Sürdürülebilir Kalkınma Amaçları
Yazarlar
- Yildirim, Ozal(19)
- Utku, Anıl(11)
- Talo, Muhammed(10)
- Utku, Anil(10)
- Baloglu, Ulas Baran(9)
- Acharya, U. Rajendra(7)
- Ergen, Burhan(7)
- Doğan, Gürkan(5)
- Karakose, Mehmet(5)
- Yilmazer, Merve(5)
- Can, Umit(4)
- Demir, Yakup(4)
- Imak, Andaç(4)
- Acharya, U Rajendra(3)
- Aydin, Galip(3)
- Ertargin, Merve(3)
- Gunay, Mihriban(3)
- Tan, Ru-San(3)
- Yıldırım, Özal(3)
- Akcayol, M. Ali(2)
- Ay, Betul(2)
- Barut, Abdulkadir(2)
- Iliev, Teodor(2)
- Karabatak, Murat(2)
- Kaya, Sema Kayapinar(2)
- Korkmaz, Adem(2)
- Kosunalp, Selahattin(2)
- Kutlu, Banu(2)
- Orhan, Ahmet(2)
- Pilatin, Abdulmuttalip(2)
- Rajendra Acharya, U.(2)
- Sengur, Abdulkadir(2)
- Tan, Ru San(2)
- Türkoğlu, Muammer(2)
- Çelebi, Adalet(2)
- Çelik, Yusuf(2)
- Şengür, Abdulkadir(2)
- Agdas, Mehmet T.(1)
- Agdas, Mehmet Tevfik(1)
- Akcayol, Muhammet Ali(1)
Yayınlar
- MakaleYalnız künye2026Scopus SJRQ1MILA: An Innovative Approach to Identifying AI-Generated Content Using BERT, CNN, and BiLSTM
Utku, Anil
AI ethical principles are essential in balancing the social impacts of rapidly developing technology today. The widespread use of AI technologies has heightened the need for ethical and legal regulations in this area. In this process, fundamental principles such as data privacy, accountability, transparency, and human oversight take center stage to enhance the reliability of artificial intelligence and foster social trust. Generative AI refers to machines producing text, images, audio, and video content. However, generative AI brings ethical discussions since it can produce original content. Detecting AI-generated content is becoming increasingly complex due to the ability of AI models to produce increasingly more fluent and meaningful texts. This study created a hybrid MILA model using BERT, CNN, and BiLSTM models to detect AI-generated content. MILA was compared in detail with SVM, XGBoost, CNN, LSTM, and BiLSTM using sentence-level and article-level datasets. Experiments demonstrated that MILA outperformed the compared models, achieving 94.34% accuracy at the sentence-level and 93.62% accuracy at the article-level.
AtıfOpenAlex1
İncele - MakaleAçık Erişim2026Scopus SJRQ1Multi-Task NILM with Anomaly Detection Using a Hybrid CNN–BilSTM–Transformer Model
Gunay, Mihriban; Demir, Yakup; Zhilevski, Marin
Non-Intrusive Load Monitoring (NILM) enables estimation of the energy use of individual appliances in smart buildings from a single aggregate meter. In practice, however, this task is not straightforward. Signals from different appliances can overlap, and the measured data may also include distortions such as spikes, drops, and noise. To address these issues, this study presents a multi-task triple-hybrid deep learning framework that handles appliance classification and anomaly detection together. The model brings together 1D-CNN, BiLSTM, and Transformer Attention so that local patterns, temporal dependencies, and wider contextual information can be learned within the same structure. It also uses a dual-output design to classify appliance categories and detect anomaly types simultaneously. Experiments were carried out on Building 1 of the UK-DALE dataset with four appliances: kettle, microwave, washer dryer, and fridge freezer. For the anomaly task, synthetic disturbances were added to segmented signal windows and grouped as normal, spike, drop, and noise. To check how well the proposed framework handled different scenarios, it was tested on both the UK-DALE and REDD datasets. Looking at the main UK-DALE results, the model correctly identified appliances 99.48% of the time and spotted anomalies with 98.80% accuracy. A secondary test on the REDD dataset yielded an 86.44% classification score. This proves the architecture can adjust to completely new power grid environments without losing its edge. On top of that, when pitted against standard benchmark models like Seq2Point, this triple-hybrid design clearly does a better job of mapping out complex signal changes. As a result, it yields much stronger anomaly detection metrics.
