Ilker Turker
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ilker Turker has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Ali Melih Kanca
Department of Computer Engineering Faculty of Engineering, Karabuk University Karabük Türkiye
1 shared publication Baha Alturan
Engineering Faculty Department of Computer Engineering Cankiri Karatekin University Cankiri Türkiye
1 shared publicationBurhan Selçuk
Department of Computer Science, Faculty of Science Ankara University Ankara Turkey
1 shared publication- Eftal Sehirli
Department of Computer Engineering Karabük University, Computing and Information Sciences Faculty Karabük Türkiye
1 shared publication - Ehssan Aljanabi
Department of Computer Engineering, Karabuk University, Karabuk 78050, Turkey
1 shared publication Elif Dorukbaşı
Department of Computer Engineering, Faculty of Computer and Information Sciences Karabuk University Turkey
1 shared publication
Latest Publications
- Open Access
- 14 July 2026
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 38 (14)
Swish‐Gumbel: A New Activation Function for Image Classification
In this study, we present Swish‐Gumbel, a new family of activation functions generated by combining the Swish and Gumbel functions through both multiplication and linear composition. We evaluate their performance on several image classification benchmarks, including MNIST, CIFAR‐10, Imagenette, and Beans, using different deep learning architectures. The results show that the multiplicative version consistently outperforms well‐known activation functions such as ReLU, Swish, Mish, and Gumbel, while the linear combination variants provide competitive results particularly on simpler datasets, with performance gains becoming more pronounced for the multiplicative form on complex benchmarks. The strength of the proposed approach lies in its ability to maintain more stable gradients in the positive domain, which improves feature extraction and generalization. Overall, our findings suggest that Swish‐Gumbel can serve as an effective alternative to existing activation functions for image classification, with potential to be applied in other machine learning tasks as well.
- 14 July 2026
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 38 (14)
Open AccessSwish‐Gumbel: A New Activation Function for Image Classification
In this study, we present Swish‐Gumbel, a new family of activation functions generated by combining the Swish and Gumbel functions through both multiplication and linear composition. We evaluate their performance on several image classification benchmarks, including MNIST, CIFAR‐10, Imagenette, and Beans, using different deep learning architectures. The results show that the multiplicative version consistently outperforms well‐known activation functions such as ReLU, Swish, Mish, and Gumbel, while the linear combination variants provide competitive results particularly on simpler datasets, with performance gains becoming more pronounced for the multiplicative form on complex benchmarks. The strength of the proposed approach lies in its ability to maintain more stable gradients in the positive domain, which improves feature extraction and generalization. Overall, our findings suggest that Swish‐Gumbel can serve as an effective alternative to existing activation functions for image classification, with potential to be applied in other machine learning tasks as well.
- 29 October 2025
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 37 (27-28)
A Systematic Review of Graph‐Based Representation Techniques for Cyber‐Attack Detection Across Application Domains
Cyber‐attack detection faces increasing challenges due to more advanced threats beyond traditional signature‐based methods. Machine learning and deep learning have become essential in enhancing network security. Graph representation techniques capture structural and relational patterns across various data sources such as network traffic, system logs, and malware behaviors. Graph‐based metrics and algorithms enable deeper analysis and more accurate classification of malicious activities. This study offers a systematic review of research articles published between 2019 and 2024 on graph‐based representation methods for detecting cyber‐attacks. Unlike previous reviews that classify studies by algorithm type or detection approach, this work groups them into four application domains: (i) general IDS/NIDS systems, (ii) botnet and DDoS detection, (iii) IoT‐related threats, and (iv) advanced threats including Android malware, APTs, and encrypted traffic. Each domain is analyzed using graph modeling techniques, feature extraction methods (manual vs. automated), learning algorithms, and detection effectiveness. The review also introduces a five‐stage framework outlining how graph‐based methods can be integrated into cyber‐attack detection: data acquisition, graph construction, feature extraction, detection/classification, and threat identification. This framework offers a structured view of methodological variety across different application areas. Results indicate that graph‐based approaches provide a powerful alternative to traditional techniques, enabling richer structural modeling in cybersecurity. Finally, the study presents a future research roadmap advocating for interpretable, graph‐driven solutions tailored to the evolving landscape of cyber threats.
