Gülfem Işıklar Alptekin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Gülfem Işıklar Alptekin has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Günce Keziban Orman
Galatasaray University
5 shared publications - Afra Arslan
ILab
3 shared publications - Ozlem Durmaz Incel
University of Twente, Enschede, Netherlands
3 shared publications - Arda Yücel
ILab
2 shared publications - Asli Sari
Istanbul Technical University, Istanbul, Turkiye
2 shared publications Ayse Basar Bener
Ryerson University, School of Information Technology Management, Toronto, Canada
2 shared publications
Latest Publications
- Open Access
- 8 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 166-167
Automated Annotation of Test Types using LLMs: A Multi-LLM Workflow for Machine Learning Test-Related Commits
Test-type annotation classifies commits based on the type of software test introduced or modified. While there are established taxonomies for testing machine learning (ML) software systems, implementing these manually at the commit level can be costly. This makes large language models (LLMs) a promising alternative. Utilizing an eight-category test-type taxonomy, we evaluate 97 manually labeled commits from scikit-learn using GPT-4o, Claude Sonnet, and Qwen 2.5-7B-Instruct-Turbo LLMs through four prompting strategies. This analysis focused on their classification performance, stability, agreement with experts, top-2 candidate recovery, and multi-LLM voting workflows (S1-S4) along with a corresponding analysis on Keras. Notably, GPT-4o in a zero-shot setting demonstrated the highest stability, and permitting a second label led to a reduction in single-prediction errors (e.g., Claude CoT-ZS Hit@1 = [59.79]%, Hit@2 = [77.32]%). Our most effective voting strategy (S1) surpassed the solo baseline, with its superior macro-F1 score (S1 > S2 > S3 > S4) also holding for Keras.
- 8 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 166-167
Open AccessAutomated Annotation of Test Types using LLMs: A Multi-LLM Workflow for Machine Learning Test-Related Commits
Test-type annotation classifies commits based on the type of software test introduced or modified. While there are established taxonomies for testing machine learning (ML) software systems, implementing these manually at the commit level can be costly. This makes large language models (LLMs) a promising alternative. Utilizing an eight-category test-type taxonomy, we evaluate 97 manually labeled commits from scikit-learn using GPT-4o, Claude Sonnet, and Qwen 2.5-7B-Instruct-Turbo LLMs through four prompting strategies. This analysis focused on their classification performance, stability, agreement with experts, top-2 candidate recovery, and multi-LLM voting workflows (S1-S4) along with a corresponding analysis on Keras. Notably, GPT-4o in a zero-shot setting demonstrated the highest stability, and permitting a second label led to a reduction in single-prediction errors (e.g., Claude CoT-ZS Hit@1 = [59.79]%, Hit@2 = [77.32]%). Our most effective voting strategy (S1) surpassed the solo baseline, with its superior macro-F1 score (S1 > S2 > S3 > S4) also holding for Keras.
- Open Access
- 19 June 2026
- Published by Sakarya University Journal of Computer and Information Sciences in Sakarya University Journal of Computer and Information Sciences
- No. Advanced O,p. 690-699
Feature Extraction for Real Estate Images and Titles with LLMs
Images and titles often contain rich latent information about their associated objects, particularly on web-based platforms. Real estate websites provide a clear example, where listing images and titles provide important details that assist users in their decision-making. However, these unstructured elements cannot be directly utilized in downstream machine learning tasks, since their contextual meaning is not directly interpretable. This work aims to transform listing images and titles into structured, tabular representations, making them suitable for analytical and predictive modeling. To this end, we propose a modular framework based on state-of-the-art large language models. The framework incorporates ReAct, LLM-as-a-Judge, and few-shot prompting techniques. Its performance is evaluated on a real-world real estate dataset and compared with BERT and CLIP-based baselines. Experimental results demonstrate that our framework achieves up to a 44.26% improvement in recall for listing attributes, such as the presence of a balcony or the furnishing status of a property.
