Research Papers
Explore peer-reviewed, open-access research papers published instantly on PaperNova by students and scholars from over 60 countries. Every paper is citable, downloadable as PDF, and issued with a certificate of publication.
Looking to publish your own work? Submit your research paper and go live in minutes.
AI Powered Student Assistance Chatbot for Department of Technical Education
This paper presents Saviour, an AI-powered multilingual student assistance chatbot designed for institutes under the Directorate of Technical Education (DOTE), Maharashtra. The system uses a hybrid NLP approach combining rule-based FAQ matching with a large language model to deliver fast and accurate responses. It supports multiple Indian languages and integrates features such as result checking, timetable access, and syllabus retrieval. Experimental results show high accuracy (~91%), low response latency (<2 seconds), and improved student satisfaction, demonstrating its effectiveness in enhancing student support services.
By Nilima Mahendra Patil · Mumbai
An Intelligent Framework for AI-Based Crime Prediction and Real-Time Emergency Response in Smart Policing Systems
The increasing complexity of urban environments and rising crime rates demand intelligent, data-driven solutions for effective law enforcement. This paper presents an AI-driven Crime Management System that integrates machine learning techniques — Random Forest for crime classification and K-Means clustering for hotspot detection — to analyze historical and real-time crime data. A real-time SOS emergency module powered by Socket.IO enables instant communication between citizens and law enforcement with live location tracking. Additionally, the system incorporates automated FIR processing with risk-based prioritization. Experimental evaluation demonstrates significant improvements in response time (from 8–10 min to 2–3 min), crime prediction accuracy (≈90%), and resource allocation efficiency, providing a scalable and intelligent framework for smart policing.
By Omkar Rathod · Vishwaniketan’s Institute of Management Entrepreneurship and Engineering Technology (VIMEET)
AI-Powered Chatbot System for Tourism Ticket Booking
This research presents the design and comprehensive implementation of an intelligent, AI-powered chatbot-based ticketing ecosystem specifically engineered for the modern museum and cultural heritage sector. By synthesizing the capabilities of Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning (ML), the proposed system transcends the limitations of traditional, menu-driven booking platforms. The core objective of this study is to replace static web forms with a dynamic, conversational interface that facilitates real-time reservations, automated multi-turn query handling, and a frictionless user journey. At the architectural heart of the system lies a sophisticated integration of high-performance technologies. We utilize the Gemini 1.5 Flash generative model for advanced intent recognition and semantic understanding, ensuring the chatbot can interpret complex human requests with high contextual accuracy. This is supported by a robust Flask-based backend that serves as the central orchestrator for processing asynchronous HTTP requests and managing session states. For data persistence and real-time synchronization, the system leverages Firebase, a NoSQL cloud database that ensures high availability and scalability for concurrent user interactions. Security and transaction integrity are prioritized through the implementation of the Razorpay API, which provides a PCI-compliant gateway for encrypted online payments. Upon successful transaction verification, the system autonomously generates a unique, QR-coded digital ticket using specialized Python libraries, thereby eliminating the need for physical paper and reducing the environmental footprint of the ticketing process. Beyond mere utility, the system incorporates personalized recommendation logic to suggest exhibits based on user preferences, further enhancing the "smart tourism" experience. Experimental evaluation and rigorous stress testing indicate that this integrated approach significantly reduces response latency, minimizes human intervention, and maximizes booking throughput compared to conventional systems. The results demonstrate that the synergy of generative AI and cloud-native infrastructure creates a highly resilient and user-centric platform. This project serves as a definitive proof-of-concept for the practical application of Large Language Models (LLMs) in the tourism industry, highlighting a scalable pathway toward fully autonomous, intelligent service solutions.
By Mehul Ramdas Pawade · Mumbai University
DDAS: A Lightweight Hash-Based Framework for Real-Time Duplicate Detection in Multi-Tenant Web Applications
As the need for storage grows in multi-tenant web apps, redundant data ingestion is still a major problem. This paper presents the Data Download Duplicate Alert System (DDAS), a streamlined framework designed to alleviate storage inefficiencies via real-time duplicate detection. DDAS uses SHA-256 cryptographic hashing in a Flask-based backend to make unique digital signatures for each file that comes in. Before allocating storage, these hashes are checked against a central MySQL relational database. The results of the experiment show that DDAS works well to get rid of unnecessary uploads while keeping latency to a minimum. This study offers a scalable, low-overhead way to improve data integrity and make better use of storage in distributed environments.
By Basavakruti Umesh Mantale · University of Mumbai
MindMitra: An NLP and Sentiment Analysis-Driven AI Chatbot for Mental Wellness Support in the Indian Context
Mental health support remains inaccessible to millions due to high costs, stigma, and lack of professionals. This paper presents MindMitra, an AI-powered chatbot that uses Natural Language Processing (NLP) and multi-class sentiment analysis to provide real-time, empathetic, and culturally adapted mental wellness support. The system employs a BERT-based model for six-class emotion classification (Anxiety, Sadness, Anger, Happiness, Neutral, Crisis), integrated with a CBT-based recommendation engine and GPT-based response generation. A key feature is passive emotion detection, eliminating the need for explicit mood input. The system also incorporates an India-specific cultural lexicon to improve relevance. Experimental evaluation shows significant improvements in mental well-being, including a 62% reduction in anxiety and 63% reduction in depression, with 91.4% classification accuracy. The results highlight the potential of AI in scalable and stigma-free mental health care.
