Daily Archives: May 5, 2026

Authors: Mrs. K. Senbagam, Dhanush S, Gopinathan S, Dilli Babu K

Abstract: Flooding is one of the most severe natural hazards, leading to significant losses in human life, infrastructure, and economic resources, particularly in flood-prone regions such as India. Rapid and reliable identification of inundated areas is essential for effective disaster response, mitigation planning, and resource allocation. Conventional flood mapping techniques are often labor-intensive, time-consuming, and limited by environmental constraints. In particular, optical satellite imagery is highly affected by cloud cover and poor visibility during extreme weather conditions. To address these limitations, this study proposes an automated flood assessment framework utilizing satellite-based remote sensing data. The approach primarily leverages Synthetic Aperture Radar (SAR) imagery, which enables consistent data acquisition irrespective of weather conditions or illumination. The proposed framework integrates image preprocessing, change detection, and region extraction techniques to identify flood-affected areas by analyzing temporal variations between pre-event and post-event images. The system is designed to efficiently highlight newly formed water bodies and quantify flood impact through statistical and visual outputs. A web-based interface is incorporated to enhance accessibility and interpretation of results. Experimental observations demonstrate that the proposed method provides reliable flood detection across diverse terrains, including urban and vegetation-covered regions. This work contributes toward developing a scalable and efficient solution for large-scale flood monitoring, supporting timely decision-making and improving disaster management strategies.

Published by: vikaspatanker

Authors: Mr. D. Harsha, N. Soumya, G. Nandavardhan Reddy, K. Sai Sreesh

Abstract: A novel protection coordination approach utilizing artificial neural networks (ANNs) is introduced in this work for meshed high-voltage transmission systems. Existing overcurrent and distance relay coordination methods in meshed topologies are prone to relay blinding, zone overreach, and incorrect operation during power swing events. The developed ANN model is trained using an extensive fault scenario dataset generated through simulation of a 9-bus, 230 kV benchmark network in MATLAB/Simulink. The proposed architecture—with 18 inputs, three hidden layers containing 36, 24, and 12 neurons respectively, and a 9-output trip signal layer—delivers improved coordination speed, selectivity, and sensitivity over traditional relay configurations. Testing results demonstrate a fault classification accuracy of 98.54% on previously unseen data. On average, fault clearance times are shortened by 56.8% in comparison to conventional coordination approaches, and dependable detection of high-impedance faults is also achieved. The approach provides a flexible and adaptive protection solution well-suited to contemporary interconnected power grids.

DOI: https://doi.org/10.5281/zenodo.20045310

Published by: vikaspatanker

Authors: Swati Pawar, Shreyash Karpe, Thanshu Agarkar, Mohit

Abstract: In today’s world, where we’re flooded with information, having a smart and efficient search system is more important than ever. Traditional search engines like Google rely on keywords and fixed ranking systems such as PageRank. While these methods work well, they often fail to truly understand what a user means, handle complex multi-step questions, or deliver deeply personalized results beyond just rewording queries. Recent advancements in AI, especially large language models (LLMs), have given rise to tools like Perplexity.ai and You.com, which combine search results into easy-to-read summaries. However, these tools still have limitations they lack deep personalization, emotional understanding, field-specific tuning, and adaptability to a user’s evolving search journey. This study presents a next-generation AI-powered search engine that bridges these gaps. It combines Google’s Custom Search API for scalability with advanced natural language processing for contextual understanding and intelligent recommendation systems. What sets this system apart is its ability to build a growing map of a user’s knowledge over time. It dynamically adapts to multi-step queries and continuously refines results to match the user’s needs and learning path. Our approach aims to connect the precision of keyword-based searches with the flexibility of conversational, chat-style searches. The result is more relevant answers, reduced search fatigue, and a smoother, more personalized experience especially valuable for academic research, technical exploration, and other knowledge-intensive tasks.

