Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
13 results
Search Results
Item A Weighted Hybrid Centrality for Identifying Influential Individuals in Contact Networks(Institute of Electrical and Electronics Engineers Inc., 2022) Shetty, R.D.; Bhattacharjee, S.In the context of infectious disease spread, one of the most challenging tasks is to identify influential nodes in the human contact networks. In recent years, the research community has acquired strong evidences of complex and heterogeneous connections and diverse patterns in a variety of real human contact networks. The heterogeneity of network topologies has a deep influence on understanding the spreading dynamics of disease. In the process of infection spread, network edges are the critical communication channels. Many existing approaches for identifying prominent nodes in such networks rely on node attributes. Further, in unweighted networks, all the edges are considered to be equal, which imposes an unrealistic assumption for epidemic spreading through frequent interactions between humans. Here, we present an edge weighting technique, named as, Weighted Hybrid Centrality left( {{W_{{C_H}}}} right), that takes into account multiple centrality indicators, including degree, k-shell, and eigenvector centrality, as well as the frequency of interactions be-tween any two users (nodes) which is considered as potential edge weight. To analyze the performance of {W_{{C_H}}}, we have utilized the weighted susceptible-infected-recovered (W-SIR) simulator and have carried out a comparative experiment on six real-world heterogeneous networks. The proposed technique outperforms the baseline methods with 0.067 to 0.641 averaged Kendall's Ï„ score. This method is useful for modeling disease dynamics and identifying highly influential contacts by considering both node and edge properties. © 2022 IEEE.Item Pattern Analysis of COVID-19 Based On Geotagged Social Media Data with Sociodemographic Factors(Institute of Electrical and Electronics Engineers Inc., 2022) Sabareesha, S.S.S.; Bhattacharjee, S.; Shetty, R.D.The world has faced a catastrophic global crisis of COVID-19 caused by coronavirus and called for analyzing the affected areas in any country. The study helps to understand how the second wave affected different states in India concerning sociodemographic factors, such as population density, economy, and unemployment rate. During the lockdown, the sudden impact of staying at home has led to increased social media usage, where people expressed their opinions on multiple topics. Twitter provides timestamp and sometimes spatial information of the tweets generated. Using the geotagged Twitter dataset, a study in India is performed in this work considering the second wave of COVID-19, which occurred approximately from April to June 2021. It analyses the temporal and spatial patterns of the geotagged tweets generated from all the states during the period mentioned above. Also, topic modeling and sentiment analysis are performed to understand the concerns discussed by the people. We use different states' sociodemographic factors and machine learning algorithms to divide the population into high and low categories to understand the topic prevalence in different socioeconomic groups. This study reveals that the low socioeconomic groups have shared more concerns, urging the government to help fight the COVID-19 pandemic. © 2022 IEEE.Item Weighted GNN-based Betweenness Centrality Considering Stability and Connection Structure(Institute of Electrical and Electronics Engineers Inc., 2023) Shetty, R.D.; Bhattacharjee, S.Epidemic spreading or information spreading in a network leads to a broader range of propagation due to the presence of bridge nodes, as defined by the betweenness centrality. The existing literature shows that computing betweenness centrality by considering connection topology is computationally expensive and less efficient. Moreover, exploiting the edge features in the network can determine how strongly the nodes are connected. This work proposes a technique to leverage connection topology and tie strength for modeling Weighted Betweenness Graph Neural Network (WBGNN) for a faster-ranking estimation. This measure is based on shortest path calculations, which take connection strength into account for determining the fastest path for contagion or information spread in the network. A variant of statistical weighted betweenness centrality, namely Stable Betweenness Centrality (CSB) and the proposed GNN-based WBGNN are compared to analyze the effectiveness of the proposed model. Further, the WBGNN model is also compared with different machine learning regressors and a neural network-based model to examine the efficacy of the proposed model. The experimental outcome has revealed that the WBGNN model has achieved a 0.203 to 0.536 improvement in Kendall's Ï„ score, and also takes less time for node ranking estimation compared to CSB and other machine learning-based regressor