Ayca Kumluca Topalli
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ayca Kumluca Topalli has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Timur Ozkul
Izmir Ekonomi Universitesi
2 shared publications - Ayça Dernek
Zmir Ekonomi niversitesi,Elektrik ve Elektronik Mh. Bl.,zmir,Trkiye
1 shared publication - Bahar Kaya
Elektrik ve Elektronik Mh. Bl., zmir Ekonomi niversitesi,stanbul,Trkiye
1 shared publication - Berra Beşer
Elektrik ve Elektronik Mh. Bl., zmir Ekonomi niversitesi,stanbul,Trkiye
1 shared publication - Ceren Özgür
Zmir Ekonomi niversitesi,Bilgisayar Mh. Bl.,zmir,Trkiye
1 shared publication - Doğukan Özyurt1 shared publication
Latest Publications
- 14 October 2026
- Published by Journal of the Faculty of Engineering and Architecture of Gazi University in Gazi Üniversitesi Mühendislik-Mimarlık Fakültesi Dergisi
- No. Advanced O
Türkçede edebî parmak izleri: Dğal dil işleme ve makine öğrenmesi kullanarak kitap türü bulma ve yazar eşleştirme
Edebi içerik temelli yazar eşleştirme, tür sınıflandırması ve öneri sistemleri doğal dil işlemede giderek daha fazla önem kazanmaktadır. Bu çalışma, kitap metinlerini vektör uzayında temsil ederek hem sınıflandırma hem de içerik tabanlı öneri yapabilen bir sistem sunmaktadır. Türkçe romanlardan oluşan kapsamlı bir veri kümesi derlenmiş, ön işlemden geçirilmiş ve Doc2Vec algoritması kullanılarak yüksek boyutlu vektörlere dönüştürülmüştür. Genellenebilirliği sağlamak amacıyla, her bir yazarın test kümesinde yer almasını garanti eden özel bir eğitim/test bölme stratejisi uygulanmıştır. Elde edilen belge vektörleri, Doğrusal Destek Vektör Makinesi Sınıflandırıcısı, Lojistik Regresyon, Rastgele Orman, Gauss Naive Bayes ve En Yakın Komşular gibi sınıflandırıcılarla yazar ve tür tespitinde kullanılmıştır. Deneysel sonuçlar, ilk iki yöntemin hem tür hem de yazar tahmininde diğer üç yönteme göre daha yüksek performans gösterdiğini ortaya koymuştur. Bu çalışma ayrıca, geleneksel kosinüs benzerliğinin ötesine geçen özgün bir sınıflandırma yaklaşımı önermektedir. Geliştirilen yöntem, okuyucunun daha önce beğendiği eserlerle benzer anlatım tarzına sahip kitapları önererek daha az bilinen yazarların görünürlüğünün artırılmasına katkı sağlamaktadır. Araştırma, belge gömme tabanlı yöntemlerin dijital beşerî bilimler ve öneri sistemlerinde etkili bir biçimde uygulanabileceğini göstermektedir.