- 19 June 2026
- Published by Sakarya University Journal of Computer and Information Sciences in Sakarya University Journal of Computer and Information Sciences
- No. Advanced O,p. 690-699
Open AccessFeature Extraction for Real Estate Images and Titles with LLMs
Images and titles often contain rich latent information about their associated objects, particularly on web-based platforms. Real estate websites provide a clear example, where listing images and titles provide important details that assist users in their decision-making. However, these unstructured elements cannot be directly utilized in downstream machine learning tasks, since their contextual meaning is not directly interpretable. This work aims to transform listing images and titles into structured, tabular representations, making them suitable for analytical and predictive modeling. To this end, we propose a modular framework based on state-of-the-art large language models. The framework incorporates ReAct, LLM-as-a-Judge, and few-shot prompting techniques. Its performance is evaluated on a real-world real estate dataset and compared with BERT and CLIP-based baselines. Experimental results demonstrate that our framework achieves up to a 44.26% improvement in recall for listing attributes, such as the presence of a balcony or the furnishing status of a property.
- 22 August 2025
- Published by Springer Nature
- p. 66-83
Modeling the Relationship Between Traffic Intensity and Urban Air Pollution with LSTM Networks
- 22 August 2025
- Published by Springer Nature
- p. 66-83
Modeling the Relationship Between Traffic Intensity and Urban Air Pollution with LSTM Networks
- Open Access
- 30 June 2025
- Published by Dicle Universitesi Muhendislik Fakultesi Muhendislik Dergisi in Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi
- Vol. 16 (2) , 273-285
Online Gıda Perakende Sektöründe Talep Tahmin Analizi
Gıda talebinin tahmin edilmesi, özellikle e-ticaret alanında iş süreçlerinin optimizasyonu ve sürdürülebilir kalkınma için büyük önem taşımaktadır. Bu çalışmada, öğrenme tabanlı algoritmalar ve zaman serisi analizlerini kullanarak geliştirilen bir talep tahmin sistemi sunulmuştur. Bu sistem, müşteri taleplerini daha doğru bir şekilde öngörmeyi ve perakende şirketlerine stratejik öneriler sunmayı amaçlamaktadır. Yapılan sayısal uygulamalar, Türkiye'deki bir gıda perakende şirketinin son yedi yıllık verileri üzerinde gerçekleştirilmiştir. Çalışmada, özellikle izolasyon ormanı, aglomeratif ve spektral kümeleme, Uzun Kısa Süreli Bellek ve Ekstrem Gradyan Arttırma gibi yöntemler kullanılmıştır. Geliştirilen modeller, satış verilerinin doğru tahmin edilmesine ve stok yönetimi gibi kritik iş süreçlerine önemli katkılar sağlamaktadır. Sonuçlar, özellikle pandemi sürecinde ve mevsimsel etkilerin dikkate alındığı durumlarda gıda talebinin daha etkin bir şekilde yönetilmesine olanak tanımaktadır.
- 30 June 2025
- Published by Dicle Universitesi Muhendislik Fakultesi Muhendislik Dergisi in Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi
- Vol. 16 (2) , 273-285
Open AccessOnline Gıda Perakende Sektöründe Talep Tahmin Analizi
Gıda talebinin tahmin edilmesi, özellikle e-ticaret alanında iş süreçlerinin optimizasyonu ve sürdürülebilir kalkınma için büyük önem taşımaktadır. Bu çalışmada, öğrenme tabanlı algoritmalar ve zaman serisi analizlerini kullanarak geliştirilen bir talep tahmin sistemi sunulmuştur. Bu sistem, müşteri taleplerini daha doğru bir şekilde öngörmeyi ve perakende şirketlerine stratejik öneriler sunmayı amaçlamaktadır. Yapılan sayısal uygulamalar, Türkiye'deki bir gıda perakende şirketinin son yedi yıllık verileri üzerinde gerçekleştirilmiştir. Çalışmada, özellikle izolasyon ormanı, aglomeratif ve spektral kümeleme, Uzun Kısa Süreli Bellek ve Ekstrem Gradyan Arttırma gibi yöntemler kullanılmıştır. Geliştirilen modeller, satış verilerinin doğru tahmin edilmesine ve stok yönetimi gibi kritik iş süreçlerine önemli katkılar sağlamaktadır. Sonuçlar, özellikle pandemi sürecinde ve mevsimsel etkilerin dikkate alındığı durumlarda gıda talebinin daha etkin bir şekilde yönetilmesine olanak tanımaktadır.
