Evaluating research trends in urban agricultural indicators through bibliometrics Pravin Dange, Naval Lawande Discover Sustainability, 2025 The role of Urban Agriculture (UA) in achieving food security and sustainability couldn't be overemphasized. Further, indicators are critical to the measurement of the role of Urban Agriculture in enhancing food security and sustainability and to understand the trade-offs between the two. The purpose of the paper is to identify, analyze, and synthesize research on Urban Agriculture Indicators (UAI) to identify gaps in information and scout for future directions and scope for research. This paper presents a review of the body of literature on Urban Agriculture Indicators using bibliometric analysis; the search was made on the Web of Science and Scopus databases up to May 2025. After removal of duplicates and cleaning the dataset 504 records were subjected to bibliometric analysis using the Biblioshiny tool to address the research questions. Our findings reflect a significant growth, underlining the increasing importance of UAI in addressing issues of urban resilience and sustainability. We observed that the UAI research is incidental to research in Agriculture, Urbanisation and Sustainability related areas and is yet to mature as a separate discipline. Multifunctional and diverse nature of UA, lack of methodological robustness and ad hoc borrowing of indicators from other domains make indicators less relevant in assessment of UA. There is a thematic shift from a siloed disciplinary focus to a systems approach in UAI development. The data used was sourced form WoS and Scopus database only. The paper points out the lopsided nature of indicator development for UA, identifies challenges associated with UAI development and suggests measures to overcome the limitations to make it more relevant. We also suggest areas for future research to improve their adaptability.
Giving the hostel mess a miss: investigating students’ attitude and behaviour towards mess and skipping mess meals, using the theory of planned behaviour Pravin Dange, Chanakya Kumar International Journal of Adolescence and Youth, 2025 The ‘hostel mess food’ provides for both economy and health. Yet, a significant number of students miss their scheduled meals at the mess to eat outside food. This is a significant concern for mess managers, university authorities, and parents. Understanding the antecedents of the mess meal skipping behaviour is crucial to design interventions to promote healthier eating habits and optimize mess utilization. This study investigates these antecedents using the theory of planned behaviour and employs the Structural Equation Modelling (SEM) technique to analyse the data collected from Hostel Students of a prominent Pune (India) based Higher Education Institution. The study reveals that the attitude towards mess food negatively influences mess meal skipping behaviour. Subjective norms significantly impact mess food skipping behaviour and moderate the influence of Perceived Behavioural Control on meal skipping behaviour. For mess managers, it entails focusing on subjective norms-based interventions to influence the mess meals skipping behaviour.
The Ramayana & Bhagvad Gita as training tools for effective & ethical leadership Pradnya Vishwas Chitrao, Pravin Kumar Bhoyar, Pravin Dange, Rajiv Divekar Cogent Arts and Humanities, 2025 The 21st C has speedy technological inventions. Success or failure is decided by the competency of managers based on excellent leadership. Excellent leadership is not just efficient control of strategy and tactics, but includes morals, feeling for persons, resisting temptations, and just, judicious behaviour with opponents. Ramayana and Bhagvad Gita postulate best leadership. The ideas of yoga of action and duty in Bhagvad Gita guide professionals. Shree Ram exemplifies efficient, ethical leadership. Different characters in Ramayana and incidents like protecting rishis doing rituals, constructing an overpass across seawater for reaching Suvarna Lanka, accepting Vibhishana as a friend, Shree Ram’s refusal to rule Sri Lanka after Ravana’s demise are classic management situations that enable student managers hone their management-based reasoning. The study analyses the central characters and incidents in Ramayana and in Bhagvad Gita and studies critical managerial leadership tenets for applying in current business scenarios. It scrutinizes the moral standards in these two major epics and gleans learnings for contemporary leaders. The study will guide modern leaders to lead ethically and competently handle critical situations. The study shows a need for including sessions on these two epics for training MBA student managers and corporate executives for leadership and ethical governance.
Visual Perception-Based Data-Driven Path Tracking Using Behavior Analysis and the ANFIS Approach Pooja Bhatt, Vinita Tiwari, T Vinoth, Pravin Dange Proceedings of 3rd IEEE International Conference on Knowledge Engineering and Communication Systems Ickecs 2025, 2025 This study presents an advanced path-tracking framework for autonomous vehicles, integrating visual perception, behavior analysis, and the Adaptive Neuro-Fuzzy Inference System (ANFIS) to enhance trajectory control in dynamic environments. By leveraging real-time data from visual sensors, the system captures critical environmental information, while behavior analysis evaluates vehicle dynamics and external influences. ANFIS synergizes neural networks and fuzzy logic to optimize trajectory predictions, ensuring adaptability and precision under varying conditions. The proposed data-driven approach incorporates feature extraction, dimensionality reduction, and hybrid learning techniques to improve computational efficiency. Experimental validation across urban and highway scenarios demonstrates a substantial performance gain over conventional methods, achieving a 96.5% tracking accuracy and a response time of 0.35 seconds. The results underscore the system’s ability to seamlessly adjust to environmental changes, making it a promising solution for real-world autonomous navigation. Future research will explore deep learning integration to further optimize performance and expand applicability.
