This study presents a novel approach to the increasingly important task of distinguishing AI-generated images from authentic photographs. The detection of such synthetic content is critical for combating deepfake misinformation and ensuring the authenticity of digital media in journalism, forensics, and online platforms. A custom-designed Vision Transformer (ViT) model, termed Patch-Based Vision Transformer for Identifying Synthetic Media (PV-ISM), is introduced. Its performance is benchmarked against innovative transfer learning methods using 60,000 authentic images from the CIFAKE dataset, which is derived from CIFAR-10, along with a corresponding collection of images generated using Stable Diffusion 1.4. PV-ISM incorporates patch extraction, positional encoding, and multiple transformer blocks with attention mechanisms to identify subtle artifacts in synthetic images. Following extensive hyperparameter tuning, an accuracy of 96.60% was achieved, surpassing the performance of ResNet50 transfer learning approaches (93.32%) and other comparable methods reported in the literature. The experimental results demonstrate the model’s balanced classification capabilities, exhibiting excellent recall and precision throughout both image categories. The patch-based architecture of Vision Transformers, combined with appropriate data augmentation techniques, proves particularly effective for synthetic image detection while requiring less training time than traditional transfer learning approaches.

Federated learning is a collaborative machine learning paradigm where multiple parties jointly train a predictive model while keeping their data. On the other hand, multi-label learning deals with classification tasks where instances may simultaneously belong to multiple classes. This study introduces the concept of Federated Multi-Label Learning (FMLL), combining these two important approaches. The proposed approach leverages federated learning principles to address multi-label classification tasks. Specifically, it adopts the Binary Relevance (BR) strategy to handle the multi-label nature of the data and employs the Reduced-Error Pruning Tree (REPTree) as the base classifier. The effectiveness of the FMLL method was demonstrated by experiments carried out on three diverse datasets within the context of animal science: Amphibians, Anuran-Calls-(MFCCs), and HackerEarth-Adopt-A-Buddy. The accuracy rates achieved across these animal datasets were 73.24%, 94.50%, and 86.12%, respectively. Compared to state-of-the-art methods, FMLL exhibited remarkable improvements (above 10%) in average accuracy, precision, recall, and F-score metrics.

Gastroesophageal reflux disease (GERD) is a multidisciplinary disease; therefore, when treating GERD, a large amount of data needs to be monitored and managed.The aim of our study was to develop a novel automation and decision support system for GERD, primarily to automatically determine GERD and its Chicago Classification 3.0 (CC 3.0) phenotypes. However, phenotyping is prone to errors and is not a strategy widely known by physicians, yet it is very important in patient treatment. In our study, the GERD phenotype algorithm was tested on a dataset with 2052 patients and the CC 3.0 algorithm was tested on a dataset with 133 patients. Based on these two algorithms, a system was developed with an artificial intelligence model for distinguishing four phenotypes per patient. When a physician makes a wrong phenotyping decision, the system warns them and provides the correct phenotype. An accuracy of 100% was obtained for both GERD phenotyping and CC 3.0 in these tests. Finally, since the transition to using this developed system in 2017, the annual number of cured patients, around 400 before, has increased to 800. Automatic phenotyping provides convenience in patient care, diagnosis, and treatment management. Thus, the developed system can substantially improve the performance of physicians.

Planning a journey by integrating route and timetable information from diverse sources of transportation agencies such as bus, ferry, and train can be complicated. A user-friendly, informative journey planning system may simplify a plan by providing assistance in making better use of public transportation. In this study, we presented the service-oriented, multimodel Intelligent Journey Planning System, which we developed to assist travelers in journey planning. We selected Izmir, Turkey, as the pilot city for this system. The multicriteria problem is one of the well-known problems in transportation networks. Our study proposes a gradual path-finding algorithm to solve this problem by considering transfer count and travel time. The algorithm utilizes the techniques of efficient algorithms including round based public transit optimized router, transit node routing, and contraction hierarchies on transportation graph. We employed Dijkstra’s algorithm after the first stage of the path-finding algorithm by applying stage specific rules to reduce search space and runtime. The experimental results show that our path-finding algorithm takes 0.63 seconds of processing time on average, which is acceptable for the user experience.

Noisy data is a prevalent issue in data mining, significantly impacting the performance of classification algorithms. Mathematical methods are crucial in tackling this obstacle, particularly in optimizing noise detection and data preprocessing. This study proposes a novel approach—Joint Tomek Links (JTL)— to identify and eliminate noisy instances by detecting pairs of nearest neighbors from different classes. It first finds the Tomek links and then refines a probabilistic method to determine which instance from a pair will be removed. In our approach, a random tree classifier serves as the base model. We conducted experiments on 40 benchmark datasets spanning various domains, achieving an average classification accuracy of 83.26% for JTL. The results demonstrate that the JTL attains an average improvement of 5.33% in accuracy compared to the original classification with a random tree. Furthermore, JTL surpasses existing techniques, delivering a noteworthy gain in accuracy by 12.30% on the same datasets. These findings underscore the effectiveness of JTL in enhancing data quality and boosting classification performance in data mining tasks.

Large numbers of job postings with complex content can be found on the Internet at present. Therefore, analysis through natural language processing and machine learning techniques plays an important role in the evaluation of job postings. In this study, we propose a novel data structure and a novel algorithm whose aims are effective storage and analysis in data warehouses of big and complex data such as job postings. State-of-the-art approaches in the literature, such as database queries, semantic networking, and clustering algorithms, were tested in this study to compare their results with those of the proposed approach using 100,000 Kariyer.net job postings in Turkish, which can be considered to have an agglutinative language with a grammatical structure differing from that of other languages. The algorithm proposed in this study also utilizes stream logic. Considering the growth potential of job postings, this study aimed to recommend new sub-qualifications to advertisers for new job postings through the analysis of similar postings stored in the system. Finally, complexity and accuracy analyses demonstrate that the proposed approach, using the Cluster Feature approach, can obtain state-of-the-art results on Turkish job posting texts.