ABSTRACT Link prediction is a fundamental task in social and complex network analysis, focused on forecasting the likelihood of unseen or future connections between nodes. Accurate link prediction can enhance understanding of network dynamics, reveal hidden structures, and improve recommendations in social and information networks. This paper proposes a novel Meta‐Learning‐Based Link Prediction model that utilizes a Centrality‐Aware connectivity matrix and incorporates Deep Graph Infomax (DGI) embeddings with the CatBoost classifier. The connectivity matrix is constructed using node centrality measures like closeness centrality, degree centrality, and betweenness centrality by capturing the network's local and global structural properties. The DGI embedding algorithm efficiently learns the network's latent features, while the CatBoost classifier is employed to enhance prediction performance. To address the challenge of imbalanced datasets in social networks, we apply downsampling to create balanced training and testing datasets, ensuring robust model learning. Our framework demonstrates improved accuracy, scalability, and adaptability compared to traditional link prediction methods. Extensive experiments on real‐world social network datasets show that the proposed model achieves superior performance in link prediction tasks, making it a promising approach for various network analysis applications.

ABSTRACT Link prediction (LP) is a prominent research topic in network science and complex network analysis, focused on predicting future connections between unconnected node pairs by examining network topology and related characteristics. This paper introduces an advanced LP model combining the graph attention network (GAT) with Jaccard similarity and node centrality measures, such as local interaction density (LID) and hubs and authority centrality (HAC). Initially, a weight matrix incorporating structural and topological information is created using a feature matrix derived from similarity measures and node centrality. Then, node embeddings are generated using GAT, allowing the model to learn detailed representations that consider local and global contexts. GAT's multi‐head attention mechanism enables the model to focus on various aspects of the node neighbourhood, capturing diverse structural information. A well‐defined dataset is created from these embeddings, representing nodes at the endpoints of edges labelled as positive or negative. Linear discriminant analysis (LDA) is applied to this well‐balanced and labelled dataset to perform classification tasks, leading to accurate link prediction. The model's performance is evaluated across eight different datasets, and the obtained results reveal the proposed model's superiority over several baseline and recently proposed LP models.

Abstract In recent years, the study of artificial intelligence (AI) has undergone a paradigm shift. This has been propelled by the groundbreaking capabilities of generative models both in supervised and unsupervised learning scenarios. Generative AI has shown state-of-the-art performance in solving perplexing real-world conundrums in fields such as image translation, medical diagnostics, textual imagery fusion, natural language processing, and beyond. This paper documents the systematic review and analysis of recent advancements and techniques in Generative AI with a detailed discussion of their applications including application-specific models. Indeed, the major impact that generative AI has made to date, has been in language generation with the development of large language models, in the field of image translation and several other interdisciplinary applications of generative AI. Moreover, the primary contribution of this paper lies in its coherent synthesis of the latest advancements in these areas, seamlessly weaving together contemporary breakthroughs in the field. Particularly, how it shares an exploration of the future trajectory for generative AI. In conclusion, the paper ends with a discussion of Responsible AI principles, and the necessary ethical considerations for the sustainability and growth of these generative models.

Activity recognition (AR) using surveillance cameras remains challenging because of the variation between the data required for deep learning and the capabilities of CCTV-based frameworks for data recording. Most research on skeleton-based activity focuses on improving recognition using complete data, often overlooking performance on incomplete skeleton data due to occlusion or noise. However, in practical situations, it is inevitable to obtain a noisy or occluded human skeleton. When some informative keyjoints are unavailable or disturbed, it can compromise the efficiency and effectiveness of existing methods. This article proposes a skeleton-based model for activity recognition to deal with the occlusion problem. The main approach of the proposed model is to address occlusion by treating it as a missing value imputation problem within the feature matrix. First, it extracts features such as normalized joints, relative joint coordinates, angles, and velocity using a CNN-based technique and encodes the skeleton data into a feature matrix. The innovative use of generative adversarial imputation networks (GAIN) with bidirectional long short-term memory (Bi-LSTM), enhanced by spatial and temporal attention mechanisms, effectively handled missing data due to occlusions, ensuring a complete feature matrix for subsequent classification. Additionally, this article introduces autoencoder-based Bi-LSTM with attention modules that improve activity classification by capturing and leveraging spatial and temporal dependencies. A comprehensive experiment was conducted on both occluded and unoccluded datasets to validate the effectiveness and improvements of the proposed model. Comparative results show that the proposed model outperforms the existing state-of-the-art methods, achieving superior recognition accuracy in an occluded environment.