- 29 October 2025
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 37 (27-28)
A Systematic Review of Graph‐Based Representation Techniques for Cyber‐Attack Detection Across Application Domains
Cyber‐attack detection faces increasing challenges due to more advanced threats beyond traditional signature‐based methods. Machine learning and deep learning have become essential in enhancing network security. Graph representation techniques capture structural and relational patterns across various data sources such as network traffic, system logs, and malware behaviors. Graph‐based metrics and algorithms enable deeper analysis and more accurate classification of malicious activities. This study offers a systematic review of research articles published between 2019 and 2024 on graph‐based representation methods for detecting cyber‐attacks. Unlike previous reviews that classify studies by algorithm type or detection approach, this work groups them into four application domains: (i) general IDS/NIDS systems, (ii) botnet and DDoS detection, (iii) IoT‐related threats, and (iv) advanced threats including Android malware, APTs, and encrypted traffic. Each domain is analyzed using graph modeling techniques, feature extraction methods (manual vs. automated), learning algorithms, and detection effectiveness. The review also introduces a five‐stage framework outlining how graph‐based methods can be integrated into cyber‐attack detection: data acquisition, graph construction, feature extraction, detection/classification, and threat identification. This framework offers a structured view of methodological variety across different application areas. Results indicate that graph‐based approaches provide a powerful alternative to traditional techniques, enabling richer structural modeling in cybersecurity. Finally, the study presents a future research roadmap advocating for interpretable, graph‐driven solutions tailored to the evolving landscape of cyber threats.
- Open Access
- 5 June 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (11) , 1441
Connectogram-COH: A Coherence-Based Time-Graph Representation for EEG-Based Alzheimer’s Disease Detection
Background: Alzheimer’s disease (AD) is a neurological disorder that affects the brain in the elderly, resulting in memory loss, mental deterioration, and loss of the ability to think and act, while being a cause of death, with its rates increasing dramatically. A popular method to detect AD is electroencephalography (EEG) signal analysis thanks to its ability to reflect neural activity, which helps to identify abnormalities associated with the disorder. Originating from its multivariate nature, EEG signals are generally handled as multidimensional time series, and the related methodology is employed. Methods: This study proposes a new transformation strategy that generates a graph representation with time resolution, which handles EEG recordings as relatively small time windows and converts these segments into a similarity graph based on signal coherence between available channels. The retrieved adjacency matrices are further flattened to form a 1-pixel image column, which represents the coherence activity from the available electrodes within the given time window. These pixel columns are concatenated horizontally for all available sliding time windows with 50% overlap, resulting in a grayscale image representation that can be input to well-known deep learning architectures specialized for images. We name this representation Connectogram-COH, a coherence-based version of the previously proposed time graph representation, Connectogram. Results: The experimental results demonstrate that the proposed Connectogram-COH representation effectively captures the coherence dynamics of multichannel EEG data and achieves high accuracy in detecting Alzheimer’s disease. The time graph images serve as robust input for deep learning classifiers, outperforming traditional EEG representations in terms of classification performance. Conclusions: Connectogram-COH offers a powerful and interpretable approach for transforming EEG signals into image representations that are well suited for deep learning. The method not only improves the detection of AD but also shows promise for broader applications in EEG-based and general time series classification tasks.
- 5 June 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (11) , 1441
Open AccessConnectogram-COH: A Coherence-Based Time-Graph Representation for EEG-Based Alzheimer’s Disease Detection
Background: Alzheimer’s disease (AD) is a neurological disorder that affects the brain in the elderly, resulting in memory loss, mental deterioration, and loss of the ability to think and act, while being a cause of death, with its rates increasing dramatically. A popular method to detect AD is electroencephalography (EEG) signal analysis thanks to its ability to reflect neural activity, which helps to identify abnormalities associated with the disorder. Originating from its multivariate nature, EEG signals are generally handled as multidimensional time series, and the related methodology is employed. Methods: This study proposes a new transformation strategy that generates a graph representation with time resolution, which handles EEG recordings as relatively small time windows and converts these segments into a similarity graph based on signal coherence between available channels. The retrieved adjacency matrices are further flattened to form a 1-pixel image column, which represents the coherence activity from the available electrodes within the given time window. These pixel columns are concatenated horizontally for all available sliding time windows with 50% overlap, resulting in a grayscale image representation that can be input to well-known deep learning architectures specialized for images. We name this representation Connectogram-COH, a coherence-based version of the previously proposed time graph representation, Connectogram. Results: The experimental results demonstrate that the proposed Connectogram-COH representation effectively captures the coherence dynamics of multichannel EEG data and achieves high accuracy in detecting Alzheimer’s disease. The time graph images serve as robust input for deep learning classifiers, outperforming traditional EEG representations in terms of classification performance. Conclusions: Connectogram-COH offers a powerful and interpretable approach for transforming EEG signals into image representations that are well suited for deep learning. The method not only improves the detection of AD but also shows promise for broader applications in EEG-based and general time series classification tasks.