By Vishruti Santosh Murudkar · Vishwaniketan Institute of Management Entrepreneurship & Engineering Technology
Academic Assistant AI Platform Integrated Using Chatbot.
The Academic Assistance AI Platform is designed to revolutionize student support through the use of intelligent conversational artificial intelligence. This platform leverages advanced natural language processing and large language models (LLMs) to provide real-time, context-aware academic guidance. Students can interact with the AI assistant to resolve academic queries, access learning materials, check schedules, and receive personalized assistance. Developed using Python, Flask, MongoDB, and DeepSeek LLM API, the system seamlessly integrates AI with college data systems to deliver fast, reliable, and efficient responses. This report outlines the motivation, design, implementation, and future scope of the project, emphasizing its potential to enhance academic engagement and accessibility. This paper details the research, design, implementation and evaluation of the platform, demonstrating its potential to revolutionize student support service.
By Soham Sanjay Kank · Vishwaniketan's Institute of Management Entrepreneurship and Engineering Technology.
Precision Agriculture Using Machine Learning
Precision agriculture uses machine learning and technologies like IoT sensors, drones, and satellite data to improve farming efficiency. It helps farmers monitor soil, weather, and crop health in real time, enabling better decisions for irrigation, disease control, and yield prediction. This data-driven approach reduces resource waste, lowers costs, and increases productivity. Overall, it supports sustainable farming and improves food security.
By Keshav Kolekar · University of Mumbai
Farmatics
Plant diseases are one of the leading causes of agricultural losses worldwide, annually destroying an estimated 20–40% of global crop production and threatening the food security of billions. Early and accurate identification of plant diseases is critical to enabling timely intervention and minimizing crop losses. Farmatics is a deep learning-based plant disease detection system specifically designed to assist farmers in identifying diseases at an early stage across multiple crop varieties directly from leaf photographs captured via a mobile or web interface. The system leverages a custom Convolutional Neural Network (CNN) architecture trained on the PlantVillage benchmark dataset comprising 54,306 images across 38 disease classes and 14 crop varieties. Unlike existing solutions that are limited to single-crop detection, require constant internet connectivity, or lack early-stage detection capability, Farmatics supports multi-crop inference and is optimized for practical agricultural deployment. Experimental evaluation demonstrates that the proposed system achieves 96.4% classification accuracy, outperforming established baselines including VGG16 (91.2%), ResNet50 (92.5%), MobileNetV2 (89.7%), and InceptionV3 (93.1%). The system further provides confidence-scored disease identification, crop-specific treatment recommendations, and a confidence thresholding mechanism for quality assurance. Farmatics represents a significant step toward democratizing precision agricultural technology for smallholder farmers in India and comparable developing economies. Index Terms — Plant Disease Detection, Convolutional Neural Network, Deep Learning, Precision Agriculture, Image Classification, Early Detection, Multi-Crop Analysis, PlantVillage, Transfer Learning, Computer Vision.
By Ayush Santosh Naik · Vishwaniketan Institute ,khalapur
Suraksha Setu: AI-Driven Criminal Identification Using Facial Recognition Technology with Aadhaar Integration for Scalable Intelligent Public Safety in Urban India
Suraksha Setu is an AI-powered dual-platform public safety ecosystem designed to address the critical operational gaps in Indian law enforcement. The system integrates Facial Recognition Technology (FRT) using the DeepFace framework (VGGFace2 backbone), real-time multi-class object detection via YOLOv8 fine-tuned on a 14,200-image custom dataset, and biometric identity verification through the UIDAI Aadhaar Authentication API 2.5. The architecture comprises two primary subsystems: (1) a Citizen Safety Portal providing geofence-triggered crime alerts via Firebase Cloud Messaging, SOS GPS triangulation, and a community gamification layer; and (2) a Police Operational Dashboard delivering live AI-monitored CCTV analytics, LIDAR/UGS-based restricted-zone intrusion detection, crime heatmap visualization via MongoDB GeoJSON, and centralized Aadhaar-linked offender database access. Comprehensive evaluation demonstrates that DeepFace achieves 94.8% facial recognition accuracy under operational CCTV conditions, surpassing FaceNet (91.7%) and ArcFace (89.3%). YOLOv8 attains a mean Average Precision (mAP@0.5) of 91.3% across five threat categories, with weapon detection reaching 92.1% F1-score. System response latency averages 1.2 seconds for SOS dispatch and 2.8 seconds for FRT matching — a 75–80% reduction over conventional systems. An ablation study confirms that each architectural component contributes measurably to system performance. A security assessment across seven threat vectors demonstrates compliance with NIST SP 800-175B, UIDAI regulations, and OWASP Top-10 standards. Suraksha Setu represents a replicable, nationally scalable framework for AI-integrated, constitutionally compliant smart policing
By Sachin Chaurasiya · Vishwaniketan iMEET
About PaperNova's research archive
PaperNova is an international platform where students and researchers can publish research papers online instantly and receive a verifiable certificate of publication. Our open-access archive covers engineering, computer science, artificial intelligence, biotechnology, management, finance, and more.
Every paper on PaperNova is peer-reviewed, citable, and indexed on the open web so it can be discovered by future researchers. If you are looking for the fastest way to publish a journal paper internationally, start by browsing our AI research papers, machine learning papers, or computer science archive.