AtıfOpenAlex1İncele - MakaleAçık Erişim2026Scopus SJRQ2A Genetic Algorithm-Optimized ConvNeXtV2-YOLOv8 Framework for Dental Caries Detection in Panoramic Radiographs
Sobahi, Nebras; Küçük, Deniz Bora; Kılıç, Kazım; İmak, Andaç; Çelebi, Adalet; Alghamedi, Yazyd; Yazicioglu, Cafer; Türkoğlu, Muammer; Şengür, Abdulkadir
Background/Objectives: Dental caries is one of the most common oral diseases worldwide, and early diagnosis is essential for effective treatment. However, detecting carious lesions in panoramic radiographs is challenging because of low image contrast, anatomical complexity, and overlapping structures. This study aimed to develop an improved object detection framework for dental caries localization in panoramic radiographs. Methods: A modified YOLOv8 architecture was developed by integrating ConvNeXtV2 blocks into the Cross-Stage Partial (C2f) modules to enhance feature extraction and gradient propagation. To improve detection performance and reduce overfitting, key training hyperparameters were optimized using a genetic algorithm. The proposed model was evaluated on a dataset of 474 panoramic dental radiographs annotated by experts using bounding boxes. Results: The optimized model achieved a box precision of 78.4%, a recall of 53.6%, and a mean average precision at 50% intersection-over-union threshold (AP50) of 62.6%. Compared with the baseline YOLOv8 model, the proposed approach improved precision and AP50. Visual and quantitative analyses demonstrated that the ConvNeXtV2-enhanced architecture enabled more accurate localization of carious regions. Conclusions: The results indicate that combining ConvNeXtV2-based architectural enhancement with genetic algorithm-based hyperparameter optimization is an effective strategy for dental caries localization in panoramic radiographs. Although recall remains a limitation, the proposed framework shows potential as a computer-aided diagnostic tool for supporting clinical caries assessment.
- MakaleAçık Erişim2026Scopus SJRQ1Surface EMG-Based Hand Gesture Recognition Using a Hybrid Multistream Deep Learning Architecture
Çelik, Yusuf; Can, Umit
Surface electromyography (sEMG) enables non-invasive measurement of muscle activity for applications such as human–machine interaction, rehabilitation, and prosthesis control. However, high noise levels, inter-subject variability, and the complex nature of muscle activation hinder robust gesture classification. This study proposes a multistream hybrid deep-learning architecture for the FORS-EMG dataset to address these challenges. The model integrates Temporal Convolutional Networks (TCN), depthwise separable convolutions, bidirectional Long Short-Term Memory (LSTM)–Gated Recurrent Unit (GRU) layers, and a Transformer encoder to capture complementary temporal and spectral patterns, and an ArcFace-based classifier to enhance class separability. We evaluate the approach under three protocols: subject-wise, random split without augmentation, and random split with augmentation. In the augmented random-split setting, the model attains 96.4% accuracy, surpassing previously reported values. In the subject-wise setting, accuracy is 74%, revealing limited cross-user generalization. The results demonstrate the method’s high performance and highlight the impact of data-partition strategies for real-world sEMG-based gesture recognition.