Online social media allows users to connect with a large number of people across the globe and facilitate the exchange of information efficiently. These platforms cater to many of our day-to-day needs. However, at the same time, social media have been increasingly used to transmit negative stances such as derogatory language, hate speech, and cyberbullying. The task of identifying the negative stances from social media posts or comments or tweets is termed negative stance detection . One of the major challenges associated with negative stance detection is that most of the content published on social media is often in a multilingual format. This work aims to identify negative stances from multilingual data streams in low-resource languages on social media using a hybrid transfer learning and deep convolutional neural network approach. The proposed work starts by preprocessing the multilingual datasets by removing irrelevant information such as special characters and hyperlinks. The processed dataset is then passed through a pretrained BERT (bidirectional encoder representations from Transformers) model to generate embeddings by fine-tuning the model as per the dataset under consideration. The generated word embeddings are then passed to a deep convolutional neural network for extracting the latent features from the texts and removing the unessential information. This helps our model to achieve robustness and effectiveness for efficient learning on the given dataset and make appropriate predictions on zero-shot data. The article utilizes several optimization strategies for examining the impact of fine-tuning different BERT layers on the model’s performance. Intensive experiments on a variety of languages — namely, English, French, Italian, Danish, Arabic, Spanish, Indonesian, German, and Portuguese — are performed. The experimental results demonstrate the effectiveness and efficiency of the proposed framework.

Humour is a crucial aspect of human speech, and it is, therefore, imperative to create a system that can offer such detection. While data regarding humour in English speech is plentiful, the same cannot be said for a low-resource language like Hindi. Through this article, we introduce two multimodal datasets for humour detection in the Hindi web series. The dataset was collected from over 500 minutes of conversations amongst the characters of the Hindi web series Kota-Factory and Panchayat . Each dialogue is manually annotated as Humour or Non-Humour. Along with presenting a new Hindi language-based Humour detection dataset, we propose an improved framework for detecting humour in Hindi conversations. We start by preprocessing both datasets to obtain uniformity across the dialogues and datasets. The processed dialogues are then passed through the Skip-gram model for generating Hindi word embedding. The generated Hindi word embedding is then passed onto three convolutional neural network (CNN) architectures simultaneously, each having a different filter size for feature extraction. The extracted features are then passed through stacked Long Short-Term Memory (LSTM) layers for further processing and finally classifying the dialogues as Humour or Non-Humour. We conduct intensive experiments on both proposed Hindi datasets and evaluate several standard performance metrics. The performance of our proposed framework was also compared with several baselines and contemporary algorithms for Humour detection. The results demonstrate the effectiveness of our dataset to be used as a standard dataset for Humour detection in the Hindi web series. The proposed model yields an accuracy of 91.79 and 87.32 while an F1 score of 91.64 and 87.04 in percentage for the Kota-Factory and Panchayat datasets, respectively.