models. © 2023 IEEE.Item A Temporal Metric-Based Efficient Approach to Predict Citation Counts of Scientists(Springer Science and Business Media Deutschland GmbH, 2023) Dewangan, S.K.; Bhattacharjee, S.; Shetty, R.D.Citation count is one of the essential factors in understanding and measuring the impact of a scientist or a publication. Estimating the future impact of scientists or publications is crucial as it assists in making decisions about potential awardees of research grants, appointing researchers for several scientific positions, etc. Many studies have been proposed to estimate publication’s future citation count; however, limited research has been conducted on forecasting the citation-based influence of the scientists. The authors of the scientific manuscripts are connected through common publications, which can be captured in dynamic network structures with multiple features in the nodes and the links. The topological structure is an essential factor to consider as it reveals important information about such dynamic networks, such as the rise and fall in the network properties like in-degree, etc., over time for nodes. In this work, we have developed an approach for predicting the citation count of scientists using topological information from dynamic citation networks and relevant contents of individual publications. This framework of the citation count prediction is formulated as the node classification task, which is accomplished by using seven machine learning-based classification models for various class categories. The highest average accuracy of 85.19% is achieved with the XGBoost classifier on the High Energy Physics - Theory citation network dataset. © 2023, IFIP International Federation for Information Processing.Item YOLOv5 Model-based Ship Detection in High Resolution SAR Images(Institute of Electrical and Electronics Engineers Inc., 2023) Sapna, S.; Sandhya, S.; Shetty, R.D.; Pais, S.M.; Bhattacharjee, S.Detection of ships in Synthetic Aperture Radar (SAR) images play a crucial role in maritime surveillance, most importantly under complex sea conditions. SAR permits observation in any weather conditions, at all hours of the day and night. At present, the ship detection from SAR images is a notable area of research since it is very difficult to detect the ships in the SAR images using traditional object or target detection algorithms. In this work, a You Only Look Once version 5 (YOLOv5) based ship detection model from SAR images with faster training speed and higher accuracy is implemented and tested. This model achieved a mean average precision (mAP) of 96.2% with a training time of 8.63 hours. This work also provides a comparative analysis with the existing methods for detection of ships in SAR images. The comparison shows that the YOLOv5 based model performs better in terms of both mean average precision and training time when compared to the existing models. © 2023 IEEE.Item Prediction of Drug Interactions Using Graph-Topological Features and GNN(Springer Science and Business Media Deutschland GmbH, 2023) Balamuralidhar, N.; Surendran, P.; Singh, G.; Bhattacharjee, S.; Shetty, R.D.The risk of side effects is sometimes inevitable every time two or more drugs are prescribed together, and these side effects of varying adversity levels can be referred to as drug-drug interactions (DDI). Massive amounts of data and the constraints of experimental circumstances result in clinical trials for medication compatibility being time-consuming, risky, expensive, and impractical. Recent research has demonstrated that DDI can be modelled as graphs and experimentally shown that deep learning on graphs can be a practical choice for determining the correlation and side effects of taking multiple medications simultaneously. We propose a novel approach to use inductive graph learning with GraphSAGE, along with topological features, to leverage the structural information of a graph along with the node attributes. An experimental study of the approach is done on a publicly available subset of the DrugBank dataset. We achieve our best results that are comparable with state-of-the-art works using degree, closeness and PageRank centrality measures as additional features with less computational complexity. This study can provide a reliable and cost-effective alternative to clinical trials to predict dangerous side effects, ensuring the safety of patients. © 2023, IFIP International Federation for Information Processing.Item Analyzing Derived Network Feature Importance to Identify Location Influence in LBSN(Institute of Electrical and Electronics Engineers Inc., 2023) Shetty, R.D.; Dewangan, S.K.; Bhattacharjee, S.Location-based social networks (LBSN) data is of greater importance in various domains of studies, such as viral marketing, location recommendation, influence maximization problem, etc. Identifying influential or hotspot locations of disease spread is of great importance during epidemic outbreaks, in case of