- Open Access
- 1 December 2024
- Published by Walter de Gruyter GmbH in Applied Computer Systems
- Vol. 29 (2) , 30-42
Unveiling Trends of Chatbot and Conversational Agents: A Bibliometric Study
Recent years have seen remarkable growth and diversification in the study of chatbots and conversational agents. This research employs bibliometric and network -analytical methodology to thoroughly investigate the latest trends and themes in chatbot technology, a topic that has gained prominence in contemporary research discourse. The primary aim for this paper is to examine the evolution, prevailing trends, and provide an extensive overview of the chatbot field. Using the Web of Science core collection database, this study evaluates articles published from 1980 to 2024 by scanning over 7327 journal articles, ultimately focusing on 2622 key articles from prominent journals, institutions, and authors in the field. Key findings indicate a consistent increase in publication count related to chatbots recently. The study also identifies discrimination for critical areas such as advancements in artificial intelligence, machine learning, and natural language processing and underscores the importance of quantitatively assessing their impact and applications in diverse areas. Additionally, it sheds light on the collaboration among researchers, institutions, and nations in the development of this field. Furthermore, an analysis of written abstracts indicates a concentrated effort on enhancing user interactions and the technological progression of chatbots. The findings of this study provide insight into various sectors related to the development of chatbot technology in digital communication and AI advancement. Therefore, this bibliometric analysis offers a unique and in-depth view of the evolving chatbot research landscape, serving as a valuable guide for future research and strategic planning in this rapidly advancing area.
- 1 December 2024
- Published by Walter de Gruyter GmbH in Applied Computer Systems
- Vol. 29 (2) , 30-42
Open AccessUnveiling Trends of Chatbot and Conversational Agents: A Bibliometric Study
Recent years have seen remarkable growth and diversification in the study of chatbots and conversational agents. This research employs bibliometric and network -analytical methodology to thoroughly investigate the latest trends and themes in chatbot technology, a topic that has gained prominence in contemporary research discourse. The primary aim for this paper is to examine the evolution, prevailing trends, and provide an extensive overview of the chatbot field. Using the Web of Science core collection database, this study evaluates articles published from 1980 to 2024 by scanning over 7327 journal articles, ultimately focusing on 2622 key articles from prominent journals, institutions, and authors in the field. Key findings indicate a consistent increase in publication count related to chatbots recently. The study also identifies discrimination for critical areas such as advancements in artificial intelligence, machine learning, and natural language processing and underscores the importance of quantitatively assessing their impact and applications in diverse areas. Additionally, it sheds light on the collaboration among researchers, institutions, and nations in the development of this field. Furthermore, an analysis of written abstracts indicates a concentrated effort on enhancing user interactions and the technological progression of chatbots. The findings of this study provide insight into various sectors related to the development of chatbot technology in digital communication and AI advancement. Therefore, this bibliometric analysis offers a unique and in-depth view of the evolving chatbot research landscape, serving as a valuable guide for future research and strategic planning in this rapidly advancing area.