DOI: https://doi.org/10.5281/zenodo.20044340

Published by: vikaspatanker

Authors: Khushbu Rajput, Bhavesh Jain

Abstract: Agriculture is an important sector in terms of food security and economic development, especially in developing nations. Precise crop yield estimation is required for efficient agricultural planning and management in the context of the increasing effects of climate change. Crop yield is affected by various factors, including climate variability, soil type, and availability of nutrients. Conventional crop yield estimation techniques, which rely on average values and traditional knowledge, are not reliable due to the complexities involved in crop yield estimation. Proposed in this paper is a framework for crop yield prediction using machine learning, incorporating climatic and soil variables. The climatic variables of rainfall, temperature, and humidity, and soil variables of soil pH and necessary nutrients (nitrogen, phosphorus, and potassium) are used as input variables. Three supervised machine learning algorithms—Linear Regression, Random Forest, and Gradient Boosting—are applied and compared to assess their predictive capability. Linear Regression is applied as a baseline algorithm, while ensemble methods are applied to deal with non-linearities in agricultural data. The performance of the models is measured using typical regression evaluation criteria, including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared (R²). The experimental outcomes show that the models based on ensemble methods perform better than the baseline model in terms of prediction accuracy and generalization ability. The results confirm that the combination of climatic and soil properties helps to improve crop yield prediction.

DOI: https://doi.org/10.5281/zenodo.20044195

Published by: vikaspatanker

Authors: Bantu Tejasr, K. Shree Vatsal, Dharavath Mahesh, Assistant Professor Dr. Sukanth T.

Abstract: This paper presents the design and practical implementation of a solar-driven, self-navigating vacuum robot intended for use in indoor settings. The system harnesses photovoltaic energy to eliminate grid dependency, uses a multi-sensor arrangement for real-time obstacle detection, and incorporates an Arduino Mega 2560 microcontroller for centralized decision-making. The prototype was subjected to rigorous testing across multiple indoor scenarios, where it recorded a 97% obstacle detection accuracy, approximately 94% cleaning coverage, and a continuous runtime of 60–70 minutes following a 3–4 hour solar charge. The outcomes confirm that merging renewable energy with embedded robotics yields a cost-effective and sustainable alternative to conventional cleaning appliances.

DOI: http://doi.org/

Published by: vikaspatanker

Authors: Research Scholar Gurpreet Singh

Abstract: In a MANET number of devices or mobiles are connected to each other with wireless medium. This network is a temporary network. In the MANET, there is not any centralized device which control all netwrok. MANET using dynamic topology. The Sybil attack is characterized as a malicious node misguidedly taking on various MANET . A Malicious device acts as though it's anything but a bigger number of nodes, for instance by mimicking different devices or basically by asserting bogus MANET. In this,a terrible device present more than one character in MANET. So it is not much safe Network. The attacker are easily attacks on the MANET. Consequently, Security is an essential worry to give ensured correspondence between nodes in impromptu organizations and shots at having the weaknesses are additionally more. In this paper we complete analysis and classification of the various Sybil attack techniques and decline the network performance and throughput.

DOI: https://zenodo.org/records/20042738

Published by: vikaspatanker

Authors: Jay Pradip Deshmukh, Siddhi Kadam, Prajakta Badhe, Professor Snehal Awatade, Professor Bhagyashri Ramchandra Gunjal

