Category Archives: Uncategorized

Authors: Vikash, Mrs Sheela Malik, Mr Atul Dubey

Abstract: Earthquakes are one of the most devastating forces on the planet. The seismic waves that travel through the ground can demolish buildings, kill people, and cost billions of dollars in damage and restoration. According to the National Earthquake Information Centre, there are over 20,000 earthquakes every year on average, including 16 major disasters. The damage was caused by the collapse of buildings with people inside, as in previous earthquakes, prompting the development of earthquake-resistant constructions. Constructions intended to withstand earthquakes are known as earthquake-resistant structures. While no structure can be completely safe from earthquake damage, earthquake-resistant construction aims to build structures that perform better than their conventional equivalents during seismic activity. Building rules state that earthquake-resistant constructions must be able to withstand the greatest earthquake with a reasonable chance of occurring at their site. In this paper we are marking comparison between Static and dynamic analysis. It is concluded from paper Dynamic analysis is economical as compared to static analysis.

DOI: http://doi.org/10.5281/zenodo.20817202

Published by: vikaspatanker

Authors: Research Scholar Namrata Ramrao Pawar, Dr. Ganesh R. Teltumbade

Abstract: Precise stock prediction and efficient portfolio optimization are still problematic in practice because of the non-stationarity, volatility, and complexity of the financial time series. In this study, we present a Hybrid Artificial Intelligence Framework (HAIF), which is the combination of three different frameworks: (1) a Graph Neural Network (GNN) with an attention layer for modelling relationships between stock prices; (2) a Transformer with convolutional layers for predicting price movements in different periods ahead; and (3) a Deep Reinforcement Learning (DRL) model called Proximal Policy Optimization for managing transactions and balancing the portfolio in different conditions. Based on 5 years of daily S&P 500 time series from 2020 to 2025 with 50 constituent stocks, our model obtains Sharpe ratio = 1.84, annual return = 28.4%, and maximum drawdown = -11.2% while outperforming

DOI: https://doi.org/10.5281/zenodo.20811838

Published by: vikaspatanker

Authors: Assistant Professor Mr Devendra Kumar Pandey, Assistant Professor Dr. Swarna Surekha

Abstract: However, the growing problem of the size of deep learning models has brought the issue of energy use, in which a single large transformer model can produce over 500 metric tons of CO₂ equivalent when training. We, in this work, propose the first green awareness framework, named GreenAI-Framework, that alters the precision level of a given model, making it sparse, and scheduling its computations on low-carbon energy sources using the carbon intensity signal. There are three proposed algorithms in our proposed framework, and these are as follows: (1) Adaptive Precision Scaling (APS) with the use of reinforcement learning to decrease the number of FLOPS between 40% to 60% with no accuracy cost, (2) Energy Aware Early Exiting (EAEE) to exit from low confidence inference requests, and (3) Carbon-Aware Task Scheduling (CATS) for executing non-urgent tasks in low-carbon energy slots. Experimental analysis demonstrates that our framework helps reduce energy use by 47.3%, having only 0.9% loss in accuracy for ResNet-50, BERT, and GPT-2 on GPU clusters.

DOI: https://doi.org/10.5281/zenodo.20811568

Published by: vikaspatanker

Authors: Mayuresh More, Piyush Punchmukhe , Shantanu Wagh , Rajashree kumbhar

Abstract: Recent developments in conversational Artificial Intelligence have ushered in a new era in the field of natural and easy man–computer interaction. Traditional desktop interfaces are sometimes cumbersome to navigate and operate using the keyboard, potentially making them less accessible and less efficient to use. This paper introduces AURA, an intelligent voice-driven desktop assistant that combines the capabilities of speech recognition, intent understanding using the large language model (LLM), desktop automation, and adaptive voice feedback into a single system. The system uses wake word detection, speech recognition, context-based intent understanding and automatic command execution to allow for hands-free interaction with the desktop. AURA can be used to manage applications, navigate websites, manipulate text, retrieve information and provide conversational assistance using natural language commands. The proposed architecture is realized in Python, with the voice processing and intent analysis modules, command execution, and user interaction management being modularized. Experimental assessment shows that the system interprets the commands accurately, responds quickly to the user and is more user-friendly than traditional command-based systems. The results demonstrate the promise of voice assistants that are powered by LLM for intuitive and inclusive computing experiences. Key Words: Large Language Models, Voice Assistants, Desktop Automation, Human–Computer Interaction, Speech Recognition, Conversational AI, Accessibility.

Published by: vikaspatanker

Authors: Associate Professor Dr. Baby. M. S

Abstract: This study explores the causal connection between financial literacy (FL) and investment decision quality among urban Indian young adults aged 18 to 30 years. This mixed-methods study uses a structured survey sample of 450 participants and an experimental investment decision scenario involving 120 individuals to assess the influence of budgeting skills, risk assessment knowledge, compound interest awareness, and digital financial skills on decision making. We find a significant positive correlation (r = 0.72, p < 0.001) between FL score and prudence, defined in terms of diversification of portfolio, risk-return tradeoff, and non-speculative nature of investment decisions. Only 34% of individuals could understand compound interest, and 58% failed to define mutual funds. Comparing investment decision quality in four FL intervention groups, we find that gamified simulation was significantly more effective than traditional lecture-based instruction in improving decision quality by 41%.