- Open Access
- 20 June 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (12) , 1313
Lung Disease Detection Using U-Net Feature Extractor Cascaded by Graph Convolutional Network
Computed tomography (CT) scans have recently emerged as a major technique for the fast diagnosis of lung diseases via image classification techniques. In this study, we propose a method for the diagnosis of COVID-19 disease with improved accuracy by utilizing graph convolutional networks (GCN) at various layer formations and kernel sizes to extract features from CT scan images. We apply a U-Net model to aid in segmentation and feature extraction. In contrast with previous research retrieving deep features from convolutional filters and pooling layers, which fail to fully consider the spatial connectivity of the nodes, we employ GCNs for classification and prediction to capture spatial connectivity patterns, which provides a significant association benefit. We handle the extracted deep features to form an adjacency matrix that contains a graph structure and pass it to a GCN along with the original image graph and the largest kernel graph. We combine these graphs to form one block of the graph input and then pass it through a GCN with an additional dropout layer to avoid overfitting. Our findings show that the suggested framework, called the feature-extracted graph convolutional network (FGCN), performs better in identifying lung diseases compared to recently proposed deep learning architectures that are not based on graph representations. The proposed model also outperforms a variety of transfer learning models commonly used for medical diagnosis tasks, highlighting the abstraction potential of the graph representation over traditional methods.
- 20 June 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (12) , 1313
Open AccessLung Disease Detection Using U-Net Feature Extractor Cascaded by Graph Convolutional Network
Computed tomography (CT) scans have recently emerged as a major technique for the fast diagnosis of lung diseases via image classification techniques. In this study, we propose a method for the diagnosis of COVID-19 disease with improved accuracy by utilizing graph convolutional networks (GCN) at various layer formations and kernel sizes to extract features from CT scan images. We apply a U-Net model to aid in segmentation and feature extraction. In contrast with previous research retrieving deep features from convolutional filters and pooling layers, which fail to fully consider the spatial connectivity of the nodes, we employ GCNs for classification and prediction to capture spatial connectivity patterns, which provides a significant association benefit. We handle the extracted deep features to form an adjacency matrix that contains a graph structure and pass it to a GCN along with the original image graph and the largest kernel graph. We combine these graphs to form one block of the graph input and then pass it through a GCN with an additional dropout layer to avoid overfitting. Our findings show that the suggested framework, called the feature-extracted graph convolutional network (FGCN), performs better in identifying lung diseases compared to recently proposed deep learning architectures that are not based on graph representations. The proposed model also outperforms a variety of transfer learning models commonly used for medical diagnosis tasks, highlighting the abstraction potential of the graph representation over traditional methods.
- Open Access
Association rule mining with earthquake data collected from Turkiye region
Earthquakes are evaluated among the most destructive disasters for human beings, as also experienced for Turkiye region. Data science has the property of discovering hidden patterns in case a sufficient volume of data is supplied. Time dependency of events, specifically being defined by co-occurrence in a specific time window, may be handled as an associate rule mining task such as a market-basket analysis application. In this regard, we assumed each day's seismic activity as a single basket of events, leading to discovering the association patterns between these events. Consequently, this study presents the most prominent association rules for the earthquakes recorded in Turkiye region in the last 5 years, each year presented separately. Results indicate statistical inference with events recorded from regions of various distances, which could be further verified with geologic evidence from the field. As a result, we believe that the current study may form a statistical basis for the future works with the aid of machine learning algorithm performed for associate rule mining.
Open AccessAssociation rule mining with earthquake data collected from Turkiye region
Earthquakes are evaluated among the most destructive disasters for human beings, as also experienced for Turkiye region. Data science has the property of discovering hidden patterns in case a sufficient volume of data is supplied. Time dependency of events, specifically being defined by co-occurrence in a specific time window, may be handled as an associate rule mining task such as a market-basket analysis application. In this regard, we assumed each day's seismic activity as a single basket of events, leading to discovering the association patterns between these events. Consequently, this study presents the most prominent association rules for the earthquakes recorded in Turkiye region in the last 5 years, each year presented separately. Results indicate statistical inference with events recorded from regions of various distances, which could be further verified with geologic evidence from the field. As a result, we believe that the current study may form a statistical basis for the future works with the aid of machine learning algorithm performed for associate rule mining.
