Ali Nizam
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Ali Nizam has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Oğuzhan Oktay Büyük
Department of Computer Engineering, Fatih Sultan Mehmet Vakif University, Turkey
2 shared publications - Abdelrahman Taha Abdeltawab Abdellatif
Fatih Sultan Mehmet Vakif University
1 shared publication - Ertuğrul Islamoğlu
Fatih Sultan Mehmet Vakif University
1 shared publication - Ertuğrul İslamoğlu
Artificial Intelligence and Data Engineering Department, Fatih Sultan Mehmet Vakıf University, İstanbul, Türkiye
1 shared publication - İbrahim Enes Aydoğdu
FATİH SULTAN MEHMET VAKIF ÜNİVERSİTES
1 shared publication - Musa Aydin
Computer Engineering Department, Fatih Sultan Mehmet Vakıf University, İstanbul, Türkiye
1 shared publication
Latest Publications
- Open Access
- 28 May 2026
- Published by Elsevier in SSRN Electronic Journal
A Novel Deep Learning and Agent-Based Consensus System for Estimating Story Points
Accurate estimation of story points is critical for effective sprint planning in Agile Scrum. Previous empirical studies suggest that cognitive biases and incomplete historical data can lead to inconsistent and inaccurate estimates. This study empirically investigates whether incorporating consensus-driven learning mechanisms and shared, persistent knowledge representations can improve story point estimation accuracy in real-world software projects.We present a multi agent consensus-based estimation framework in which individual estimators are modeled using their historical task assignments to capture personalized estimation behavior. Agents iteratively exchange predictions in a planning-card–inspired process and compute an agreement score reflecting the level of consensus. When agreement exceeds a predefined threshold, the estimate is finalized; otherwise, agents update their models through a peer-informed adjustment process that integrates consensus-weighted feedback with individual prediction errors. This repository structurally encodes individual estimation experience conditioned on estimator-capabilities, capturing patterns such as implied task difficulty, time-allocation tendencies, assignment preferences. We evaluate the proposed approach using a dataset of 23,300 completed user stories collected from 16 open-source software projects, which is larger than those used in many prior empirical studies on story point estimation. The approach is compared against baseline deep neural models, including GPT2SP and Deep-SE, as well as existing collaborative learning techniques. Our framework achieves average estimation accuracies 78.5%, corresponding to improvements of up to 19% over a no-consensus variant, 35% over federated learning, 22% over multi-agent reinforcement learning, and 8% on spectral expectation-maximization. It even outperforms well-known industrial solutions, achieving an improvement of 10% on Deep-SE and 6% on GPT2SP.
- 28 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Novel Deep Learning and Agent-Based Consensus System for Estimating Story Points
Accurate estimation of story points is critical for effective sprint planning in Agile Scrum. Previous empirical studies suggest that cognitive biases and incomplete historical data can lead to inconsistent and inaccurate estimates. This study empirically investigates whether incorporating consensus-driven learning mechanisms and shared, persistent knowledge representations can improve story point estimation accuracy in real-world software projects.We present a multi agent consensus-based estimation framework in which individual estimators are modeled using their historical task assignments to capture personalized estimation behavior. Agents iteratively exchange predictions in a planning-card–inspired process and compute an agreement score reflecting the level of consensus. When agreement exceeds a predefined threshold, the estimate is finalized; otherwise, agents update their models through a peer-informed adjustment process that integrates consensus-weighted feedback with individual prediction errors. This repository structurally encodes individual estimation experience conditioned on estimator-capabilities, capturing patterns such as implied task difficulty, time-allocation tendencies, assignment preferences. We evaluate the proposed approach using a dataset of 23,300 completed user stories collected from 16 open-source software projects, which is larger than those used in many prior empirical studies on story point estimation. The approach is compared against baseline deep neural models, including GPT2SP and Deep-SE, as well as existing collaborative learning techniques. Our framework achieves average estimation accuracies 78.5%, corresponding to improvements of up to 19% over a no-consensus variant, 35% over federated learning, 22% over multi-agent reinforcement learning, and 8% on spectral expectation-maximization. It even outperforms well-known industrial solutions, achieving an improvement of 10% on Deep-SE and 6% on GPT2SP.
