Computer Science, Computer Networks and Communications, Computer Engineering, Multidisciplinary
24
Scopus Publications
Scopus Publications
Node Localization in Underwater Wireless Sensor Networks using Machine Learning Manni Kumar, Sachin Ahuja, Nitin Goyal Proceedings of the 4th IEEE International Conference on Interdisciplinary Approaches in Technology and Management for Social Innovation Iatmsi 2026, 2026 The UWSNs are a type of submersible wireless sensor networks that can support important marine functions like environmental monitoring, inspection of underwater infrastructure and disaster pre-emption. Nonetheless, proper localization of nodes is still a problem because of acute acoustic noise, multipath propagation, movement that is brought about by the currents of a water body, and the impossibility of GPS application in the sub-aquatic environment. In this paper, a hybrid localization model (Hybrid ML+EKF) will be proposed to combine a deep regression learning block to learn nonlinear acoustic fingerprints (ToA/RSSI patterns) and an Extended Kalman Filter (EKF) smooth noisy measurements and reduce node drift. The suggested solution is compared to the traditional algorithms (ToA-LS, TDoA, DV-Hop) and isolated learning models (SVM-Loc, DNN-Loc). The results of the experiments indicate sustained improvements of accuracy and high resilience with the noise of the channel. In normalized timing noise 2.0, the hybrid model proposed has a RMSE of 2.30 m which is better than the ToA-LS (4.90 m) and DV-Hop (5.90 m). The CDF analysis also indicates smaller tail errors with increased reliability. In general, hybrid architecture has a viable tradeoff about localization accuracy, robustness when moving, and energy efficiency in long term underwater deployments.
Enhancing Earthquake Prediction with Ensemble Learning: A Hybrid XGBoost and LightGBM Approach Manni Kumar, Nivedita Sharma, Mahatava Saxena, Samaira Handa, Tamanna Sarawag, Tamanna Kashyap Proceedings of the 4th IEEE International Conference on Interdisciplinary Approaches in Technology and Management for Social Innovation Iatmsi 2026, 2026 Anticipated prediction of earthquakes is also a critical component of disaster management that requires accurate and timely forecasts to minimize the possible damages. The study adopts a two-machine learning system that comprises of XGBoost and LightGBM to attain better accuracy in the process of predicting earthquakes. The data of historical seismic activity is merged with seismic parameters such as magnitude and depth and location-based information which was acquired at Kaggle. The combination of XGBoost and LightGBM functions in an ensemble model allows more sophisticated seismic patterns to be detected that eventually adds value to the prediction quality. When ensemble learning approaches are applied, the results of improved generalization and reduced overfitting issues are achieved, which in turn allows the application in the real world. The mixture of XGBoost and LightGBM works in the ensemble model permits to rent finer seismic patterns and is more enhanced than individual models in prediction. Ensemble learning through the assistance of a special preprocessing pipeline of unequal seismic data and strong hyperparameter optimization that improve generalization and reduce overfitting. The quick detection of earthquakes that will be evident through the assistance of this strategy will assist the authorities in charge of disaster management in undertaking pre-emptive preventive measures in order to minimize the number of victims and safeguard infrastructure. As has been shown, the proposed ensemble outperforms the classic machine learning baselines, and it means that the methods of AI- based ensemble are applicable in the real world when it comes to earthquake prediction.
Leveraging Data Science for Small Business Growth: Applications of ML, IOT, and Data Visualization Nitish Lambra, Nivedita Sharma, Manni Kumar 2025 4th Opju International Technology Conference on Smart Computing for Innovation and Advancement in Industry 5 0 Otcon 2025, 2025 Small enterprises are found to form the backbone of the world's economy and, consequently, prove a source of innovation, employment, and community development. However, these firms face strong constraints that go against scalability and competition with big firms through resource poverty, low-profit margins, and unstable market conditions. This paper examines the revolutionary power of data science technologies to empower small businesses to transcend their limitations and grow sustainably. This case study of the representative case of grocery stores will illustrate how machine learning, the Internet of Things, and data visualization tools can be applied practically in solving key operational challenges. Smart shelves and temperature sensors are just a few of the many IoT-enabled devices that monitor stocks in real time, reducing waste spoilage, and enhancing greater compliance with health standards. Data visualization tools, such as Tableau, enable the business owner to understand complex datasets through user-friendly dashboards, providing actionable insights into sales trends, customer demographics, and operational inefficiencies. The integration of these technologies enhances not only the operational efficiency but also the decision-making and customer satisfaction levels. This research highlights the potential capability of data science to level the playing field for small business organizations, making them equipped to adapt rapidly to their market conditions and compete strongly with established organizations. Thereby, by embracing the concept of data science, small business organizations can open up potential opportunities, increase resilience, and place themselves at the center of competitive forces in this modern economy.