Machine Learning and Graph Theory-Based Approach for Detecting and Diagnosing Hardware Trojans in Gate-Level Netlists Jay Gandhi, V. Prabakaran, Mahalakshmi S, Pravin Dange Proceedings of 3rd IEEE International Conference on Knowledge Engineering and Communication Systems Ickecs 2025, 2025 Hardware Trojans (HTs) represent a significant threat to the integrity of hardware systems, often inserted during the design or fabrication stages. These malicious modifications can compromise system functionality, performance, and security. Detecting and diagnosing hardware Trojans in gate-level netlists is crucial for ensuring the reliability of hardware devices. This paper introduces an innovative approach that combines machine learning (ML) techniques and graph theory to enhance the detection and diagnosis of HTs. The hardware design is modeled as a graph, where gates are represented as nodes and their interconnections as edges, capturing the structural relationships of the system. Features such as node degree, betweenness centrality, and clustering coefficient are extracted from this graph-based representation. These features are then used as input to machine learning classifiers, including Random Forest and Support Vector Machine (SVM), to classify gates as Trojan or nonTrojan. The proposed method is evaluated on benchmark hardware designs, demonstrating its effectiveness in detecting HTs with high accuracy, sensitivity, and specificity. The results indicate that the integration of ML and graph theory significantly improves Trojan detection performance compared to traditional techniques. Additionally, the approach enables the precise diagnosis of Trojans, helping to locate their specific position and behavior within the hardware design. This work provides a robust, scalable solution for enhancing hardware security.
Design and Evaluation of VR-Enhanced Hand Therapy with a Custom Biomechatronic 3D-Printed Orthosis Chintan Thacker, Kishore Ravikumar, K. Bharat Ramkumar, Pravin Dange Proceedings of 3rd IEEE International Conference on Knowledge Engineering and Communication Systems Ickecs 2025, 2025 Hand rehabilitation plays a crucial role in the recovery of patients with hand injuries or neurological disorders. Traditional methods, while effective, often fail to engage patients and can be time-consuming. This paper introduces a novel approach to hand therapy that integrates Virtual Reality (VR) with a custom-made biomechatronic orthosis, 3D-printed to match the individual’s hand anatomy. The VR system offers an immersive environment, where patients perform hand exercises that improve strength, flexibility, and dexterity, while the biomechatronic orthosis provides physical feedback and support. The orthosis assists the patient by adjusting the resistance based on real-time sensor data, ensuring that the exercises are performed correctly and efficiently. The system’s ability to offer personalized rehabilitation, coupled with the interactive nature of VR, aims to improve patient motivation and enhance therapy outcomes. An experimental evaluation of this system was conducted, comparing it with conventional therapy methods. The results showed a significant improvement in hand function and range of motion (ROM), along with higher levels of patient motivation. This approach represents an innovative solution that combines advanced technology with patient-centered care, providing a more effective, engaging, and personalized rehabilitation experience. The study suggests that VR-enhanced hand therapy using a custom biomechatronic orthosis has the potential to revolutionize the rehabilitation process.
Decentralized Federated Learning Framework with Blockchain for Verifiable and Trusted Collaboration Kruti Sutaria, Mahalakshmi S, Manisha Paliwal, Pravin Dange Proceedings of 3rd IEEE International Conference on Knowledge Engineering and Communication Systems Ickecs 2025, 2025 Federated Learning (FL) is an innovative distributed machine learning approach that allows multiple participants to collaboratively train a global model without sharing their private data. While effective, traditional FL frameworks often rely on centralized servers for model aggregation, making them susceptible to single points of failure, security breaches, and trust concerns. To address these challenges, this paper introduces a Decentralized Federated Learning Framework (DFLF) integrated with blockchain technology. The proposed framework eliminates the need for a central coordinator by utilizing blockchain’s immutable ledger and decentralized architecture. This integration ensures secure model sharing, verifiable updates, and incentivized collaboration among participants. Smart contracts automate key operations, including model aggregation, contribution validation, and reward distribution, thereby enhancing efficiency and transparency. The framework was implemented using a permissioned blockchain and tested on a healthcare dataset for patient diagnosis prediction. Experimental results demonstrate improved model accuracy, reduced latency, and enhanced trust compared to conventional FL approaches. The inclusion of blockchain ensures robust privacy preservation and builds trust in collaborative machine learning. This paper provides a comprehensive analysis of the framework, including its design, algorithms, mathematical formulations, and performance metrics. The results highlight the potential of combining blockchain with FL to establish a scalable, secure, and trustworthy ecosystem for decentralized machine learning applications.