- 14 October 2026
- Published by Journal of the Faculty of Engineering and Architecture of Gazi University in Gazi Üniversitesi Mühendislik-Mimarlık Fakültesi Dergisi
- No. Advanced O
Türkçede edebî parmak izleri: Dğal dil işleme ve makine öğrenmesi kullanarak kitap türü bulma ve yazar eşleştirme
Edebi içerik temelli yazar eşleştirme, tür sınıflandırması ve öneri sistemleri doğal dil işlemede giderek daha fazla önem kazanmaktadır. Bu çalışma, kitap metinlerini vektör uzayında temsil ederek hem sınıflandırma hem de içerik tabanlı öneri yapabilen bir sistem sunmaktadır. Türkçe romanlardan oluşan kapsamlı bir veri kümesi derlenmiş, ön işlemden geçirilmiş ve Doc2Vec algoritması kullanılarak yüksek boyutlu vektörlere dönüştürülmüştür. Genellenebilirliği sağlamak amacıyla, her bir yazarın test kümesinde yer almasını garanti eden özel bir eğitim/test bölme stratejisi uygulanmıştır. Elde edilen belge vektörleri, Doğrusal Destek Vektör Makinesi Sınıflandırıcısı, Lojistik Regresyon, Rastgele Orman, Gauss Naive Bayes ve En Yakın Komşular gibi sınıflandırıcılarla yazar ve tür tespitinde kullanılmıştır. Deneysel sonuçlar, ilk iki yöntemin hem tür hem de yazar tahmininde diğer üç yönteme göre daha yüksek performans gösterdiğini ortaya koymuştur. Bu çalışma ayrıca, geleneksel kosinüs benzerliğinin ötesine geçen özgün bir sınıflandırma yaklaşımı önermektedir. Geliştirilen yöntem, okuyucunun daha önce beğendiği eserlerle benzer anlatım tarzına sahip kitapları önererek daha az bilinen yazarların görünürlüğünün artırılmasına katkı sağlamaktadır. Araştırma, belge gömme tabanlı yöntemlerin dijital beşerî bilimler ve öneri sistemlerinde etkili bir biçimde uygulanabileceğini göstermektedir.
- 1 August 2026
- Published by American Society of Civil Engineers (ASCE) in Natural Hazards Review
- Vol. 27 (3)
Determining Impact Scores of Tweets in Turkish at Disaster Times
- 1 August 2026
- Published by American Society of Civil Engineers (ASCE) in Natural Hazards Review
- Vol. 27 (3)
Determining Impact Scores of Tweets in Turkish at Disaster Times
- 10 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Retrieval Augmented Generation Based System to Analyze Turkish News Texts According to Sustainable Development Goals Using Large Language Models (RAGaze)
Due to its geographical location and social structure, Türkiye is a country where events that deeply affect many people happen every day. For example, climate changes are now clearly observed in our country where four seasons are experienced; this situation affects agriculture and the economy, while at the same time, since Türkiye has borders with many countries due to its geographic location, it makes our country a bridge country that contributes to peace. However, the recent increase in the number of women murders shows how essential it is for the press to cover important issues facing the society and thus raise public awareness about these issues. For this reason, a system that aims to identify the extent to which the written press in Türkiye pays attention to social sensitivities has been made. The developed system is a chatbot application that works on a user interface. It analyzes columns published in various newspapers based on user questions within the framework of Sustainable Development Goals (SDG) such as equality, justice, economy, climate change and gender equality and responds accordingly. It uses a large language model (LLM) infrastructure for this purpose. The model classifies the extent to which these issues are covered in Turkish newspapers using Retrieval Augmented Generation (RAG) and vector embeddings. In addition, two different embedding models, text-embedding-004 and gemini-embedding-exp-03-07, were used and it was determined which one was better for Turkish texts.As a result of this project, a successful chatbot application was developed to analyze press content related to SDG. The chatbot application RAGaze provides accurate and meaningful answers to user queries. In addition, the answers given to the questions always remain within the context of SDG and RAGaze does not act outside this context.
- 10 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Retrieval Augmented Generation Based System to Analyze Turkish News Texts According to Sustainable Development Goals Using Large Language Models (RAGaze)
Due to its geographical location and social structure, Türkiye is a country where events that deeply affect many people happen every day. For example, climate changes are now clearly observed in our country where four seasons are experienced; this situation affects agriculture and the economy, while at the same time, since Türkiye has borders with many countries due to its geographic location, it makes our country a bridge country that contributes to peace. However, the recent increase in the number of women murders shows how essential it is for the press to cover important issues facing the society and thus raise public awareness about these issues. For this reason, a system that aims to identify the extent to which the written press in Türkiye pays attention to social sensitivities has been made. The developed system is a chatbot application that works on a user interface. It analyzes columns published in various newspapers based on user questions within the framework of Sustainable Development Goals (SDG) such as equality, justice, economy, climate change and gender equality and responds accordingly. It uses a large language model (LLM) infrastructure for this purpose. The model classifies the extent to which these issues are covered in Turkish newspapers using Retrieval Augmented Generation (RAG) and vector embeddings. In addition, two different embedding models, text-embedding-004 and gemini-embedding-exp-03-07, were used and it was determined which one was better for Turkish texts.As a result of this project, a successful chatbot application was developed to analyze press content related to SDG. The chatbot application RAGaze provides accurate and meaningful answers to user queries. In addition, the answers given to the questions always remain within the context of SDG and RAGaze does not act outside this context.