AtıfScopus4WoS4OpenAlex7
İncele - MakaleAçık Erişim2026Scopus SJRQ1Quantifying agricultural resilience under climate variability: a data-driven climate resilience index for European cereal systems
Utku, Anil; Radulescu, Magdalena; Barut, Abdulkadir; Alofaysan, Hind
Climate variability and extreme weather events pose an increasingly serious threat to agricultural productivity in Europe. Although many studies have focused on the mean yield response to climate change, relatively few have attempted to offer a comprehensive, multi-dimensional view of agricultural resilience, separating long-term productivity trends from short-term climate-induced fluctuations. In this paper, a data-driven Climate Resilience Index is proposed to assess the resilience of European cereal systems under climate stress. Annual yield time series are decomposed into trend and anomaly components to identify interannual deviations driven by climate variability. Machine learning and deep learning algorithms, including RF, CatBoost, CNN, LSTM, TCN, and a TCN-LSTM hybrid model, are used to assess the predictive validity of climate extreme indices and persistence patterns. Among the evaluated models, the TCN-LSTM architecture achieved the highest predictive performance with 0.8347 R2, outperforming standalone temporal and ensemble-based models. The PCA-based exposure framework explained 68.4% of the total climate variability within the first principal component, indicating a highly concentrated climate-stress structure. Feature importance analyses further revealed that lagged yield anomalies and rolling volatility indicators were the dominant predictors of resilience dynamics. The findings show that temporal models perform better than traditional ensemble methods, emphasizing the significance of multi-year recovery mechanisms in anomaly evolution. To provide a consistent measure of exposure, principal component analysis is used to reduce correlated climate indicators into a multi-dimensional climate stress index. The CRI captures these three essential elements: climate exposure, yield variability, and recovery. The results at the country level show substantial variability across Europe, indicating that while exposure intensity is important, recovery dynamics are also critical to understanding resilience. The results also show that there are structurally distinct resilience profiles, and robustness tests indicate that country rankings are not sensitive to different weights and recovery functions. The results clearly show that recovery and adaptive capacity are critical to agricultural sustainability amid growing climate uncertainty. The CRI developed in this paper provides a clear, empirically valid framework for comparing resilience in agricultural systems.
AtıfScopus1OpenAlex2İncele - MakaleYayıncıda açık erişim2026Scopus SJRQ1Explainable AI-based deep learning architectures for hydrogen prediction in biomass syngas and photocatalysis processes: Towards a sustainable and efficient energy economy
Imak, Andaç
This study, focusing on sustainable energy production and particularly the hydrogen (H2) economy, presents an advanced artificial intelligence (AI) framework for accurately predicting H2 production from sucrose photocatalysis catalyzed by LaFeO3 (GLFO) and the H2 concentration in syngas derived from biomass conversion. The research was conducted along two main directions: (i) deep learning (DL) architectures comprising nine specially designed recurrent neural network-based models and (ii) six conventional machine learning (ML) algorithms. For both H2 production systems, the best-performing models were identified and further interpreted using explainable AI (XAI) based on SHapley Additive exPlanations (SHAP). Among all investigated algorithms, a hybrid One-dimensional Convolutional Neural Network–Bidirectional Long Short-Term Memory–Bidirectional Gated Recurrent Unit (1D-CNN–BiLSTM–BiGRU) model demonstrated the highest predictive performance. Using 10-fold cross-validation, the proposed model achieved outstanding accuracy, with an R2 value of 0.9987 ± 0.0011 for photocatalytic H2 production prediction and 0.9980 ± 0.0006 for syngas H2 concentration prediction. These results highlight the potential of the proposed framework as a precise, reliable, and practical methodological approach for modeling and optimizing complex thermochemical energy systems.
