready framework for automatically discovering optimal neural network architectures using the Non-dominated Sorting Genetic Algorithm II (NSGA-II) multi-objective optimization ap‐ proach. Unlike traditional Neural Architecture Search (NAS) methods that optimize for a single objective (typically accuracy), MO-NAS simultaneously optimizes multiple competing objectives including accuracy, computational cost (FLOPs), model size (parameters), inference latency, and memory footprint. The framework supports multiple data modalities including image, text, sequence, and tabular data, making it a universal NAS solution. We incorporate advanced techniques such as zero-cost proxies for rapid evaluation, Bayesian guidance for search efficiency, and weight sharing to reduce training costs. Our approach produces a Pareto-optimal front of architectures, allowing practitioners to select the best trade-off for their specific deployment constraints.

 

Published by: vikaspatanker

Authors: Nitish Kumar Bharadwaj

Abstract: Decision-making in contemporary organizations is increasingly complex due to technological advances, volatile markets, and uncertainty in social, political, and economic environments. Relying solely on intuition or experience often leads to inefficient allocation of scarce resources and costly managerial errors. Operations Research (OR) provides a scientific and quantitative framework that supports rational decision-making through modeling, analysis, and optimization. This paper reviews the historical evolution of OR, clarifies conceptual foundations, and explains the processes through which OR transforms real-world problems into structured decision models. Different classes of models, deterministic, probabilistic, static, dynamic, descriptive, and normative are examined with reference to their applicability and limitations. The paper further highlights implementation challenges and discusses how computing advances have broadened the scope of OR in domains such as supply chains, healthcare, energy, and public policy. The study contributes by presenting a synthesized framework that links classical OR principles with contemporary decision environments, emphasizing how quantitative modeling can improve managerial effectiveness and support evidence-based decisions. Recommendations for future research and practice are also discussed.

DOI: https://doi.org/10.5281/zenodo.18081819

Published by: vikaspatanker

Authors: Nithish Kumar R, Gokul Kanna Sm

Abstract: Smart Healthcare represents a transformative shift in modern medical systems by integrating artificial intelligence (AI), machine learning (ML), deep learning, and Internet of Things (IoT) technologies into healthcare delivery. Early disease detection, accurate diagnosis, and personalized treatment remain critical challenges in healthcare systems worldwide. Traditional healthcare practices largely rely on manual diagnosis, clinician expertise, and time-consuming diagnostic procedures, which may lead to delayed detection, human error, and increased healthcare costs. With the rapid growth of AI and medical imaging technologies, automated disease detection and health monitoring systems have gained significant attention. Medical images such as X-rays, MRI scans, CT scans, ultrasound images, and skin lesion images contain rich visual information that can be effectively analyzed using machine learning and deep learning techniques. This paper presents an intelligent Smart Healthcare framework that utilizes AI-driven medical image analysis for early disease detection and clinical decision support. The proposed system includes image acquisition, preprocessing, feature extraction, disease classification, and result visualization. Experimental studies indicate that AI-based healthcare systems significantly improve diagnostic accuracy, reduce workload on healthcare professionals, and enhance patient outcomes. The system aims to support early diagnosis, reduce medical errors, optimize treatment planning, and promote efficient and patient-centric healthcare services.

 

 

Published by: vikaspatanker

Authors: Sukesh G, Sanjay S

Abstract: Early detection of crop diseases is a critical requirement for ensuring agricultural productivity, food security, and economic stability for farmers. Crop diseases caused by fungi, bacteria, viruses, and pests often spread rapidly and remain unnoticed during their initial stages, leading to severe yield loss and financial damage. Traditional crop disease detection methods rely on manual inspection by farmers or agricultural experts, which is time-consuming, subjective, and often inaccurate due to human limitations and environmental variations. Moreover, expert support is not always accessible to farmers in rural and remote areas. With the rapid advancement of machine learning and image processing technologies, automated crop disease detection systems have gained significant attention in recent years. Leaf images contain rich visual information such as color variation, texture patterns, and shape irregularities that can be effectively analyzed using computer vision techniques. This paper presents an automated crop disease detection framework using machine learning and image processing techniques to identify plant diseases at an early stage. The proposed system involves image preprocessing, feature extraction, and classification using both traditional machine learning algorithms and deep learning models. The system aims to reduce crop loss, minimize excessive pesticide usage, and assist farmers in making timely and informed decisions. Experimental evaluation demonstrates that the proposed approach achieves improved accuracy and reliability, making it suitable for real-world agricultural applications.