- Open Access
- 25 June 2025
- Published by Marmara University in International Journal of Advances in Engineering and Pure Sciences
- Vol. 37 (2) , 196-205
EkoBot: Türkçe Destekli Akıllı Sanal Akademik Danışman
Bu çalışmada üniversite öğrencilerine sanal danışman olarak yardımcı olabilecek, Türkçe destekli bir akıllı yazılım, EkoBot sunulmuştur. Bu yazılımla öğrencilerin sorularına doğru ve hızlı bir şekilde yanıt almaları hedeflenmiştir. Bunun için yapay zekâ destekli büyük dil modellerinden yararlanılmıştır. Büyük dil modelinin, eğitimi sırasında kullanılmamış, üniversite yönetmeliklerine dayalı yanıtlar verebilmesi için bu belgeler modele dışardan verilmiş ve “almayla artırılmış üretim” yöntemi kullanılmıştır. Önerilen sistemin performansını ölçmek için ucu açık ya da olumlu ve olumsuz yanıtlara sahip 100 adet soru üretilmiştir. Alma kısmında, soruya en çok benzeyen beş bağlam metni ile %100 başarım elde edilmiştir. Üretme kısmında, yanıt ile bağlam benzerlikleri 0,82 olarak bulunmuştur. Ayrıca önerilen çözümün bir Web sayfası olarak çalışan bir prototipi hazırlanmış ve öğrencilerin kullanımına sunulmuştur.
- 25 June 2025
- Published by Marmara University in International Journal of Advances in Engineering and Pure Sciences
- Vol. 37 (2) , 196-205
Open AccessEkoBot: Türkçe Destekli Akıllı Sanal Akademik Danışman
Bu çalışmada üniversite öğrencilerine sanal danışman olarak yardımcı olabilecek, Türkçe destekli bir akıllı yazılım, EkoBot sunulmuştur. Bu yazılımla öğrencilerin sorularına doğru ve hızlı bir şekilde yanıt almaları hedeflenmiştir. Bunun için yapay zekâ destekli büyük dil modellerinden yararlanılmıştır. Büyük dil modelinin, eğitimi sırasında kullanılmamış, üniversite yönetmeliklerine dayalı yanıtlar verebilmesi için bu belgeler modele dışardan verilmiş ve “almayla artırılmış üretim” yöntemi kullanılmıştır. Önerilen sistemin performansını ölçmek için ucu açık ya da olumlu ve olumsuz yanıtlara sahip 100 adet soru üretilmiştir. Alma kısmında, soruya en çok benzeyen beş bağlam metni ile %100 başarım elde edilmiştir. Üretme kısmında, yanıt ile bağlam benzerlikleri 0,82 olarak bulunmuştur. Ayrıca önerilen çözümün bir Web sayfası olarak çalışan bir prototipi hazırlanmış ve öğrencilerin kullanımına sunulmuştur.
