Authors - Lakshmi Priya G G, Padma Lakshmi G, Thomas Felix K Abstract - The conservation of natural habitats and the coexistence of humans and wildlife are vital for biodiversity preservation. In the Mudumalai Forest Conservation region of Tamil Nadu, India, achieving harmony between human activities and wildlife preservation presents a significant challenge. A framework for real-time monitoring of human-wildlife interactions leveraging a comprehensive integration of satellite imagery, Internet of Things (IoT) sensors, and deep learning techniques is discussed in this paper. The proposed system utilizes high-resolution satellite imagery to identify the hotspots, where human-wildlife interactions are most likely to occur or where conflicts are already prevalent. Deep learning algorithms are applied to analyze the satellite imagery data and detect patterns indicative of potential human-wildlife conflict areas. By training and continuously updating the models with real-time information, the system can accurately identify areas of heightened risk. Throughout the identified hotspots, IoT sensor networks are deployed strategically to monitor the real time human activities, wildlife movements, and predict the possibilities of human - wildlife interactions by employing light weight deep learning models. Based on the prediction, real-time alerts and early warnings with location are communicated via message and mobile apps notification to the relevant stakeholders for necessary actions. Moreover, by incorporating feedback loops, the system can adapt and improve its performance over time.
Authors - N. Mangaiyarkarasi, J. Arputha Vijaya Selvi, T. Pasupathi Abstract - Free Space Optical (FSO) communication systems offersultra high bandwidth, large data rate and very secure data transmission, making them a feasible solution for next generation communication networks. However, performance of the FSO communication system is greatly impacted by adverse atmospheric conditions such as heavy turbulence, rain, and fog, which set up errors and degrade the quality of the signal. In this paper an adaptive neural network-based approach is presented to mitigate errors under different seasons of atmospheric conditions. This method exploits a convolutional neural network (CNN) based architecture to predict and compensate the atmospheric induced distortions, thereby improving the performance of the FSO communication system. It significantly improves the Bit Error Rate (BER) and Signal-to-Noise Ratio (SNR). Real-time atmospheric data such as temperature (T in C), relative humidity (%), atmospheric pressure (Pa), and wind speed (ms-1) are collected and the CNN dynamically changes the parameters to optimize performance. Achieved result shows that the neural network model significantly improves the robustness and reliability of the communication system. This method triggers the way for more resilient FSO networks, which is more crucial for the implementation of 5G/6G and beyond communication infrastructures.
Authors - Shivani Kania, Yesha Mehta Abstract - Augmented analytics, managed by machine learning and natural language processing, handles data analysis findings, reducing the time-consuming pre-processing and feature development processes. The article focuses on the importance of Augmented Data Science (ADS), an interactive, data-driven system that combines personal judgement with analysis of statistics to improve decision-making in data interpretation. The challenges are developing the requirements for assessment, developing defined review methods, and comparing suggested methodologies to real-world datasets and use cases. The goal is to create and develop a model for data interpretation and natural language-based generated output in Augmented Analytics, with objectives including data processing, model design, query processing, and component analysis.
Authors - Saroj S. Date, Sachin N. Deshmukh, Mahesh B. Shelke, Daivat D. Sawant, Chatrabhuj B. Kadam, Kailas M.Ambhure Abstract - Analyzing text data in regional languages is essential for uncovering sociocultural insights. However, languages like Marathi face considerable challenges due to the scarcity of computational resources. Over the past few decades, Linguistic Inquiry and Word Count (LIWC) software has become a gold standard for text data analysis. It uses a dictionary-based approach to classify words into predefined psychological and linguistic categories, enabling researchers to explore aspects of personality, behavior, emotions, and social interactions. This research paper introduces the application of MR-LIWC2015 for analyzing Marathi text data. MR-LIWC2015 is a translation of the English LIWC dictionary into the Marathi language. In this paper, by leveraging a dictionary-based approach, MR-LIWC2015 analyzes psychological and emotional dimensions in textual feedback. Using a dataset of 185 student feedback entries, translated from English to Marathi, this research evaluates Marathi LIWC's efficacy by comparing its results with the original English LIWC. Findings indicate a high positive correlation between the two software, demonstrating the reliability of the Marathi LIWC. This advancement not only facilitates feedback analysis in Marathi but also opens avenues for diverse domains like sentiment analysis, mental health analysis, product review analysis, expressive writing analysis, etc. Future work aims to expand the lexicon set of Marathi LIWC and explore cross-domain applications. Furthermore, the software may be adapted for other regional Indian languages, marking MR-LIWC2015 as the first LIWC translation for any Indian language.