Abstract: The use of Generative AI has started changing how software is developed today. Many tools are now available that help developers write code faster or fix errors, but most of them only support specific tasks. They do not handle the complete software development process from In this paper, we propose a system called JellyfishAI ,which aims to automate the full software development lifecycle. The idea is simple — a user can describe their requirements in normal language, and the system will generate a complete, ready-to-use application. It combines different technologies like natural language processing, code generation models, testing, and deployment into one system. Technology is advancing very fast, and modern applications need to be scalable, reliable, and developed quickly. Because of this, traditional development methods are under a lot of pressure. Usually, software development includes many steps like requirement analysis, design, coding, testing, and deployment. These steps often require manual work and skilled developers, which makes the process slow and costly. It can also lead to mistakes and inconsistent results. With the introduction of Generative AI, things have started to improve. These systems can understand user input and generate code, documentation, and even design ideas. They help developers save time by assisting in coding and debugging tasks. However, these tools still have limitations. Most of them only focus on one part of development, like code suggestions or bug fixing, and do not provide a complete solution. Another issue is that developers still need to use multiple tools separately and connect them manually. This makes the workflow complicated and less efficient. Also, the code generated by AI tools often needs to be checked by humans to ensure it is correct, secure, and follows proper standards. Even though CI/CD pipelines help automate deployment, they mostly follow fixed rules and do not have intelligent decision-making capabilities. Similarly, testing tools can find errors but cannot automatically improve the quality of the code. Because of this, there is a need for a system that can intelligently manage all stages of development together. Today’s software systems are also becoming more complex, and there is a growing need to build applications quickly. Startups and companies want to turn their ideas into working products as fast as possible. However, depending on skilled developers and manual work often slows things down. This becomes even more difficult when dealing with large systems that require proper coordination and integration. To solve these problems, we introduce JellyfishAI, which is designed to automate the entire development process. Unlike existing tools, this system brings everything together in one place. It can understand user requirements, generate code, test it, check for errors, and deploy the application automatically. The system is built using multiple layers that include language processing, AI-based code generation, validation, and deployment. This ensures that the generated code is not only functional but also tested and ready to use. The system also improves over time using feedback and learning mechanisms. JellyfishAI also focuses on important factors like security, scalability, and maintainability. It includes features to detect vulnerabilities, manage dependencies, and improve performance, so that the final application meets industry standards. This system can change how software development is done. By automating repetitive tasks, developers can focus more on designing and solving problems instead of doing routine work. It also makes it possible for people without strong technical skills to build applications. In conclusion, AI has the potential to improve software development by making it faster, cheaper, and more efficient. However, to achieve full automation, we need systems that combine all stages of development in one place. JellyfishAI is an attempt to do that by providing a complete, end-to-end solution for building software automatically.

DOI: https://zenodo.org/records/20042299

Published by: vikaspatanker

Authors: Aaryan Kansal, Aditya Garg, Rajat Takkar

 

 

Abstract: Modern software systems must handle unpredictable user loads while maintaining performance and reliability. This paper proposes an AI-based load testing framework using asynchronous load generation, machine learning models, anomaly detection, and digital twin simulation. The system predicts latency, detects anomalies, and estimates failure thresholds, enabling proactive performance optimization for high-traffic applications.

DOI:

 

 

Published by: vikaspatanker

Authors: Dr. Anand Singh Rajawat, Mayur Devare, Vaibhav Ingale, Mohit Deshmukh, Prathmesh Ingle

Abstract: Traditional attendance marking is a labor-intensive process prone to human error and proxy attendance. This paper presents an automated Smart Attendance Tracking System (SATS) that utilizes Quick Response (QR) Code technology for instant identification. The system works by generating unique encrypted QR codes for each student, which are then scanned using a standard high-definition webcam. Built on a Java-based technical stack with a MySQL backend, the system processes image data to decode information in real-time. Experimental results demonstrate that the system can process an individual scan in less than 0.8 seconds with 100% accuracy in standard lighting. By eliminating physical registers and dedicated scanners, this system offers a cost-effective and highly scalable solution for educational institutions.

DOI: https://doi.org/10.5281/zenodo.20038072

Published by: vikaspatanker

Authors: Sufiyan Ansari, Arhaan Shaikh, Usaid Khairdi, Moaiz Kazi

Abstract: Text classification has numerous applications in real world scenarios. Emotion detection from text is one of the vital tasks within natural language processing, which has gained significant attention of researchers over the years. In this study, emotions are detected and classified for better human–computer interaction, sentiment analysis, health management, and smart chattingbots. A number of deep learning models are developed and outperform the traditional models for the classification. Con- volutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and hybrid of CNN and LSTM are used for emotion classification. Furthermore, two traditional machine learning ap- proaches including Logistic Regression and Naive Bayes are also implemented for comparison purpose. Preprocessing is a very important step for a good model. Text normalization, stop words removal, tokenization, and padding are used for data preparation. Word embeddings, specifically pre-trained word2vec, are used to capture the semantic relationship of text features learned from deep learning models. The performance evaluation of these models is done using accuracy, precision, recall, F1-score, and confusion matrix. The experimental results show that the deep learning models have outperformed the traditional models. The hybrid CNN-LSTM model achieved the best results to classify emotions in multi-class problem.

Published by: vikaspatanker