 

 

Published by: vikaspatanker

Authors: Dharunika G, Jack Robin J

Abstract: The increasing scale and dynamic complexity of modern cloud computing environments pose significant challenges for ensuring system reliability and availability, making traditional manual fault diagnosis and recovery insufficient. This paper explores advanced methodologies, contrasting established statistical monitoring techniques with emerging Artificial Intelligence (AI)-driven autonomous self-healing frameworks designed to manage faults, minimize downtime, and optimize resource utilization. Early methods employed correlation analysis using Canonical Correlation Analysis (CCA) and Exponentially Weighted Moving Average (EWMA) control charts for anomaly detection, followed by feature selection using ReliefF and SVM-RFE for problem location. The evolution toward AI-driven solutions leverages machine learning (ML), deep learning (DL), and reinforcement learning (RL) to achieve automated failure prediction and recovery. Recent advancements feature hybrid AI models and self- reflective multi-agent systems (SR-MACHA) that demonstrate enhanced accuracy and self- optimization capabilities, achieving significant reductions in Mean Time to Repair (MTTR) and improving system resilience. However, challenges remain regarding the scalability, computational cost of training complex models, and ensuring real-time performance in dynamic, heterogeneous cloud infrastructures.

Published by: vikaspatanker

Authors: Semran Ojha, Professor Rahul Patidar, Professor Jayshree Boaddh

Abstract: The rapid expansion of the Internet of Things (IoT) has created significant challenges in managing computation and data processing efficiently. As billions of interconnected devices generate massive workloads, traditional cloud infrastructures experience latency and performance bottlenecks. To address these limitations, this research introduces an intelligent edge-level load distribution model using Random Forest techniques. The proposed system leverages edge computing to process tasks closer to data sources, thereby reducing dependency on centralized cloud servers. The Random Forest algorithm is utilized to learn from historical task patterns and predict optimal job scheduling sequences. This is integrated with a wolf optimization strategy that dynamically adjusts load distribution across heterogeneous edge nodes without prior training requirements. Experimental evaluation conducted using MATLAB demonstrates that the proposed model effectively minimizes makespan by 0.79% and enhances edge utilization by 16.25% compared to the existing Preference-Based Stable Matching (PBSM) model. These improvements confirm that machine learning- driven edge load balancing can significantly improve resource allocation, task completion time, and overall network efficiency in large-scale IoT environments.

Published by: vikaspatanker

Authors: Gayatri dhasade, Soham vishe, Soham Bhintade, Darshan Bhamare, Prof.Poonam Chavan

Abstract: An academic event management system is A digital platform designed to plan, organize, manage and monitor academic and non-academic events held at educational institutions. College Event Management System is a digital platform designed to plan, organize, manage, and monitor academic and non-academic events conducted in educational institutions. The Colleges regularly organize seminars, workshops, cultural programs, sports events, technical competitions and guest lectures. competitions, and guest lectures. Handling these events manually using paperwork or scattered communications often results in inefficiency, communication issues, and data loss. Managing these events manually using paperwork or scattered communication often leads to inefficiency, miscommunication, and data loss. The proposed system provides a centralized solution to manage event creation, registration, scheduling, notifications and reporting. notifications, and reporting. It allows students, faculty and event coordinators to interact through a single platform, improving transparency and coordination. enables students, faculty, and event coordinators to interact through a single platform, improving transparency and coordination. By automating event workflows, the system reduces administrative workload, ensures timely communication and improves overall event execution. communication, and enhances overall event execution.

 

Published by: vikaspatanker

Authors: Aadhithyan K, Pranauv Raaj N

Abstract: Card-based transactions and modern digital payment systems face sophisticated and rapidly evolving security threats, necessitating advanced fraud detection methods. Traditional approaches, often reliant on fixed rules and descriptive analytics, are slow to adapt to new fraud schemes and struggle with the volume of real-time transactions. This presentation analyzes the effectiveness of AI-driven fraud detection, specifically focusing on the integration of Real-Time Analytics, Deep Learning (DL), and Agentic AI systems to enhance security and prevent financial losses. The study highlights that DL models, such as hybrid Recurrent Neural Networks (RNNs) combined with attention mechanisms, offer superior performance by modeling sequential data and addressing challenges like data imbalance. Furthermore, adopting the Deep Learning–Sector–Governance (DLSG) framework is crucial, as it ensures that technical innovations are aligned with sector-specific constraints and regulatory requirements, such as the need for explainability and data privacy. The synthesis of these technologies provides a proactive, adaptive solution to safeguard complex financial ecosystems.