Authors - Hardik I. Patel, Dharmendra Patel Abstract - Higher education institutions' reputation, financing, and student achievement are all impacted by student retention, which has grown to be a major problem. In order to properly identify at-risk students, traditional methods to retention issues frequently lack the predictive capacity and flexibility required. In order to predict student retention rates, this study makes use of machine learning approaches, giving academic leaders useful information. The suggested approach builds a strong prediction model by combining a variety of information, such as financial, behavioral, academic, and demographic factors. The model finds important patterns and trends related to retention outcomes by using sophisticated techniques like gradient boosting and neural networks. A methodical procedure that includes feature selection, data preparation, and model assessment guarantees excellent accuracy and scalability. By employing Explainable AI technologies to make forecasts clear and actionable, the study also highlights the significance of interpretability. This method allows institutions to apply timely interventions, such academic help, counseling, or financial aid changes, by turning raw data into useful forecasts. The results show how predictive analytics has the power to transform retention tactics and promote an inclusive and effective educational system. This study offers a guide for incorporating machine learning into higher education's strategic decision-making process.
Authors - Dhanashree Joshi, Nilesh B. Korde, Pratibha Jape, Gayatri Newase, Mitali Gajbhiye, Maitriyee Kadam, Gayatri Gujar Abstract - Agriculture is a vital sector in many countries, especially in developing economies like India, where it contributes significantly to GDP and supports over half the population. The sector includes a wide range of crops, such as sugarcane, banana, apple, and pomegranate, which are crucial for food security and economic stability. However, the cultivation of these crops presents challenges, particularly in crop management and pesticide application. For instance, sugarcane fields are dense and tall, making it difficult for farmers to access all areas, and banana plantations in states like Tamil Nadu, Kerala, and Andhra Pradesh also have densely packed leaves that complicate manual spraying efforts. Similarly, apple orchards in hilly regions such as Himachal Pradesh and Uttarakhand, and pomegranate fields in Maharashtra and Karnataka, are challenging due to their dense canopies and rugged terrain. These obstacles make it difficult for farmers to spray pesticides safely and efficiently, and the presence of wild animals hiding within these fields adds an additional layer of risk. To address these challenges, advanced technologies like drone-based pesticide spraying systems are proposed. These drones, equipped with GPS, IoT sensors, and deep learning capabilities, can autonomously navigate dense crop fields and varied terrains. By targeting specific farm coordinates and adjusting their spraying patterns based on real-time environmental data, the drones ensure precise and even pesticide application. This technology not only enhances the safety of farmers by minimizing their exposure to hazardous environments but also improves the efficiency and effectiveness of pest control. As a result, the adoption of drone technology can reduce labour costs, increase crop yields, and promote sustainable farming practices in regions where traditional methods are inadequate and can improve efficiency of farming and reduce human load. Additionally, by taking drone in consideration while spraying pesticides in farming can improve overall efficiency.