Federated Ensembled Learning for Intrusion Detection Using IoT Network Sujan Samanta, Rizwan Yousuf, Nivedita Sharma, Manni Kumar International Conference on Emerging Trends in Engineering and Technology Icetet, 2025 This paper focuses on exploring the integration of federated learning with ensemble methodologies for the improvement of intrusion detection systems in distributed IoT systems, solved data privacy concerns, scalability, and detection efficacy differences. The proposed framework allows the IoT devices to collectively build a collective intrusion detection model with no need to share local data, thus enhancing privacy and minimizing the utilization of bandwidth. In additional to detection performance problem, the ensemble component compounding predictions from multiple models making the system more accurate and robust against intelligent attacks. Consequently, profound assessment of the model show high level of effectiveness with accuracy of 0.92 in the detection of intrusion events. Due to privacy preservation, scalability, and robustness, this work suggests a pragmatic framework for enhancing IoT networks' cybersecurity without compromising data sensitivity and with decentralized data sources. The proposed framework therefore offers a secure efficient intrusion detection system that can adequately meet the rising cybersecurity demand in more interconnected IoT systems.
Fuzzy Decision-Based Clustering for Efficient Data Aggregation in Mobile UWSNs Aadil Mushtaq Pandith, Manni Kumar, Naveen Kumar, Nitin Goyal, Sachin Ahuja, Yonis Gulzar, Rashi Rastogi, Rupesh Gupta Computers Materials and Continua, 2025 Underwater wireless sensor networks (UWSNs) rely on data aggregation to streamline routing operations by merging information at intermediate nodes before transmitting it to the sink. However, many existing data aggregation te... | Find, read and cite all the research you need on Tech Science Press
Analysis and performance evaluation of computation models for node localization in deep sea using UWSN Manni Kumar, Nitin Goyal, Ashutosh Kumar Singh, Rakesh Kumar, Arun Kumar Rana International Journal of Communication Systems, 2024 SummaryUnderwater wireless sensor network (UWSN) connects the real world to the hidden resources available deep into the sea using sensor nodes deployed sparsely and are interconnected to gather information for any movement/change in the ocean. The computed information on nodes obtained from concurrently running UWSN's applications will be meaningful, only if the location of this change will be identified. Nevertheless, it is always difficult to obtain the exact location coordinates of any misadventure like tectonic plate's movement but using localization algorithms in UWSN helps to obtain the coordinates. Since localization algorithms for terrestrial networks are not feasible for UWSN because of environmental challenges. Moreover, GPS systems do not work after 15 m of depth and the radio frequency gets suppressed. So, the requirement arises to analyze existing localization approaches in the deep sea where acoustic signals are used for communication and data transfer. In this paper, along with describing the UWSN's applications and challenges, the underwater localization schemes are reviewed to present, summarize, and mention the scope of improvement. Further, classification into range‐based and range‐free categories of these schemes is depicted with implementation in the NS2.30 simulation environment of some of the recent techniques to showcase the reasons for better performance.
Enhancing Human-Computer Interaction with Generative AI Ronit, Parth Agrawal, Manni Kumar 2024 IEEE 4th International Conference on ICT in Business Industry and Government Ictbig 2024, 2024 Modern Chatbots represent a significant advancement in human-computer interaction by integrating emotional intelligence into their design. Unlike traditional chatbots that rely on predefined rules and keyword recognition, these advanced systems use affective computing and sentiment analysis to understand and respond to user’s emotional states. By using large LLMs, this innovation brings several key benefits. It enhances efficiency by providing immediate, relevant answers without users needing to search through multiple sources. The chatbot delivers precise information tailored to the emotional context of the interaction. Additionally, these chatbots offer greater convenience by consolidating information into a single query, streamlining the user experience and reducing the need to navigate through various platforms. They have improved so much yet are so much limited. An emotion sensitive chatbots should recognize and respond to emotional cues, fostering a deeper connection with users, encouraging regular interaction, creating a more interactive and personalized experience while also offering valuable insights into user behavior and preferences, enabling organizations to refine their marketing strategies and product development efforts. Overall, emotion-based chatbots should bridge the gap for human digital interactions. In this view, this paper explores how Generative AI, its technology, various applications, how it can be improved via enhancing the Human-Computer Interface, how does the improved version fare against current version, its various new applications, and additional key challenges and research prospectives.
Brain Tumor Detection Using AIML Shristi Kumari, Pooja Bharti, Gaurav Dixit, Vishwas Rohilla, Manni Kumar, Chahil Choudhary 2024 MIT Art Design and Technology School of Computing International Conference Mitadtsocicon 2024, 2024