- 1 December 2024
- Published by Springer Nature in Journal of Intelligent Manufacturing
- Vol. 36 (8) , 5967-5980
Discovering Granger causality with convolutional neural networks
- 1 December 2024
- Published by Springer Nature in Journal of Intelligent Manufacturing
- Vol. 36 (8) , 5967-5980
Discovering Granger causality with convolutional neural networks
- 17 November 2024
- Published by Omer Halisdemir Universitesi in Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi
Comparison of Machine Learning Methods for Limited Predictive Maintenance
Predictive maintenance has gained increasing attention recently with the availability of sensors and connectivity of equipment. Yet, it would be difficult to obtain a wide range of data, especially with legacy devices. This paper describes an intelligent method for predicting a near future condition using the past information for an environment in which data are limited to the alarm logs from industrial machinery. Since machine learning methods are proven to be efficient in classification tasks using time series data, three of them are selected to predict an alarm two hours in advance using the past occurrences. These methods are neural networks, random forests, and extreme gradient boosting. The performances of these three methods are compared, and it is aimed to find the optimal configuration among hyper-parameter values. According to the obtained results, extreme gradient boosting gives the highest F1-score of 0.767 with number of trees equal to 500, maximum depth of 128, and an input window of alarm occurrences from the last day. This work consists of a comparative study aiming to identify the best machine learning method for alarm predictions, which potentially provides important insights into the operation and maintenance of machinery, bringing the possibility of considerable cost reductions.
- 17 November 2024
- Published by Omer Halisdemir Universitesi in Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi
Comparison of Machine Learning Methods for Limited Predictive Maintenance
Predictive maintenance has gained increasing attention recently with the availability of sensors and connectivity of equipment. Yet, it would be difficult to obtain a wide range of data, especially with legacy devices. This paper describes an intelligent method for predicting a near future condition using the past information for an environment in which data are limited to the alarm logs from industrial machinery. Since machine learning methods are proven to be efficient in classification tasks using time series data, three of them are selected to predict an alarm two hours in advance using the past occurrences. These methods are neural networks, random forests, and extreme gradient boosting. The performances of these three methods are compared, and it is aimed to find the optimal configuration among hyper-parameter values. According to the obtained results, extreme gradient boosting gives the highest F1-score of 0.767 with number of trees equal to 500, maximum depth of 128, and an input window of alarm occurrences from the last day. This work consists of a comparative study aiming to identify the best machine learning method for alarm predictions, which potentially provides important insights into the operation and maintenance of machinery, bringing the possibility of considerable cost reductions.
- Open Access
- 23 October 2024
- Published by Duzce Universitesi Bilim ve Teknoloji Dergisi in Düzce Üniversitesi Bilim ve Teknoloji Dergisi
- Vol. 12 (4) , 2016-2032
VOTEMAT: A Blockchain Based Voting System
This study aims to show that a secure, trustable and immutable voting system can be established with Blockchain technology. Decentralized structure of the Blockchain excludes the central authority and provides transparency. Moreover, its cryptographic functions enable secure transactions. Therefore, the operation is prevented from potential frauds, such as multiple votes, fake vote attempts, and fraudulent vote counts. The proposed method, VOTEMAT, covers both electronic voting and paper ballot as a complete solution. A mobile application and a Website, connected to Ethereum private Blockchain network, were developed for the voters who prefer to cast their votes remotely. It is also possible to vote in the voting centres via the mobile device or paper ballot placed in the vote boxes; but these votes are also recorded in the same Blockchain and equally secure. For the remote users, a two-step authentication is designed, based on the information on the national identity card and face recognition. An additional encryption based security measure is used to avoid hacking attempts, such as man in the middle attacks. Since the proposed system is more practical than the traditional voting methods, it can increase the participation and be utilized in all kinds of local or national elections.