- MakaleAçık Erişim2026Scopus SJRQ2A Unified and Interpretable Benchmark of Classification Models for DGA-Based Power Transformer Fault Diagnosis
Doğan, Kıvanç; Ertarğın, Merve; Çalışkan, Abuzer; Bakay, Ayşenur
Power transformers are among the most critical components of electric power transmission and distribution systems, and unexpected failures can lead to substantial economic losses and prolonged power outages. Dissolved Gas Analysis (DGA) is the most widely used diagnostic technique for transformer fault diagnosis and involves interpreting gases dissolved in insulating oil. However, conventional interpretation methods, such as the Rogers ratio, the Doernenburg ratio, the IEC 60599 ratio method, and the Duval Triangle, rely heavily on expert knowledge, may produce inconsistent diagnoses for the same oil sample, and may fail to provide a diagnosis in certain cases. In this study, 12 classification models were evaluated using the publicly available Power Transformers Fault Detection and Diagnosis (FDD) and Remaining Useful Life (RUL) dataset and a unified evaluation protocol. Model performance was assessed using Accuracy, Balanced Accuracy, and Macro-F1 score, while model interpretability was investigated through Shapley Additive Explanations (SHAP) analysis. The results showed that ensemble tree-based methods achieved the best overall performance. LightGBM and Random Forest both attained an Accuracy of 0.969 and a Macro-F1 score of 0.924, while LightGBM further achieved the highest Balanced Accuracy of 0.940. XGBoost exhibited the most stable performance under cross-validation. SHAP analysis revealed that engineered relative concentration features, particularly the CO/H2 ratio and the combined gas ratio, were among the most influential features for fault classification. These findings demonstrate that, for datasets of this scale, ensemble tree-based models combined with well-designed features provide strong and interpretable performance for imbalanced DGA-based fault diagnosis, highlighting the effectiveness of feature-based ensemble learning for small- to medium-sized datasets.
- MakaleYalnız künye2026Scopus SJRQ1Analysis of the impact of the social justice index on the COVID-19 pandemic using artificial intelligence in EU and OECD countries
Utku, Anıl; Can, Ümit; BARUT, Abdulkadir; Pilatin, Abdulmuttalip; Sharabati, Abdel-Aziz Ahmad
The COVID-19 disease has evolved into a pandemic that affects the world in multiple dimensions. While COVID-19 spread rapidly to several nations, distinct patterns of spread emerged in each nation. This spreading behavior is related to various characteristics of countries, including social, political, and economic factors. In this study, an analysis of the COVID-19 pandemic in EU and OECD countries was conducted, taking into account the social justice index for the first time. The impact and significance of the social justice index on the COVID-19 analysis results were attempted to be revealed. As a result, it is understood that the countries with high Social Justice Index scores and sub-dimension scores are more successful in combating the COVID-19 pandemic. In addition, a new LSTM-based model was developed to predict the number of cases and deaths in these countries. The designed LSTM approach was compared with some well-known deep learning and machine learning algorithms. As a result of the experimental studies, the LSTM model obtained the best prediction results. Finally, the similarities between these countries in terms of the characteristics of the COVID-19 pandemic, including death and case spread, were analyzed, and exciting results were disclosed.
- MakaleAçık Erişim2026Scopus SJRQ1EBA-Net: A hybrid EfficientNetB3-BiLSTM-Attention model for species-level diatom classification
Balo Utku, Esen Damla; Kutlu, Banu; Utku, Anil
Species identification in diatoms is important for assessing water bodies, determining environmental quality, and analyzing biodiversity. Nonetheless, diatom classification has proven difficult due to intra-class variation, interclass similarity, and data imbalance in microscopic images. In this regard, we propose a new deep learning approach for species-level diatom classification. Various deep learning models, including traditional convolutional neural networks or CNNs (ResNet50, DenseNet121), state-of-the-art CNNs (EfficientNetB3, Xception, MobileNetV3), transformer-based architectures (ViT-B16, Swin-Tiny), and our proposed EBA-Net (EfficientNetB3-BiLSTM-Attention Network), were extensively compared under identical experimental settings. The EBA-Net integrates the advantages of convolutional feature extraction, sequence modeling, and an attention mechanism to improve classification accuracy. It first uses the EfficientNetB3 model as the backbone network to obtain higher-level features, followed by a BiLSTM layer to capture dependencies among the extracted features. Then, it applies the attention mechanism to the BiLSTM output to highlight the most informative features. Experimental results showed that the proposed EBA-Net model outperforms other models considered in the study, yielding 95.8% accuracy and 85.1% macro F1-score. In addition, Top-3 (98.9%) and Top-5 (99.6%) accuracies confirmed the efficiency of the proposed model for classifying objects from visually similar classes. An analysis of the model's interpretability was conducted using Gradient-weighted Class Activation Mapping, showing that the proposed model's attention is concentrated on relevant morphological features of diatoms. To the best of our knowledge, the suggested approach achieves state-of-the-art accuracy on the data set of diatoms used.