- Open Access
- 14 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 31335-31350
Optimizing Pre-Trained Code Embeddings With Triplet Loss for Code Smell Detection
Code embedding represents code semantics in vector form. Although code embedding-based systems have been successfully applied to various source code analysis tasks, further research is required to enhance code embedding for better code analysis capabilities, aiming to surpass the performance and functionality of static code analysis tools. In addition, standard methods for improving code embedding are essential to develop more effective embedding-based systems, similar to augmentation techniques in the image processing domain. This study aims to create a contrastive learning-based system to explore the potential of a generic method for enhancing code embedding for code classification tasks. A triplet loss-based deep learning network is designed to optimize in-class similarity and increase the distance between classes. An experimental dataset that contains code from Java, Python, and PHP programming languages and 4 different code smells is created by collecting code from open-source repositories on GitHub. We evaluate the proposed system’s effectiveness with widely used BERT, CodeBERT, and GraphCodeBERT pretrained models to create code embedding for the code classification task of code smell detection. Our findings indicate that the proposed system may offer improvements in accuracy, an average of 8% and a maximum of 13% for models. These results suggest that incorporating contrastive learning techniques into the generation process of code representation as a preprocessing step can enhance performance in code analysis.
- 14 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 31335-31350
Open AccessOptimizing Pre-Trained Code Embeddings With Triplet Loss for Code Smell Detection
Code embedding represents code semantics in vector form. Although code embedding-based systems have been successfully applied to various source code analysis tasks, further research is required to enhance code embedding for better code analysis capabilities, aiming to surpass the performance and functionality of static code analysis tools. In addition, standard methods for improving code embedding are essential to develop more effective embedding-based systems, similar to augmentation techniques in the image processing domain. This study aims to create a contrastive learning-based system to explore the potential of a generic method for enhancing code embedding for code classification tasks. A triplet loss-based deep learning network is designed to optimize in-class similarity and increase the distance between classes. An experimental dataset that contains code from Java, Python, and PHP programming languages and 4 different code smells is created by collecting code from open-source repositories on GitHub. We evaluate the proposed system’s effectiveness with widely used BERT, CodeBERT, and GraphCodeBERT pretrained models to create code embedding for the code classification task of code smell detection. Our findings indicate that the proposed system may offer improvements in accuracy, an average of 8% and a maximum of 13% for models. These results suggest that incorporating contrastive learning techniques into the generation process of code representation as a preprocessing step can enhance performance in code analysis.
- 22 November 2024
- Published by Springer Nature
- p. 351-361
Analysis of Code Similarity with Triplet Loss-Based Deep Learning System
- 22 November 2024
- Published by Springer Nature
- p. 351-361
Analysis of Code Similarity with Triplet Loss-Based Deep Learning System
- 15 September 2023
- Published by Elsevier in Journal of Systems and Software
- Vol. 206, 111851
Deep learning with class-level abstract syntax tree and code histories for detecting code modification requirements
- 15 September 2023
- Published by Elsevier in Journal of Systems and Software
- Vol. 206, 111851
Deep learning with class-level abstract syntax tree and code histories for detecting code modification requirements
- Open Access
- 28 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 34428-34441
Software Project Failure Process Definition
Many researchers have attempted to identify the factors behind software project failure and their solutions from different perspectives. However, there is a lack of a systematic and integrated process definition for failure like the process models for success. This study aims to build a process definition for software project failure as an anti-pattern by identifying the main phases and their relationship in terms of team behaviors. We researched software engineering literature and case studies to gather information about critical incidents and repeating behaviors of teams in failed projects into a novel dataset. Grounded theory was employed to build a theoretical foundation for failure phase definitions from collected data. The Design Structure Matrix and Bayesian Belief Network were utilized for quantitative assessment of transition between phases. The results revealed that common behavioral patterns occur in approximately 89 percent of the case studies supporting the decision to consider software project failure as a process. The proposed failure process has a simple structure that uses everyday concepts for phase names. It provides critical insights to software teams and even non-technical managers for evaluating the real progress of their projects and designing strategies to avoid failure.