- Open Access
- 28 December 2022
- Published by Journal of Information and Communication Technologies in Bilgi ve İletişim Teknolojileri Dergisi
- Vol. 4 (2) , 128-142
VarioGram – Zaman serileri için renkli bir zaman-graf temsili
In this study, a framework for network-based representation of time series is presented. In the proposed method, initially, a segmentation procedure is completed by dividing the signals in the time domain into fixed-width time windows with 50% overlap. Each segment is normalized based on the range defined by the absolute maximum amplitude value of the main signal and its negative counterpart, and the normalized signals are quantized to 2^n levels. This transformation, proceeding through 3 channels expressed by 3 different jump values, generates a vertical RGB image representation by combining the channels in layers. As a result of tiling these vertical RGB images from each time window horizontally, a time-graph representation called VarioGram is obtained, where the horizontal axis represents time, and the vertical axis represents signal fluctuations. Feeding a ResNet model with VarioGram representations obtained by the transformation of the audio signals in the ESC-10 dataset which is frequently used in environmental sound classification problems, a classification success of 82.08% has been obtained, while this success has been 93.33% with the VarioGram representations hybridized with mel-spectrogram images. The VarioGram representations therefore acted to slightly improve the highest classification success achievable with the mel-spectrogram alone.
- 28 December 2022
- Published by Journal of Information and Communication Technologies in Bilgi ve İletişim Teknolojileri Dergisi
- Vol. 4 (2) , 128-142
Open AccessVarioGram – Zaman serileri için renkli bir zaman-graf temsili
In this study, a framework for network-based representation of time series is presented. In the proposed method, initially, a segmentation procedure is completed by dividing the signals in the time domain into fixed-width time windows with 50% overlap. Each segment is normalized based on the range defined by the absolute maximum amplitude value of the main signal and its negative counterpart, and the normalized signals are quantized to 2^n levels. This transformation, proceeding through 3 channels expressed by 3 different jump values, generates a vertical RGB image representation by combining the channels in layers. As a result of tiling these vertical RGB images from each time window horizontally, a time-graph representation called VarioGram is obtained, where the horizontal axis represents time, and the vertical axis represents signal fluctuations. Feeding a ResNet model with VarioGram representations obtained by the transformation of the audio signals in the ESC-10 dataset which is frequently used in environmental sound classification problems, a classification success of 82.08% has been obtained, while this success has been 93.33% with the VarioGram representations hybridized with mel-spectrogram images. The VarioGram representations therefore acted to slightly improve the highest classification success achievable with the mel-spectrogram alone.
- 1 February 2021
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 64
Arithmetic success and gender-based characterization of brain connectivity across EEG bands
- 1 February 2021
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 64
Arithmetic success and gender-based characterization of brain connectivity across EEG bands
- Open Access
- 24 June 2016
- Published by Springer Nature in SpringerPlus
- Vol. 5 (1) , 1-18
Uncovering the differences in linguistic network dynamics of book and social media texts
Complex network studies span a large variety of applications including linguistic networks. To investigate the differences in book and social media texts in terms of linguistic typology, we constructed both sequential and sentence collocation networks of book, Facebook and Twitter texts with undirected and weighted edges. The comparisons are performed using the basic parameters like average degree, modularity, average clustering coefficient, average path length, diameter, average link weight etc. We also presented the distribution graphs for node degrees, edge weights and maximum degree differences of the pairing nodes. The degree difference occurrences are furtherly detailed with the grayscale percentile plots with respect to the edge weights. We linked the network analysis with linguistic aspects like word and sentence length distributions. We concluded that linguistic typology demonstrates a formal usage in book that slightly deviates to informal in Twitter. Facebook interpolates between these media by the means of network parameters, while the informality of Twitter is mostly influenced by the character limitations.
- 24 June 2016
- Published by Springer Nature in SpringerPlus
- Vol. 5 (1) , 1-18
Open AccessUncovering the differences in linguistic network dynamics of book and social media texts
Complex network studies span a large variety of applications including linguistic networks. To investigate the differences in book and social media texts in terms of linguistic typology, we constructed both sequential and sentence collocation networks of book, Facebook and Twitter texts with undirected and weighted edges. The comparisons are performed using the basic parameters like average degree, modularity, average clustering coefficient, average path length, diameter, average link weight etc. We also presented the distribution graphs for node degrees, edge weights and maximum degree differences of the pairing nodes. The degree difference occurrences are furtherly detailed with the grayscale percentile plots with respect to the edge weights. We linked the network analysis with linguistic aspects like word and sentence length distributions. We concluded that linguistic typology demonstrates a formal usage in book that slightly deviates to informal in Twitter. Facebook interpolates between these media by the means of network parameters, while the informality of Twitter is mostly influenced by the character limitations.