- MakaleYalnız künye2026Scopus SJRQ1FedMedIDS: A non-IID-resilient federated DCNN framework for securing IoMT and IoHT devices
Gencer, Burak; Dogan, Gurkan
The diversity of devices in the Internet of Medical Things (IoMT) and Internet of Health Things (IoHT) networks creates significant security vulnerabilities against cyberattacks. Traditional centralized machine learning approaches, however, are not always suitable for these environments due to data privacy regulations. In this study, a new Deep Convolutional Neural Network (DCNN) architecture is proposed for attack detection in IoMT/IoHT networks. The proposed architecture integrates the Residual Squeeze-and-Excitation (RSE) Block, the Squeeze-and-Excitation attention mechanism (SEM), and a one-dimensional multi-pooling approach. The model was evaluated within a federated learning (FL) framework called Federated Medical Intrusion Detection System (FedMedIDS), which preserves data privacy and demonstrates robustness in heterogeneous client scenarios. Four different training strategies—Centralized, FedAvg, FedProx, and Local-Only—were adopted for the evaluations. The proposed model and existing models were tested on the Canadian Institute for Cybersecurity – Internet of Medical Things 2024 (CICIoMT2024) and Edith Cowan University – Internet of Health Things (ECU-IoHT) datasets and compared with current deep learning models. The DCNN proposed for the CICIoMT2024 dataset and a Dirichlet distribution with α = 1.0 achieved F1 scores of up to 99.92% under FedAvg and 94.64% under FedProx with α = 0.5. At α = 0.05, the highest level of heterogeneity, it outperformed the Long Short-Term Memory (LSTM) by as much as 27%. On the ECU-IoHT dataset, F1 scores of up to 89.57% under FedAvg and 89.50% under FedProx were achieved at α = 1.0. This represents an improvement of up to 37% over LSTM at the same level of heterogeneity.
- MakaleAçık Erişim2026Scopus SJRQ2Data-Driven Prediction of Climate Variables in Agricultural Cities of India With Hybrid GA-TCN-LSTM Model
Utku, Anil
Accurate prediction of climate variables is important in reducing the effects of natural disasters and taking preventive measures for agriculture and food security, energy sector, public health, and water resources management. For agricultural production, it is essential for producers to determine strategies for their agricultural activities and to increase agricultural productivity. Issues such as determining planting and harvest times, determining pesticides to be used against agricultural pests, selecting products to be planted, irrigation, and sustainable agriculture are directly related to climate variables. In this study, a GA-TCN-LSTM hybrid prediction model was created to determine climate variables such as relative humidity, temperature, wind speed, and dew point in Bhopal, Indore, Kanpur, Ludhiana, and Patna, which are among the prominent cities of India in terms of agricultural production. The hyper-parameters of the developed model using the temporal convolutional networks (TCN) and long short-term memory (LSTM) were optimized with the genetic algorithm (GA), and the GA-TCN-LSTM hybrid model was created. GA-TCN-LSTM was extensively compared with the base TCN-LSTM, convolutional neural network (CNN), LSTM, TCN, and CNN-LSTM. The compared models were tested using approximately 15 years of hourly, up-to-date, and real-time data of the cities. Experiments showed that GA-TCN-LSTM outperformed the compared models and had above 0.9 R-Squared (R2) for the majority of cities and climate variables.