- 28 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 34428-34441
Open AccessSoftware Project Failure Process Definition
Many researchers have attempted to identify the factors behind software project failure and their solutions from different perspectives. However, there is a lack of a systematic and integrated process definition for failure like the process models for success. This study aims to build a process definition for software project failure as an anti-pattern by identifying the main phases and their relationship in terms of team behaviors. We researched software engineering literature and case studies to gather information about critical incidents and repeating behaviors of teams in failed projects into a novel dataset. Grounded theory was employed to build a theoretical foundation for failure phase definitions from collected data. The Design Structure Matrix and Bayesian Belief Network were utilized for quantitative assessment of transition between phases. The results revealed that common behavioral patterns occur in approximately 89 percent of the case studies supporting the decision to consider software project failure as a process. The proposed failure process has a simple structure that uses everyday concepts for phase names. It provides critical insights to software teams and even non-technical managers for evaluating the real progress of their projects and designing strategies to avoid failure.
- 30 January 2021
- Published by Marmara University in International Journal of Advances in Engineering and Pure Sciences
- Vol. 33 (1) , 73-82
GraphQL için Sorgu Oluşturma Sürecinde Kullanılan Yöntemlerin Analizi ve İyileştirilmesi
Günümüzde teknolojik gelişmeler İnternete bağlanan toplam cihaz türü ve sayısında büyük artışa yol açmıştır. Sunucu makineler daha fazla istek almaya başlamış hem ağ trafiği hem de sunucu yanıt süresi olumsuz etkilenmiştir. Bu sorunları çözmek için geliştirilen GraphQL teknolojisi tek bir istekle birden fazla tablo, koleksiyon veya veri tabanına erişim sağlayarak toplu veri sorgulama ve değiştirmeye imkân vermektedir. Bu sayede cihaz başına düşen istek sayısı ve cihazların belleklerinde tutulacak veri boyutu azalır. Ancak GraphQL yeni bir teknoloji olduğundan henüz kod geliştirme sürecini yöneten ve kolaylaştıran araçlar tam olarak gelişmemiştir. Sunucu kısmında sorguları oluşturmak ve çalıştırmak için önemli ölçüde kodun elle yazılması gerekmektedir. Bu da yazılım geliştiricilere önemli bir iş yükü oluşturmaktadır. Bu çalışmada GraphQL sorgu geliştirme süreci, bu süreci kolaylaştırmak veya otomatikleştirmek için kullanılan araçlar, bu araçların kullandığı yöntemler ve sorgu geliştirme maliyetleri analiz edilmiştir. Bu maliyeti azaltmak için kodları otomatik oluşturan bir yöntem önerilmiş ve bir araç geliştirilmiştir. Geliştirilen yöntemin etkinliği diğer yöntemlerle karşılaştırılmış, sayısal olarak incelenmiş ve yazılımcıları birçok kodu tekrar yazmaktan kurtararak zamandan tasarruf sağladığı görülmüştür.