- 28 June 2025
- Published by Springer Nature
- p. 274-285
Analyzing the Impact of Visual and Listing Features in Real Estate Listings
- 28 June 2025
- Published by Springer Nature
- p. 274-285
Analyzing the Impact of Visual and Listing Features in Real Estate Listings
- Open Access
- 30 May 2025
- Published by Bilecik Seyh Edebali Universitesi Fen Bilimleri Dergisi in Bilecik Şeyh Edebali Üniversitesi Fen Bilimleri Dergisi
- Vol. 12 (1) , 233-241
Enhancing Hotel Recommendations through Feature- based Clustering
This paper addresses the challenge of sparse interaction data in recommendation systems for the hotel industry. Due to the infrequent nature of hotel stays (often once or a few times annually), customer-product interaction data is typically sparse, hindering the effectiveness of traditional collaborative filtering techniques. We propose a novel hybrid recommendation framework specifically designed for this scenario. Unlike conventional systems that rely solely on user preference similarity, our framework leverages hotel clustering based on binary attributes to segment the product space. User interactions are analyzed within these clusters, leading to a more refined recommendation process. We take advantage of several clustering and feature reduction techniques and assign the final recommendation through ballot scoring. The experiments are performed on a real-world hotel sales data set including both sales information and hotel attributes. We evaluate our methodology and demonstrate significant improvements over baseline approaches which is the case of not using the found clusters for recommendation. The proposed framework achieves a two-fold increase in both the number of users receiving recommendations and the number of correct recommendations. These results highlight the potential of cluster- based recommendations for mitigating sparsity issues in tourism recommender systems.
- 30 May 2025
- Published by Bilecik Seyh Edebali Universitesi Fen Bilimleri Dergisi in Bilecik Şeyh Edebali Üniversitesi Fen Bilimleri Dergisi
- Vol. 12 (1) , 233-241
Open AccessEnhancing Hotel Recommendations through Feature- based Clustering
This paper addresses the challenge of sparse interaction data in recommendation systems for the hotel industry. Due to the infrequent nature of hotel stays (often once or a few times annually), customer-product interaction data is typically sparse, hindering the effectiveness of traditional collaborative filtering techniques. We propose a novel hybrid recommendation framework specifically designed for this scenario. Unlike conventional systems that rely solely on user preference similarity, our framework leverages hotel clustering based on binary attributes to segment the product space. User interactions are analyzed within these clusters, leading to a more refined recommendation process. We take advantage of several clustering and feature reduction techniques and assign the final recommendation through ballot scoring. The experiments are performed on a real-world hotel sales data set including both sales information and hotel attributes. We evaluate our methodology and demonstrate significant improvements over baseline approaches which is the case of not using the found clusters for recommendation. The proposed framework achieves a two-fold increase in both the number of users receiving recommendations and the number of correct recommendations. These results highlight the potential of cluster- based recommendations for mitigating sparsity issues in tourism recommender systems.
- Open Access
- 30 January 2025
- Published by Erciyes Universitesi (DergiPark) in Erciyes İletişim Dergisi
- Vol. 12 (1) , 53-66
The Short-Term Effect of TV Advertisements on Digital Traffic
This study investigates the short-term effects of television advertisements (TV ads) on digital platform engagement across companies from diverse sectors. Utilizing statistical analysis methods including t-test, ANOVA, and Tukey's HSD test, we examined the impact of factors such as ad timing, program type, and sector on web traffic. Our findings reveal that prime time advertisements significantly influence web traffic, with effects varying across different sectors and program types. The study demonstrates the importance of strategic media planning, highlighting significant interactions between prime time, program types, and company-specific factors. Results indicate that advertising effectiveness differs among industries, suggesting the need for sector-specific strategies. The findings underscore the importance of integrating traditional TV advertising with digital media strategies to maximize cross-platform engagement. These insights provide valuable guidance for advertisers and media planners in optimizing TV ad campaigns and their impact on digital platforms in an evolving media landscape.