AtıfScopus2WoS1OpenAlex2İncele - MakaleAçık Erişim2025Advanced Classification of Poxvirus-Based Skin Diseases Using Deep Learning Techniques
Arik, Kaan; Agdas, Mehmet T.; Korkmaz, Adem; Kosunalp, Selahattin; Iliev, Teodor
Viral infections, especially those of the poxvirus family, present significant diagnostic challenges due to their similar clinical symptoms. This study proposes an innovative deep learning-based approach to classify six categories of poxvirus-related skin diseases: chickenpox, cowpox, healthy, measles, monkeypox, and smallpox. A dataset of 9,120 augmented images was used to train, validate, and test three advanced deep-learning models-YOLOv8, YOLOv5, and ResNet32. Among the models, YOLOv8 demonstrated superior performance, achieving an accuracy of 99.80%, precision of 99.28%, and recall of 99.14%, significantly outperforming YOLOv5 and ResNet32. The results underscore the potential of YOLOv8 in medical image analysis, providing a robust and efficient tool for the early detection and accurate classification of viral skin diseases. Comparisons with related studies highlight the effectiveness of the proposed approach, making it a state-of-the-art solution for improving diagnostic accuracy in healthcare. Future work will focus on extending the dataset and evaluating the model's applicability in real-time clinical environments.
AtıfOpenAlex1İncele - BildiriAçık Erişim2025Deep Learning-Based Approaches for Short-Term Residential Electricity Consumption Prediction: A Comparative Study of LSTM, CNN-LSTM, and CNN-GRU Models
Gunay, Mihriban; Yildirim, Ozal; Demir, Yakup; Benedicenti, L.; Liu, Z.
The rapid increase in the world population has increased electricity consumption and demand for electricity. In order for electricity distribution companies to determine electricity demand, household electricity consumption must be predicted. In this study, an open access dataset was used to predict individual household electricity consumption. For this purpose, deep learning-based LSTM, CNN-LSTM and CNN-GRU models were developed and a comparative analysis was performed. The performance of the models was evaluated using four evaluation metrics commonly used in prediction model performance evaluation. According to the results obtained, it was seen that the models used in the study were successful in predicting short-term electricity consumption. In particular, the CNNLSTM model was found to have a lower error rate compared to the other two models.
- MakaleYalnız künye2025Scopus SJRQ4Deep learning-based hybrid traffic flow prediction model for freight transport
Utku, Anıl; Kaya, Sema Kayapinar; Can, Umit
Urban freight transportation is crucial to sustaining domestic and international trade and fulfilling the regular requirements of community companies and customers. Therefore, this study developed an efficient convolutional neural network (CNN) and long short-term memory (LSTM) based hybrid deep learning model for predicting traffic flow in freight transportation. The number of freight trucks that would travel along 663 Street in Anderlecht, Belgium, and their average speed was predicted. The created model was compared with support vector machine (SVM), linear regression (LR), multilayer perceptron (MLP), random forest (RF), LSTM, and CNN using the success measures. The critical finding of this study is that combining the feature extraction ability of the 1D CNN model with the learning ability of the LSTM model from multivariate traffic data increases the traffic density prediction success compared to other traditional methods. Based on the results of the experiments, the CNN-LSTM model outperformed all the other models employed in this study. With this model, freight density can be successfully predicted, assisting in planning freight transport activities.