- 30 January 2021
- Published by Marmara University in International Journal of Advances in Engineering and Pure Sciences
- Vol. 33 (1) , 73-82
GraphQL için Sorgu Oluşturma Sürecinde Kullanılan Yöntemlerin Analizi ve İyileştirilmesi
Günümüzde teknolojik gelişmeler İnternete bağlanan toplam cihaz türü ve sayısında büyük artışa yol açmıştır. Sunucu makineler daha fazla istek almaya başlamış hem ağ trafiği hem de sunucu yanıt süresi olumsuz etkilenmiştir. Bu sorunları çözmek için geliştirilen GraphQL teknolojisi tek bir istekle birden fazla tablo, koleksiyon veya veri tabanına erişim sağlayarak toplu veri sorgulama ve değiştirmeye imkân vermektedir. Bu sayede cihaz başına düşen istek sayısı ve cihazların belleklerinde tutulacak veri boyutu azalır. Ancak GraphQL yeni bir teknoloji olduğundan henüz kod geliştirme sürecini yöneten ve kolaylaştıran araçlar tam olarak gelişmemiştir. Sunucu kısmında sorguları oluşturmak ve çalıştırmak için önemli ölçüde kodun elle yazılması gerekmektedir. Bu da yazılım geliştiricilere önemli bir iş yükü oluşturmaktadır. Bu çalışmada GraphQL sorgu geliştirme süreci, bu süreci kolaylaştırmak veya otomatikleştirmek için kullanılan araçlar, bu araçların kullandığı yöntemler ve sorgu geliştirme maliyetleri analiz edilmiştir. Bu maliyeti azaltmak için kodları otomatik oluşturan bir yöntem önerilmiş ve bir araç geliştirilmiştir. Geliştirilen yöntemin etkinliği diğer yöntemlerle karşılaştırılmış, sayısal olarak incelenmiş ve yazılımcıları birçok kodu tekrar yazmaktan kurtararak zamandan tasarruf sağladığı görülmüştür.
- 28 August 2019
- Published by Emerald Publishing in The International Journal of Information and Learning Technology
- Vol. 37 (1-2) , 1-13
Evaluation and improvement of the credit transfer models in the student exchange process
Purpose: Increasing flexibility and student mobility are among the most important objectives of today’s universities. The purpose of this paper is to discuss the effects of student exchange process (SEP) on a campus management system (CMS), compare different models and recommend an improvement handling transfer process. Design/methodology/approach: A comprehensive literature review of how credit transfer is handled in today’s CMS was conducted to understand the information management methods used in the student transfer process. On this basis, a flexible task-based system design is recommended to facilitate the transfer of educational gains between universities. For evaluating the effectiveness of different system designs in the SEP, the duration of information processing process steps was measured quantitatively via on-site observations and user interview in a university’s horizontal, vertical and the Erasmus student exchange (SE) data. Findings: Building a flexible system design based on a loosely coupled mapping between curriculum and educational activities, and increasing the self-management capabilities of a student will facilitate managing SE data in an integrated environment and reduce the university staff’s workload considerably. Originality/value: To the author’s knowledge, this is the first study making quantitatively measurement and comparison of different credit transfer methods of CMSs. Based on this result, the authors have recommended a new flexible method that supports increasing a student’s self-administration capabilities, reducing the workload of university staff, and contributes academic mobility.
- 28 August 2019
- Published by Emerald Publishing in The International Journal of Information and Learning Technology
- Vol. 37 (1-2) , 1-13
Evaluation and improvement of the credit transfer models in the student exchange process
Purpose: Increasing flexibility and student mobility are among the most important objectives of today’s universities. The purpose of this paper is to discuss the effects of student exchange process (SEP) on a campus management system (CMS), compare different models and recommend an improvement handling transfer process. Design/methodology/approach: A comprehensive literature review of how credit transfer is handled in today’s CMS was conducted to understand the information management methods used in the student transfer process. On this basis, a flexible task-based system design is recommended to facilitate the transfer of educational gains between universities. For evaluating the effectiveness of different system designs in the SEP, the duration of information processing process steps was measured quantitatively via on-site observations and user interview in a university’s horizontal, vertical and the Erasmus student exchange (SE) data. Findings: Building a flexible system design based on a loosely coupled mapping between curriculum and educational activities, and increasing the self-management capabilities of a student will facilitate managing SE data in an integrated environment and reduce the university staff’s workload considerably. Originality/value: To the author’s knowledge, this is the first study making quantitatively measurement and comparison of different credit transfer methods of CMSs. Based on this result, the authors have recommended a new flexible method that supports increasing a student’s self-administration capabilities, reducing the workload of university staff, and contributes academic mobility.