DOI: https://doi.org/10.5281/zenodo.20811288

Published by: vikaspatanker

Authors: Swami Bhagwat, Shreyash Vidhate, Darshan Shinde, Professor G. K. Gaikwad

Abstract: Human Activity Recognition (HAR) focuses on automatically identifying human actions from video streams or sensor data using computer vision and machine learning techniques. With the rapid growth of intelligent healthcare, surveillance, and smart automation systems, HAR has become an important research area. This paper presents a redesigned and implementation-oriented study of a HAR system built using OpenCV and modern learning models. The work explains the complete pipeline including video acquisition, preprocessing, feature extraction, and activity classification. Instead of relying only on theoretical descriptions, the paper emphasizes a practical modular architecture and real-time considerations. The role of deep learning models combined with OpenCV preprocessing is discussed along with system challenges such as lighting varia- tion, occlusion, and computational cost. The proposed approach highlights how lightweight processing and hybrid models can support accurate and efficient recognition suitable for real-world deployment.

DOI: https://doi.org/10.5281/zenodo.20809909

Published by: vikaspatanker

Authors: Saurav khambe, Suchipriya Malge, Harshit Mishra, Ayushi Chinde, Sakshi Jadhav

Abstract: This paper presents an edge-based smart border surveillance system integrating multi-modal sensing with lightweight deep learning for real-time intrusion detection. The system combines a Raspberry Pi 5, PIR motion sensor, KY-037 acoustic sensor, and NoIR camera in an event-driven architecture. Motion or abnormal sound triggers visual analysis using a TensorFlow Lite–optimized YOLOv5 model deployed for on-device inference. Experimental evaluation across 10 controlled scenarios under daytime and low-light conditions achieved an overall detection accuracy of 80%, with precision and recall of 0.89 for human and vehicle detection. The measured end-to-end latency ranged from 1.6–1.9 s. Average CPU utilization during inference was 55–60%, with peak usage of 72%, and total power consumption measured 6–8 W during active operation. The decision-level sensor fusion approach reduced unnecessary visual processing and minimized false activations compared to continuous vision-based monitoring. The system operates entirely at the edge without cloud dependency, enabling low-latency and bandwidth-efficient deployment in remote border environments.

DOI: https://doi.org/10.5281/zenodo.20809718

Published by: vikaspatanker

Authors: Ms. Vaishnavi Kishor Patil, Ms. Rutuja Rajkumar Waghmare, Ms. Priyanka Ubhad, Mr. Ramgopal Sahu

Abstract: Material handling plays a crucial role in industrial automation, warehouse management, and logistics operations. Conventional transportation methods based on manual carts and trolleys require significant human effort, resulting in reduced operational efficiency, increased labor dependency, and limitations in confined working environments. To address these challenges, this paper presents OmniLiftBot, an autonomous Mecanum-wheeled robotic platform designed for smart load transportation, vertical load handling, and real-time payload monitoring. The proposed system is built around an ESP32 microcontroller and integrates Mecanum wheels for omnidirectional mobility, a scissor-lift mechanism for controlled elevation of loads, a load cell with HX711 module for weight measurement, and an ultrasonic sensor for obstacle detection. A Wi-Fi-based interface enables path selection and system control, allowing flexible operation in indoor environments. The combination of mobility, lifting, sensing, and monitoring functionalities within a single platform enhances the versatility of the system while reducing manual intervention. Experimental evaluation of the developed prototype demonstrates reliable navigation, effective lifting operation, accurate payload monitoring, and safe obstacle detection under controlled conditions. The proposed solution offers a compact, cost-effective, and scalable approach for modern material handling applications and can serve as a foundation for future intelligent warehouse automation systems.

Published by: vikaspatanker

Authors: Tejas Rajendra Moule, Kunal Sanjay Patekar, Abhay Ramesh Mishra, Assistant Professor Ganesh Keshav Gaikwad

Abstract: The Kumbh Mela, one of the largest human congregations on Earth, presents significant challenges in crowd management, health response, and transportation logistics. Traditional management systems[1] often rely on manual surveillance and limited communication mechanisms, which are insufficient for real-time risk detection and decision-making. Kumbh Connect proposes a comprehensive AI-powered framework that integrates computer vision, predictive analytics, Internet of Things (IoT) sensors, and natural language processing to ensure safety, efficiency, and improved pilgrim experience. This paper surveys current research and technologies relevant to large-scale event management, identifies key challenges, and outlines potential directions for future development toward a more intelligent, connected, and secure Kumbh Mela environment.

DOI: https://doi.org/10.5281/zenodo.20808940

Published by: vikaspatanker

Authors: Gaurav Kankuse, Jay Deshmukh, Om Borse, N. D. Dhamale

Abstract: This project focuses on building a smart, easy- to-use Grammatical Error Correction (GEC) system for a native Indic language, specifically Marathi. The system leverages modern transformer-based AI models, such as IndicBERT and mBART, which are fine-tuned using local language data. The primary objective of the system is to identify and correct grammatical errors in sentences in real time. The proposed solution includes a simple web-based tool where users can input text, view suggested corrections along with brief explanations, and choose which changes to accept or reject. The study outlines the system design to automated text analysis. Preliminary observations indicate the feasibility of the proposed approach, with future work focusing on extensive experimental validation.

DOI: https://doi.org/10.5281/zenodo.20808637

Published by: vikaspatanker