Scalable Subspace Clustering via Purity Kernel Tensor Learning Rachit Adhvaryu, S Rajes Kannan, Sanjay C. P, Pravin Dange 3rd International Conference on Integrated Circuits and Communication Systems Icicacs 2025, 2025 Subspace clustering is a fundamental task in machine learning, particularly for high-dimensional datasets where the goal is to group data points that lie in distinct subspaces. Traditional methods, such as Sparse Subspace Clustering (SSC) and Low-Rank Representation (LRR), have demonstrated good performance but often suffer from high computational costs, particularly when applied to large-scale data. To address these limitations, this paper introduces a novel technique called Purity Kernel Tensor Learning (PKTL). PKTL combines the power of tensor decomposition with a purity kernel function to improve both the efficiency and effectiveness of subspace clustering. By utilizing tensor-based representations, the method captures higher-order relationships in the data, allowing for a more robust clustering process. The purity kernel function measures the coherence within the clusters, enhancing the identification of subspaces in complex datasets. The proposed approach not only reduces computational complexity but also improves the clustering accuracy compared to existing methods. Experiments on synthetic and real-world datasets, including ORL, Yale, and MNIST, show that PKTL outperforms conventional subspace clustering algorithms, both in terms of clustering accuracy and scalability. The results demonstrate the potential of PKTL in handling large datasets while maintaining high performance, making it a promising approach for scalable subspace clustering in real-world applications.
Efficient and Robust Graph Metric Learning Guided by Model Selection Principles Umang Soni, Rajeswari. C, Ganesh Pandit Pathak, Pravin Dange 3rd IEEE International Conference on Data Science and Network Security Icdsns 2025, 2025 Graph Metric Learning (GML) is increasingly critical for tasks such as classification, clustering, and anomaly detection on graph-structured data. However, selecting an optimal model that balances accuracy, and computational efficiency remains a core challenge. While existing works have integrated information-theoretic criteria such as Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) with graph neural networks (e.g., Liu et al.), this study advances the state-of-the-art by presenting a unified, model-agnostic framework that systematically combines AIC/BIC with diverse GML techniques. Our proposed approach, Graph Metric Learning with Model Selection (GMLMS), begins by generating a candidate pool comprising traditional metric learning methods (e.g., LMNN, ITML), graph neural networks, and kernel-based models. Each candidate is evaluated based on its AIC and BIC scores to identify the model that best balances fit and complexity. The selected model is then fine-tuned to enhance performance. Experimental evaluation on real-world datasets, including citation networks, social graphs, and protein-protein interaction networks, demonstrates that GML-MS significantly outperforms individual baseline methods. The selected model achieved the highest accuracy (94.2%) with improved precision, recall, and F1-score, while also maintaining the lowest AIC and BIC values. This work provides a robust, scalable, and interpretable solution for graph learning applications, especially where large-scale, complex data necessitates careful model selection for optimal performance.
Twitter Bot Detection Using GCN with Adaptive Neighborhood Aggregation and Squeeze Module Mitul Patel, D Karthick Rajan, Adusupalle Muni Raju, Pravin Dange Proceedings of 3rd IEEE International Conference on Knowledge Engineering and Communication Systems Ickecs 2025, 2025 The rise of automated bots on social media platforms, especially Twitter, has posed a significant challenge in maintaining the authenticity and integrity of online content. Detecting these bots is crucial for curbing misinformation, improving user engagement, and ensuring the credibility of social media interactions. In this paper, we propose a novel bot detection framework that utilizes Graph Convolutional Networks (GCN) with Adaptive Neighborhood Aggregation and a Squeeze Module. The proposed framework models the social network of Twitter users as a graph, where each user is represented as a node and interactions between users form the edges. The GCN is employed to capture the complex relationships between users based on their interaction patterns, while the Adaptive Neighborhood Aggregation mechanism dynamically adjusts the importance of each user’s neighbors. This adjustment is critical as it enables the model to better understand diverse user behaviors, distinguishing between genuine users and bots. The Squeeze Module is introduced to enhance feature extraction, reducing redundancies and ensuring more efficient learning. Through extensive experimentation on publicly available datasets, the model’s performance is compared with existing methods, demonstrating a significant improvement in accuracy, precision, recall, and F1-score. The results confirm that the integration of GCN with Adaptive Neighborhood Aggregation and the Squeeze Module offers an effective and scalable solution for Twitter bot detection.