- 30 January 2025
- Published by Erciyes Universitesi (DergiPark) in Erciyes İletişim Dergisi
- Vol. 12 (1) , 53-66
Open AccessThe Short-Term Effect of TV Advertisements on Digital Traffic
This study investigates the short-term effects of television advertisements (TV ads) on digital platform engagement across companies from diverse sectors. Utilizing statistical analysis methods including t-test, ANOVA, and Tukey's HSD test, we examined the impact of factors such as ad timing, program type, and sector on web traffic. Our findings reveal that prime time advertisements significantly influence web traffic, with effects varying across different sectors and program types. The study demonstrates the importance of strategic media planning, highlighting significant interactions between prime time, program types, and company-specific factors. Results indicate that advertising effectiveness differs among industries, suggesting the need for sector-specific strategies. The findings underscore the importance of integrating traditional TV advertising with digital media strategies to maximize cross-platform engagement. These insights provide valuable guidance for advertisers and media planners in optimizing TV ad campaigns and their impact on digital platforms in an evolving media landscape.
- Open Access
A Comprehensive Framework for Measuring the Immediate Impact of TV Advertisements: TV-Impact
Measuring the immediate impact of television advertisements (TV ads) on online traffic poses significant challenges in many aspects. Nonetheless, a comprehensive consideration is essential to fully grasp consumer reactions to TV ads. So far, the measurement of this effect has not been studied to a large extent. Existing studies have either determined how a specific focus group, i.e., toddlers, people of a certain age group, etc., react to ads via simple statistical tests using a case study approach or have examined the effects of advertising with simple regression models. This study introduces a comprehensive framework called TV-Impact. The framework uses a Bayesian structural time-series model called CausalImpact. There are additional novel approaches developed within the framework. One of the novelties of TV-Impact lies in its dynamic algorithm for selecting control variables which are supporting data sources and presumed to be unaffected by TV ads. In addition, we proposed the concept of Group Ads to combine overlapping ads into a single ad structure. Then, Random Forest Regressor, which is a commonly preferred supervised learning method, is used to decompose the impact into single ads. The TV-Impact framework was applied to the data of iLab, a venture company in Turkey, and manages its companies’ advertising strategies. The findings reveal that the TV-Impact model positively influenced the companies’ strategies for allocating their TV advertisement budgets and increased the amount of traffic driven to company websites, serving as an effective decision support system.
Open AccessA Comprehensive Framework for Measuring the Immediate Impact of TV Advertisements: TV-Impact
Measuring the immediate impact of television advertisements (TV ads) on online traffic poses significant challenges in many aspects. Nonetheless, a comprehensive consideration is essential to fully grasp consumer reactions to TV ads. So far, the measurement of this effect has not been studied to a large extent. Existing studies have either determined how a specific focus group, i.e., toddlers, people of a certain age group, etc., react to ads via simple statistical tests using a case study approach or have examined the effects of advertising with simple regression models. This study introduces a comprehensive framework called TV-Impact. The framework uses a Bayesian structural time-series model called CausalImpact. There are additional novel approaches developed within the framework. One of the novelties of TV-Impact lies in its dynamic algorithm for selecting control variables which are supporting data sources and presumed to be unaffected by TV ads. In addition, we proposed the concept of Group Ads to combine overlapping ads into a single ad structure. Then, Random Forest Regressor, which is a commonly preferred supervised learning method, is used to decompose the impact into single ads. The TV-Impact framework was applied to the data of iLab, a venture company in Turkey, and manages its companies’ advertising strategies. The findings reveal that the TV-Impact model positively influenced the companies’ strategies for allocating their TV advertisement budgets and increased the amount of traffic driven to company websites, serving as an effective decision support system.
- 21 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Demand Forecasting for Daily Retail Orders in Fresh Food Market
In the complex landscape of fresh food retail, accurate demand forecasting stands as a critical challenge, pivotal to both financial optimization and sustainable practices. This paper introduces an advanced approach to daily retail order predictions by applying multiple machine learning paradigms: Long Short-Term Memory (LSTM) networks, Transformers, Prophet, Feedforward Neural Networks (FNN), and eXtreme Gradient Boosting (XGBoost). Leveraging LSTM’s adeptness at processing sequential sales data, we harness its capability to discern temporal patterns and long-term dependencies inherent in the sales trajectory. Transformers can handle longer dependencies, which have been shown to be beneficial for our yearly seasonal data. On the other hand, XGBoost, with its robustness against overfitting and ability to capture nonlinear relationships, provides a granular understanding of the domain-related factors. FNNs are included as a strong baseline, and the included residual connections have proven crucial due to our sales data’s correlation with the next day’s sales being the most pronounced. Through experimentations with a dataset of sales data from the past seven years, our research highlights the strength of these methodologies. The results reveal that all models perform in acceptable ranges, however, Transformers, XGBoost and FNN models outperform LSTMs. Additionally, XGBoost exhibits slightly lower variation, which may be preferable.