- 23 October 2024
- Published by Duzce Universitesi Bilim ve Teknoloji Dergisi in Düzce Üniversitesi Bilim ve Teknoloji Dergisi
- Vol. 12 (4) , 2016-2032
Open AccessVOTEMAT: A Blockchain Based Voting System
This study aims to show that a secure, trustable and immutable voting system can be established with Blockchain technology. Decentralized structure of the Blockchain excludes the central authority and provides transparency. Moreover, its cryptographic functions enable secure transactions. Therefore, the operation is prevented from potential frauds, such as multiple votes, fake vote attempts, and fraudulent vote counts. The proposed method, VOTEMAT, covers both electronic voting and paper ballot as a complete solution. A mobile application and a Website, connected to Ethereum private Blockchain network, were developed for the voters who prefer to cast their votes remotely. It is also possible to vote in the voting centres via the mobile device or paper ballot placed in the vote boxes; but these votes are also recorded in the same Blockchain and equally secure. For the remote users, a two-step authentication is designed, based on the information on the national identity card and face recognition. An additional encryption based security measure is used to avoid hacking attempts, such as man in the middle attacks. Since the proposed system is more practical than the traditional voting methods, it can increase the participation and be utilized in all kinds of local or national elections.
- 16 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Türkçe Yazılmıs Tweet İletilerinin Afetle İlgili İki Asamalı Sınıflandırılmasında Büyük Dil Modellerinin Performans Karşılaştırması Performance Comparison of Large Language Models in Disaster Related Two-Stage Classification of Tweets Written in Turkish
Natural disasters are very frequent in Turkiye, therefore it is quite vital to tackle the problems aroused after these disasters. This study proposes a system to reduce the losses caused by the natural disasters and provides a comparison method for the efficient selection of the system components. A database is formed from the tweet samples posted in the aftermath of the previous natural disasters and these tweets are classified in two stages using prompt engineering and large language models. In the first stage, the classification is done based on disaster type such as “earthquake”, “fire” or “flood”, then the tweets in these disaster types are classified for needs such as “search and rescue”, “equipment and food” in the second stage. In order to find the best model for aforementioned classifications, ChatGPT-3.5, fine-tuned ChatGPT-3.5 and ChatGPT-4 are selected and tested. Fine-tuned ChatGPT-3.5 with enhanced prompting is found to have the highest performance with 98.4% average success score for disaster classification. The success rate of the fine-tuned model for classification of needs is calculated as 95.6% in average. This study is expected not only to contribute to the Turkish language processing research area but also to support rescue organisations as well.
- 16 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Türkçe Yazılmıs Tweet İletilerinin Afetle İlgili İki Asamalı Sınıflandırılmasında Büyük Dil Modellerinin Performans Karşılaştırması Performance Comparison of Large Language Models in Disaster Related Two-Stage Classification of Tweets Written in Turkish
Natural disasters are very frequent in Turkiye, therefore it is quite vital to tackle the problems aroused after these disasters. This study proposes a system to reduce the losses caused by the natural disasters and provides a comparison method for the efficient selection of the system components. A database is formed from the tweet samples posted in the aftermath of the previous natural disasters and these tweets are classified in two stages using prompt engineering and large language models. In the first stage, the classification is done based on disaster type such as “earthquake”, “fire” or “flood”, then the tweets in these disaster types are classified for needs such as “search and rescue”, “equipment and food” in the second stage. In order to find the best model for aforementioned classifications, ChatGPT-3.5, fine-tuned ChatGPT-3.5 and ChatGPT-4 are selected and tested. Fine-tuned ChatGPT-3.5 with enhanced prompting is found to have the highest performance with 98.4% average success score for disaster classification. The success rate of the fine-tuned model for classification of needs is calculated as 95.6% in average. This study is expected not only to contribute to the Turkish language processing research area but also to support rescue organisations as well.
- 23 May 2024
- Published by Taylor & Francis in Inorganic and Nano-Metal Chemistry
- Vol. 55 (7) , 775-783
Green synthesized silver nanoparticles in two stages: Box Behnken Design to machine learning
- 23 May 2024
- Published by Taylor & Francis in Inorganic and Nano-Metal Chemistry
- Vol. 55 (7) , 775-783
Green synthesized silver nanoparticles in two stages: Box Behnken Design to machine learning