۱.
Detecting fake news on social media platforms remains a significant challenge due to the dynamic nature of these networks, evolving user-news relationships, the difficulty in distinguishing real from fake information, and the use of advanced generative models to create fake content. In this study, we propose a novel approach, the Dynamic Graph Attention Network (DynGAT), for effective fake news detection. The DynGAT model utilizes the dynamic graph structure of social networks to capture the evolving interactions between users and news sources. It includes a graph construction module that updates the graph based on temporal data and a graph attention module that assigns importance to nodes and edges within the graph. The model applies attention mechanisms to prioritize critical interactions and uses deep learning techniques to classify news articles as real or fake. Experimental results on the TweepFake dataset (20,712 samples) show that DynGAT achieves 95% accuracy, outperforming existing methods such as Static GNN (87%), Transformer-based models (91%), and Hybrid models (89%). The model also demonstrates improvements in precision, recall, and F1 score. This work contributes to the ongoing efforts to combat misinformation and promote reliable information on social media platforms.
۲.
The verification of complex systems has traditionally relied on semi-automatic theorem-proving methods. However, model checking represents a paradigm shift by enabling automated, exhaustive verification of behavioral properties through systematic state exploration. Among advanced formal verification tools, Colored Petri Net (CPN) stands out for its integration of the ML programming language, facilitating robust model checking and system validation. Nevertheless, the application of CPN to complex systems is often constrained by the state-space explosion problem, which presents a significant challenge in contemporary research. While state-space analysis offers powerful capabilities for validation and scenario extraction, its potential remains largely untapped due to computational complexity constraints. This limitation is particularly pronounced in concurrent systems with multiple interacting variables, exemplified by game systems where intricate rule sets, deadlock conditions, and termination scenarios demand sophisticated modeling approaches. This paper presents a novel methodological framework for modeling and analyzing such game riddles, introducing methods to mitigate the state-space explosion problem. We demonstrate the efficacy of our approach through a comprehensive case study of the Merchant Ship puzzle game, though the methodology generalizes across various game typologies. By synthesizing model-checking techniques with ML-based algorithmic implementations, we develop an optimized search strategy for traversing the state space graph, enabling the derivation of quantitative complexity metrics. These metrics encompass critical indicators such as the success-to-total scenario ratio and the minimal trajectory length for both successful and unsuccessful game completions. Our research contributes to both the theoretical understanding of game complexity analysis and practical applications in game design through formal methods.
۳.
Multi-label text classification is a critical challenge in natural language processing, where the goal is to assign multiple labels to a given document. Recent advances have primarily focused on deep learning approaches, yet many fail to adequately capture the intricate relationships between documents and labels. In this paper, we propose a novel method called MultiCGCN, in which we leverage Graph Convolutional Networks (GCNs) for multi-label text classification by modeling text as a heterogeneous graph. This unified graph incorporates document similarities, label relationships, and document-label associations, enabling the model to effectively capture both document and label dependencies. We transform the multi-label classification problem into a link prediction task, using Term Frequency–Inverse Document Frequency (TF-IDF) for document similarity and applying GCNs to predict label assignments. Our empirical evaluations demonstrate that MultiCGCN achieves a significant performance boost, improving F1 score by 10% over traditional baseline models. This approach opens new avenues for enhancing the accuracy of multi-label classification in various domains.
۴.
Facial emotion recognition has recently attracted considerable interest due to its wide range of applications. It plays a crucial role in supporting individuals with autism spectrum disorders and improving interactions between humans and computers. The ability to execute these applications in real-time is essential. The architecture of the model and the computational resources available are the key determinants of inference time. Consequently, the development of a real-time solution requires a concentrated effort on these elements. In this paper, we present a scalable approach that utilizes EfficientNetV2, chosen for its operational efficiency. Our methodology involves resolution scaling based on a polynomial equation, which ensures real-time performance across various computational resources and model configurations. This scalable technique employs a polynomial equation to identify the optimal resolution for designated inference times, specifically adapted to our hardware and model specifications. By implementing the polynomial equation for resolution scaling, we created two variants of EfficientNetV2. Our findings from the KDEF dataset indicate that the proposed EfficientNetV2 can accurately classify images in real time on our hardware.
۵.
Securing computer networks against malicious attacks requires an efficient Network Intrusion Detection System (IDS). While machine learning techniques are commonly used for anomaly-based intrusion detection, data imbalance challenges conventional algorithms, leading to biased predictions and reduced accuracy. This study introduces a novel approach that combines ADASYN and Tomek links to address this issue, along with specific machine learning algorithms. ADASYN generates synthetic samples for the minority class to achieve dataset balance, and Tomek links eliminate redundant instances from the majority class. Four supervised machine learning algorithms (Random Forest, J48, Multilayer Perceptron, and Bagging) were assessed on both imbalanced and balanced datasets. Results show Random Forest exhibited 99.67% accuracy, while J48 and Bagging yielded 99.30%, and MLP recorded 98.53%. Notably, Random Forest emerges as a highly effective algorithm for Intrusion Detection, demonstrating flawless accuracy with balanced data. These outcomes highlight the proposed approach's ability to enhance prediction accuracy in network intrusion detection compared to imbalanced datasets, validated through a comparative analysis with state-of-the-art solutions.
۶.
In recent years, the field of medicine in Iran has faced significant public scrutiny, influenced by two major health crises: the 2016 death of acclaimed filmmaker Abbas Kiarostami due to a medical error and the COVID-19 pandemic. This study examines shifts in the emotional and discursive climate surrounding medicine and physicians on Persian Twitter before and after these events. Using relevant medical hashtags, over 131,000 tweets from 2015 to 2020 were analyzed through sentiment analysis, employing a rule-based approach with NVivo12 software.
The findings reveal a sevenfold increase in tweets about medicine during Kiarostami’s death, accompanied by heightened negativity and associations of terms like "error," "negligence," and "mistake" with medicine which has resulted in the reconstruction of 'medical error.' Additionally, as a result of the association of the terms' error,' 'negligence,' and 'mistake' with 'medicine,' the obviousness of the physician's holiness and respect for medicine has deteriorated, and the association of the terms' value and credibility' with medicine has been de-naturalized. However, during the COVID-19 pandemic, the sentiment shifted positively, reflecting greater appreciation for the medical profession.
This study highlights how public sentiment towards medicine changes in response to major health crises, emphasizing the interplay between public sphere and trust in healthcare systems. Understanding these dynamics can inform strategies to rebuild trust and address public concerns about medical practices.
۷.
With the expansion of smart homes, Human Activity Recognition (HAR) has become a key challenge in artificial intelligence, enhancing not only the comfort and safety of residents but also contributing to the development of applications such as healthcare and smart surveillance. The Transformer architecture, with its ability to model long-term dependencies and process data in parallel, has made significant advancements in recognizing human activities. In addition, its multi-head attention mechanism enables the analysis of complex input data by allowing the model to focus on different parts of the input simultaneously, capturing diverse relationships and dependencies within the data. This paper examines the application of Transformers in HAR and analyzes recent studies (since 2019). In addition to investigating innovative architectures, feature extraction methods, and accuracy improvements, it also discusses the challenges and future prospects of these models in recognizing human activities. Rapid advancements in deep learning and access to extensive datasets have made Transformers a key tool for improving the accuracy and efficiency of HAR systems in smart environments.