- MakaleAçık Erişim2025Scopus SJRQ1Automatic Segmentation of Asphalt Cracks on Highways After Large-Scale and Severe Earthquakes Using Deep Learning-Based Approaches
Yilmaz, Mehmet; Yalcin, Erkut; Demir, Fatih; Ozdemir, Ahmet Munir; Atar, Muhammed; Gunes, Aysegul; Furtana Yalcin, Beyza; Cambay, Ertugrul
This study develops a deep learning-based automated system for detecting and segmenting earthquake-induced asphalt cracks, offering a rapid and reliable solution for post-disaster road condition assessments. Unlike traditional manual inspections, which are time-consuming and error-prone, our approach leverages advanced segmentation techniques to ensure accurate, pixel-level classification of various crack types. The main challenge of this study was determining the damage caused to highways by earthquakes with magnitudes greater than 7.0, which occur approximately once every 200 years. The most crucial step in the automatic detection of these damages is the reliable preparation of a high-accuracy dataset. To achieve this, pixel-based labels were created by experts in the construction field by analyzing each pixel value. Following two major earthquakes, a unique dataset for segmenting roadway deterioration was created through intensive and detailed studies. This study aims to present the performance results of popular deep learning-based segmentation models in an unbiased manner, providing a feasible infrastructure for future real-time applications. The innovative aspect of this research lies in the creation of a unique post-earthquake dataset, collected and labeled from highways affected by the February 6, 2023 earthquakes in Turkey (Mw = 7.7 and Mw = 7.6). Deep learning models, including SegNet, Attention SegNet, U-Net, FCN (8s), and DeepLab, were trained and tested on this dataset. Among these, the SegNet model achieved the best performance with an average accuracy of 86.72%, precision of 92.99%, and sensitivity of 78.45%. By demonstrating superior performance metrics compared to existing methods, this study provides a robust framework for future infrastructure monitoring and maintenance strategies, ensuring safer and more resilient transportation networks in disaster-prone regions.
AtıfScopus6OpenAlex4İncele - MakaleAçık Erişim2025Scopus SJRQ2Hybrid ConvViT model for classification of dental radiography images; [Dental radyografi görüntülerinin sınıflandırılmasına yönelik hibrit ConvViT modeli]
Utku, Anıl
Dental radiography is used as an important diagnostic tool in the field of dentistry. Dental radiography provides detailed visualization of teeth, jaws and surrounding structures. Dental radiographs are manually interpreted by dentists for diagnosis and treatment purposes. Artificial intelligence methods are successfully applied to automate various image-related processes such as diagnosis and classification. In this study, a hybrid ConvViT model was developed to detect caries, fillings, impacted teeth, implants and normal teeth from dental radiography images. ConvViT combined the advantages of CNN and ViT architectures and provided more successful classification of dental radiography images. ConvViT was compared with popular pre-trained deep learning models such as ResNet-50, VGG-16, EfficientNetB0, and DenseNet201. Experimental results showed that ConvViT outperformed the compared models with 95% accuracy, 95% precision, 94% recall and 94% F-score. ConvViT makes a significant contribution to the development of artificial intelligence-based automatic diagnosis systems in the field of dentistry by effectively modeling both local and long-range features thanks to its hybrid structure.
AtıfScopus1
İncele - MakaleAçık Erişim2025Scopus SJRQ2Unraveling user perceptions and biases: A comparative study of ML and DL models for exploring twitter sentiments towards ChatGPT
Baker, Mohammed Rashad; Utku, Anil
ChatGPT is a powerful chatbot that used to generate human-like text. Chat GPT, developed by OpenAI, can answer many questions honestly, like a personal teacher who knows almost everything. It can perform various tasks such as question and answer, solving mathematical equations, writing text, debugging, and translating between languages. Compared to traditional chatbots, ChatGPT has been adopted by many users since its introduction. Some users feel that ChatGPT will override most content creation professions. This study analyzed users’ feelings about ChatGPT by analyzing tweets shared about ChatGPT. For this purpose, a hybrid Deep Learning (DL) model was developed using Convolutional Neural Networks (CNN) and Bi- Long Short-Term Memory (Bi-LSTM) models. The study has been compared with a number of DL and Machine Learning algorithms, LSTM, Bi-LSTM, CNN, Gated Recurrent Unit, Random Forests and Support Vector Machines. The experimental outcomes demonstrated that, the FastText-trained CNN-Bi-LSTM model exceeded the other models in terms of accuracy, reaching 96.59%.