- 21 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Demand Forecasting for Daily Retail Orders in Fresh Food Market
In the complex landscape of fresh food retail, accurate demand forecasting stands as a critical challenge, pivotal to both financial optimization and sustainable practices. This paper introduces an advanced approach to daily retail order predictions by applying multiple machine learning paradigms: Long Short-Term Memory (LSTM) networks, Transformers, Prophet, Feedforward Neural Networks (FNN), and eXtreme Gradient Boosting (XGBoost). Leveraging LSTM’s adeptness at processing sequential sales data, we harness its capability to discern temporal patterns and long-term dependencies inherent in the sales trajectory. Transformers can handle longer dependencies, which have been shown to be beneficial for our yearly seasonal data. On the other hand, XGBoost, with its robustness against overfitting and ability to capture nonlinear relationships, provides a granular understanding of the domain-related factors. FNNs are included as a strong baseline, and the included residual connections have proven crucial due to our sales data’s correlation with the next day’s sales being the most pronounced. Through experimentations with a dataset of sales data from the past seven years, our research highlights the strength of these methodologies. The results reveal that all models perform in acceptable ranges, however, Transformers, XGBoost and FNN models outperform LSTMs. Additionally, XGBoost exhibits slightly lower variation, which may be preferable.
- Open Access
- 16 December 2023
- Published by Kent Akademisi in Kent Akademisi
- Vol. 16 (4) , 2538-2552
Using Multi-Criteria Decision-Making for Smart City Evaluation and Ranking
As urban areas continue to grow and evolve, the development of smart cities has become a critical topic in urban planning and sustainability. Cities have been forced to become as smart as possible to deal with huge amounts of civic functions and boost economic growth, while constantly increasing citizens’ quality of life. They frequently invest in data-driven smart technologies for managing the vast amount of data and improving performance and efficiency. However, a city earns its smartness status when it satisfies criteria from various dimensions. Therefore, governments and local authorities need to direct their limited resources to the right domains. One way to determine it is to position the city among its rivals. The paper proposes a methodology for evaluating and ranking the smartness of 48 cities worldwide using two techniques: TOPSIS and the entropy weight method. The proposed framework examines the smartness of the cities considering seven dimensions. The results reveal a comprehensive ranking that not only accounts for the cities' overall performance in smart city development but also highlights specific strengths and weak areas for improvement. The proposed approach can be utilized as a decision support tool for city planners, policymakers, and stakeholders to identify best practices for the development of sustainable smart cities, evaluate progress, and direct investments to the right areas.
- 16 December 2023
- Published by Kent Akademisi in Kent Akademisi
- Vol. 16 (4) , 2538-2552
Open AccessUsing Multi-Criteria Decision-Making for Smart City Evaluation and Ranking
As urban areas continue to grow and evolve, the development of smart cities has become a critical topic in urban planning and sustainability. Cities have been forced to become as smart as possible to deal with huge amounts of civic functions and boost economic growth, while constantly increasing citizens’ quality of life. They frequently invest in data-driven smart technologies for managing the vast amount of data and improving performance and efficiency. However, a city earns its smartness status when it satisfies criteria from various dimensions. Therefore, governments and local authorities need to direct their limited resources to the right domains. One way to determine it is to position the city among its rivals. The paper proposes a methodology for evaluating and ranking the smartness of 48 cities worldwide using two techniques: TOPSIS and the entropy weight method. The proposed framework examines the smartness of the cities considering seven dimensions. The results reveal a comprehensive ranking that not only accounts for the cities' overall performance in smart city development but also highlights specific strengths and weak areas for improvement. The proposed approach can be utilized as a decision support tool for city planners, policymakers, and stakeholders to identify best practices for the development of sustainable smart cities, evaluate progress, and direct investments to the right areas.