Authors - Shoib Ahmed Shourav, Shahariar Sarkar, Salekul Islam Abstract - For developing countries, maintaining road network infrastructure is an essential concern. To strengthen a nation’s economy, road infrastructure must be maintained effectively. Potholes, speed breakers, and drain holes in roads are major reasons for causing accidents, traffic jams, and car damage. In addition to improving driving safety and minimizing vehicle damage and accidents, identifying road anomalies like potholes, speed breakers, and drain holes is essential for enabling authorities to effectively manage road maintenance. Self-driving cars need to be able to handle different road conditions. In this research, a custom dataset comprising potholes, drain holes, and speed breakers was developed. The study employs YOLOv11, a cutting-edge deep learning-based object detection model, to accurately detect these anomalies, including a comparison of road anomaly detection performance under daytime and nighttime conditions. The proposed approach achieved an accuracy of 83.8% on the daytime dataset, 81.6% on the nighttime dataset, and 84.4% on the combined dataset.
Authors - Madhusmita Mishra, R. Kanagavalli Abstract - This comprehensive survey examines advancements in semi-supervised learning (SSL) techniques developed to address imbalanced multi-class classification problems across a variety of real-world applications, including healthcare, fraud detection, and industrial monitoring. Traditional machine learning models often struggle with highly skewed data distributions, leading to biased predictions that favour majority classes while overlooking minority classes. SSL, which leverages both labelled and unlabelled data, has emerged as a promising approach, reducing the need for extensive labelled datasets while improving model generalization for minority classes. This review focuses on methodologies such as re-sampling, cost-sensitive learning, ensemble learning, hybrid techniques, active learning, and evolutionary algorithms, each offering unique approaches to mitigate the impact of class imbalance. Re-sampling methods, such as SMOTE (Synthetic Minority Over-sampling Technique) and its variants, augment minority classes by creating synthetic samples, addressing imbalances within SSL frameworks. Costsensitive learning introduces penalties for misclassifications, improving sensitivity to minority classes, while ensemble learning methods, like bagging and boosting, combine multiple classifiers to enhance predictive accuracy in multi-class settings. Additionally, hybrid techniques that integrate re-sampling with cost-sensitive approaches show promise in balancing class representation and boosting model robustness. Active learning, which iteratively selects the most informative samples, and meta-learning, which enables models to adapt dynamically to different class distributions, provide further innovation in tackling imbalances in SSL applications.
Authors - Khush Mendiratta, Shweta Singh, Pratik Chattopadhyay Abstract - Early detection of brain tumors through magnetic resonance imaging (MRI) is essential for timely treatment, yet access to diagnostic facilities remains limited in remote areas. Gliomas, the most common primary brain tumors, arise from the carcinogenesis of glial cells in the brain and spinal cord, with glioblastoma patients having a median survival time of less than 14 months. MRI serves as a non-invasive and effective method for tumor detection, but manual segmentation of brain MRI scans has traditionally been a labour-intensive task for neuroradiologists. Recent advancements in computer-aided design (CAD), machine learning (ML), and deep learning (DL) offer promising solutions for automating this process. This study proposes an automated deep learning model for brain tumor detection and classification using MRI data. The model, incorporating spatial attention, achieved 96.90% accuracy, enhancing the aggregation of contextual information for better pattern recognition. Experimental results demonstrate that the proposed approach outperforms baseline models, highlighting its robustness and potential for advancing automated MRI-based brain tumor analysis.
Authors - Ninaad Nagaraj Yeligar, Rajesh Prakash Unakal, Soumya H Hooli, Prerana Girish Karoli, Kiran M R, Suneeta V Budihal Abstract - The work aims to address the need for efficient energy management by implementing a smart energy meter using LTE technology. We are developing a smart energy meter system capable of accurately measuring and transmitting energy usage data. This work features easy hardware implementation using a ESP32, a cost-effective microcontroller and a LTE module to establish a connection for transmitting data. The system utilizes LTE technology to ensure reliable and long-range communication. Energy consumption data is then monitored and the captured which could be used in future for different purposes. This work provides a comprehensive and user-friendly solution for continuous monitoring and management of energy consumption. The LTE system’s ability to capture real-time energy usage data enhances accessibility and facilitates data-driven decision-making. This innovative solution is ideal for applications where efficient energy management is crucial, such as in residential, commercial, and industrial settings.