AtıfScopus18WoS11OpenAlex13İncele - MakaleAçık Erişim2025Scopus SJRQ1Automated fault classification of asynchronous motor using mobile phone accelerometer and Parallel Residual CNN-GRU
Ertargin, Merve; Orhan, Ahmet; Yildirim, Ozal; Gurgenc, Turan
In this study, a fault detection system that classifies motor faults using deep learning without requiring any sensor connections inside or on the motor setup is proposed. The focus is on the electrical and mechanical faults of asynchronous motors, which are widely used in the industry. A mobile application developed using Flutter can collect vibration data using the three-axis accelerometer sensor in smartphones. By placing a smartphone with the application installed on the motor for which fault detection is desired, vibration data on three axes (x, y, z) is collected to train deep models. Subsequently, test data is used to identify the fault status and class of the electric motor. Different conditions and scenarios were created to test the adequacy of smartphone accelerometer sensors in detecting motor faults. Fault classification with vibration data obtained under different motor operating speeds in loaded and unloaded conditions achieved accuracy rates of over 99%.
AtıfScopus18OpenAlex15İncele - MakaleYalnız künye2025Scopus SJRQ2Enhanced gastrointestinal disease classification using a convvit hybrid model on endoscopic images
Utku, Anıl
Endoscopy is a procedure that allows examination of the gastrointestinal system, including the stomach, esophagus, large intestine, and duodenum, with the help of an endoscope. Processing of endoscopic images is important for early detection and treatment of gastrointestinal diseases. In this study, hybrid ConvViT was developed using CNN and ViT to increase the classification accuracy of pathologies in gastrointestinal endoscopic images. CNNs are well-suited for capturing local spatial features through hierarchical convolutions, making them highly effective in detecting fine-grained textures and edge patterns. These capabilities complement the ViT’s global attention mechanism, which excels at modeling long-range dependencies in images. The motivation of this study is to increase the classification accuracy and reliability with the ConvViT model, which was developed by combining the practical features of CNN and ViT models, which are individually successful in different aspects of image processing. The ConvViT model was compared with VGG-16, ResNet-50, Inception-V3 and ViT. Comparable models were tested using a gastrointestinal endoscopic image dataset containing ulcers, polyps, inflammation, bleeding, and regular anatomical features. Experiments showed that ConvViT had better prediction performance than compared models, with 95.87% classification accuracy.
AtıfScopus5WoS2OpenAlex2İncele - MakaleAçık Erişim2025Scopus SJRQ1A novel deep learning-based statistical randomness evaluation test methodology for cryptographic applications
Kaner, Seyfullah; Garipcan, Ali Murat; Erdem, Ebubekir
The security of cryptographic systems is directly linked to the statistical randomness properties of the random numbers used. Traditional statistical randomness tests can be limited in evaluating the properties of these numbers and can require long processing times. While widely used test suites such as the National Institute of Standards and Technology (NIST) Special Publication (SP) 800–22 play a crucial role in assessing data randomness, they are slow on large data sets, can only evaluate certain statistical properties, and fail to detect complex data patterns and dependency structures. In this paper, we propose a new deep learning (DL) based method to overcome these limitations. In the study, bit sequences produced by an FPGA-based true random number generator (TRNG) and different pseudo-random number generators (PRNGs) were converted into image format, and classification experiments were carried out on the AlexNet, ResNet50, and EfficientNetB0 architectures. The results showed that AlexNet, with 87% accuracy and 99% recall, was by far the most successful method compared to the other models. Furthermore, the ablation analysis revealed that the components of data augmentation, early stopping, and cross-validation play a critical role in the model’s stability and generalizability. Providing a reliable randomness evaluation with high accuracy, this new method, as an alternative to traditional statistical test suites, demonstrates that AI-based solutions can enhance the effectiveness and accuracy of randomness assessment in cryptographic applications. The success of DL methods in complex data structures offers significant potential for improving the security of modern cryptographic systems.
AtıfScopus9WoS9OpenAlex11İncele