Exposure to half-truths or lies has the potential to undermine democracies, polarize public opinion, and promote violent extremism. Identifying the veracity of fake news is a challenging task in distributed and disparate cyber-socio platforms. To enhance the trustworthiness of news on these platforms, in this article, we put forward a fake news detection model, OptNet-Fake. The proposed model is architecturally a hybrid that uses a meta-heuristic algorithm to select features based on usefulness and trains a deep neural network to detect fake news in social media. The$d$-D feature vectors for the textual data are initially extracted using the term frequency inverse document frequency (TF-IDF) weighting technique. The extracted features are then directed to a modified grasshopper optimization (MGO) algorithm, which selects the most salient features in the text. The selected features are then fed to various convolutional neural networks (CNNs) with different filter sizes to process them and obtain the$n$-gram features from the text. These extracted features are finally concatenated for the detection of fake news. The results are evaluated for four real-world fake news datasets using standard evaluation metrics. A comparison with different meta-heuristic algorithms and recent fake news detection methods is also done. The results distinctly endorse the superior performance of the proposed OptNet-Fake model over contemporary models across various datasets.

Temporal link prediction (TLP) is a prominent problem in network analysis that focuses on predicting the existence of future connections or relationships between entities in a dynamic network over time. The predictive capabilities of existing models of TLP are often constrained due to their difficulty in adapting to the changes in dynamic network structures over time. In this article, an improved TLP model, denoted as TLP-NEGCN, is introduced by leveraging network embedding, graph convolutional networks (GCNs), and bidirectional long short-term memory (BiLSTM). This integration provides a robust model of TLP that leverages historical network structures and captures temporal dynamics leading to improved performances. We employ graph embedding with self-clustering (GEMSEC) to create lower dimensional vector representations for all nodes of the network at the initial timestamps. The node embeddings are fed into an iterative training process using GCNs across timestamps in the dataset. This process enhances the node embeddings by capturing the network's temporal dynamics and integrating neighborhood information. We obtain edge embeddings by concatenating the node embeddings of the end nodes of each edge, encapsulating the information about the relationships between nodes in the network. Subsequently, these edge embeddings are processed through a BiLSTM architecture to forecast upcoming links in the network. The performance of the proposed model is compared against several baselines and contemporary TLP models on various real-life temporal datasets. The obtained results based on various evaluation metrics demonstrate the superiority of the proposed work.

The Asian social networks are dominated by the society’s collectivist culture, and this interestingly introduces an influence mechanism aided by word-of-mouth and opinion leaders. An opinion leader can help to generate and shape other people’s opinion and achieve a high information spread on any topic. In this work, a modified spider monkey optimization based opinion leader detection approach is proposed. Firstly, we employ the modified node2vec graph embedding to generate the lower dimensional vectors which act as the initial features for the nodes in a typical Asian social network. Next, the entire population is broken down into several groups using the k-means++ algorithm where the number of clusters is equal to the number of opinion leaders to be selected. The local and global leaders are chosen by using the coordinates of the cluster centres of these clusters. The coordinates of the centroids of the clusters are then used to detect the local and global leaders in the network. The local leaders then form the seed set of opinion leaders for the network. The positions of the nodes in the network, including the local and global leaders, are updated over a number of iterations. At the end of these iterations, the seed set generating the maximum influence forms the set of opinion leaders in the network. We test our proposed approach using the popular information diffusion and cognitive opinion dynamics (COD) models. We perform intensive experiments on several real-life social networks based on various performance metrics. The results obtained reveal that the proposed approach outperforms several existing techniques of opinion leader detection.

Finding influential nodes reshuffles the very notion of linear paths in business processes and replaces it with networks of business value within a smart enterprise system. There are many existing algorithms for identifying influential nodes with certain limitations for applying in large-scale networks. In this article, we propose a community structure with integrated features ranking (CIFR) algorithm to find influential nodes in the network. First, we use the community detection algorithm to find communities in the system, and then we rank the nodes of network based on three factors, namely local ranking, gateway ranking, and community ranking, collectively termed as integrated features. Our algorithm intends to select influential nodes, which are both globally and locally optimal, leading to overall high information propagation. We perform the experimental results on total eight networks using various evaluation parameters. The obtained results validate superior performance against contemporary algorithms adding value to smart enterprises.