limited vaccination, to make timely decisions about stay-at-home policy, restricting the transportation/human movement from one geographical location to another, etc. LBSN data can be used to construct the location-based weighted spatio-temporal network with its defined feature sets, including its rich collection of user movement information with respect to space and time. Here, we design two algorithms to generate spatio-temporal edge weights in such networks. We then examine four benchmark algorithms to identify the importance of these derived edge features to evaluate location influence in the context of contagious disease spread. It is observed that deriving edge features play an important role in the application scenario, especially to understand the need of dense network compared to the sparse network, generated from the LBSN data. © 2023 IEEE.Item A Comparative Study of Temporal and Static Networks in Modeling Pathogen Transmission within School Environments(Institute of Electrical and Electronics Engineers Inc., 2024) Shetty, R.D.; Bhattacharjee, S.In the domain of epidemiology and infectious disease control, understanding the dynamics of pathogen transmission is crucial. This study shows the dynamics of pathogen spread in school environments by comparing two distinct network models: temporal networks and static networks. The objective is to assess their respective effectiveness in modeling the transmission of pathogens within school settings. Temporal networks take into account the evolving nature of social interactions over time, offering a more realistic representation of human contact patterns. In contrast, static networks provide a simplified but stable view of social connections. This research investigates how these two modeling approaches impact our ability to analyze, and control the spread of pathogens, such as viruses, in school environments. We Propose T-GSI (Temporal Global Structure Influence) to measure the influence of the temporal networks, which outperformed the state-of-the-art methods. Our findings reveal the strengths and weaknesses of temporal and static networks in capturing the influence of pathogen transmission within school communities. © 2024 IEEE.Item Does Degree Capture It All? A Case Study of Centrality and Clustering in Signed Networks(Association for Computing Machinery, Inc, 2025) Murali, S.S.; Abhin, B.; Shetty, R.D.; Bhattacharjee, S.Signed graph networks are used to model systems that contain both positive and negative components. By incorporating signed information into Graph Neural Networks (GNNs), allow for the analysis of complex interactions between nodes, facilitating tasks such as sentiment analysis and trust prediction in social networks. Our main goal in this study is to improve feature selection in a benchmark GNN, Signed Graph Attention (SiGAT) by including centrality and clustering measures other than degree. Our studies reveal that using both degree and centrality features slightly improves signed link prediction performance. Further, our ablation studies revealed that 6 degree features and 16 attention heads optimally encode information and reduce noise. © 2024 Copyright held by the owner/author(s).Item GSI: An Influential Node Detection Approach in Heterogeneous Network Using Covid-19 as Use Case(Institute of Electrical and Electronics Engineers Inc., 2023) Shetty, R.D.; Bhattacharjee, S.; Dutta, A.; Namtirtha, A.The growth of COVID-19, caused by the SARS-CoV-2 virus, has turned into an unprecedented pandemic in the last century. It is crucial to identify superspreading nodes to prevent the pandemic's progress. Most available superspreader identification techniques consider only a single or few network metrics related to the complex network's topological structure. Furthermore, it is more challenging to determine influential spreaders from heterogeneous structures of networks. In a disease transmission network, the degree of heterogeneity is essential to locate the path of the infection spread. Therefore, it is required to have an extended degree of centrality to collect information from various neighborhood levels. This article presents an approach, namely, global structure influence (GSI), which considers network nodes' local and global influence. This method can gather information from multiple levels of the neighborhood. Evaluation of our proposed method is done by considering different types of networks, i.e., social networks, highly heterogeneous human contact networks, and epidemiological networks, and also by using the benchmark susceptible-infected-recovered (SIR) epidemic model. The GSI technique provides real-spreading dynamics across various network structures and has outperformed the baseline techniques with an average Kendall's τ improvement range from 0.017 to 0.278. This study will help to identify the superspeaders in real applications, where pathogens spread quickly because of close contact, such as the recently witnessed COVID-19 pandemic. © 2014 IEEE.