Published by: vikaspatanker

Authors: Shubhangi Sondhiya, Deepesh Malviya

Abstract: These seismic risks are considered the prime cause for concern in seismic zones and earthquake-prone areas around the world. Over time, a sequence of earthquake motions with different seismic intensity has been used to conduct the investigation and analyze the structural dynamics. Analyses considering the effect of isolated structures showed that the isolators restrict the lateral loads transmitted to the structure, which, in turn, has the tendency to reduce the sizes of building components. In this study, design, operation, testing, and applicability of base isolation are analyzed in detail as per Indian Standards. Base isolation has been found to be one of the popular design approaches in recent times. A building structure is taken as a case study model for this study, and contemporary design tools are also applied for the analysis. Conclusions are drawn from the results obtained. We'll discuss the probable advantages of base isolation over the conventional dynamic analysis. This chapter deals with the design details of the models and step followed to design the building. This study is conducted on a G+11-story building located on soft soil. The structure is designed as a college building (with plan irregularity) situated in Seismic Zone V and analyzed using ETABS 2022. For the analysis, two models are considered: Model M-1, which has a fixed base, and Model M-2, which incorporates a base isolator. The comparative analysis between the fixed-base model (M–1) and the base-isolated model (M–2) clearly demonstrates the effectiveness of base isolation in improving seismic performance. The base-isolated structure shows reduced base shear, displacement, and overturning moments by approximately 25–30%, indicating enhanced stability and safety. Although story drift slightly increases due to controlled base movement, this behavior helps in dissipating seismic energy and reducing damage to the superstructure. Overall, base isolation significantly enhances structural resilience, minimizes earthquake-induced forces, and provides an efficient and reliable solution for earthquake-resistant design in multi-storey buildings. Overall, the comparative study clearly demonstrates that the implementation of base isolation considerably enhances the seismic performance of structures. It reduces base shear, displacement, and overturning moments while allowing controlled drift, thereby ensuring improved safety, flexibility, and durability. These outcomes confirm that base isolation is a highly effective and reliable strategy for seismic risk mitigation in multi-storey buildings.

DOI: http://doi.org/10.5281/zenodo.18066513

Published by: vikaspatanker

Authors: A. Mohamed Sikkander, Joel J. P. C. Rodrigues, Manoharan Meena

Abstract: Artificial Intelligence (AI), also referred to as machine learning (ML) or deep learning (DL), is rapidly revolutionizing cancer treatment by using genomic information for improving diagnosis, prognosis, treatment decision, and drug discovery. Being a result of genetic and molecular changes, it is important to understand cancer’s genomic patterns and profiles. In conventional genomic analyses, common methodologies fail to handle high-dimensional genomic data produced from next-generation sequencing (NGS) and multi-omics platforms; on the other hand, AI approaches excel in detecting intricate patterns from large genomic datasets. This AI system trained from a large public genomic database such as ‘The Cancer Genome Atlas (TCGA),’ ‘Genomic Data Commons (GDC),’ or generally from the ‘Catalogue of Somatic Mutations in Cancer (COSMIC)’ has already facilitated accurate classifications of cancer subtypes and their treatment predictions or discovery of effective biomarkers for treatment of cancer subtypes that are accurate to a great extent. Deep learning from somatic mutation sequences showed an accuracy of approximately 0.98 for clinical biomarkers such as microsatellite instability (MSI), which is a considerably high improvement over other existing methodology. Integration of AI with multi-omics genomic, transcriptomic, proteomic data types further helps to increase efficiency of predictions regarding patient outcomes. Though AI is a revolution in genomic study thereby bringing a revolution in cancer treatment approaches following a detailed precise treatment decision of cancer treatment from an individual’s genomic study followed by inducing a global revolution in cancer treatment true to precision medicine practices around the world.

DOI: http://doi.org/10.5281/zenodo.18066233

Published by: vikaspatanker