Data Science Projects: A Systematic Literature Review on Characteristics, Implementation, and Challenges
19th International Conference on Information Technology and Applications, ICITA 2025, Oslo, Norveç, 14 - 16 Ekim 2025, cilt.1889 LNNS, ss.374-384, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 1889 LNNS
- Doi Numarası: 10.1007/978-3-032-21174-3_34
- Basıldığı Şehir: Oslo
- Basıldığı Ülke: Norveç
- Sayfa Sayıları: ss.374-384
- Anahtar Kelimeler: Data Science, Project Challenges, Project Characteristics, Project Management, Project Phases, Systematic Literature Review
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
Data science (DS) project implementation presents unique challenges distinguishing it from traditional software development, yet literature lacks comprehensive synthesis of these differences. This systematic review examines DS project implementation through four research questions addressing goals, execution approaches, challenges, and distinguishing characteristics. Following established methodology, we analyzed 30 publications selected from 446 studies via rigorous screening of Web of Science publications. Analysis reveals DS project challenges across four interconnected dimensions: Data (quality, availability, governance), Organization (coordination, resources, culture), Technology (scalability, infrastructure), and Strategy (vision alignment, value realization). Organizational challenges emerged as most prevalent, highlighting human and structural factors’ critical importance. DS goals demonstrate strategic alignment across financial, customer, process, and learning perspectives, primarily enabling informed decision-making. Synthesis of DS methodologies reveals four overarching phases: Business, Data, Model, and Product. However, contemporary practice shows growing adoption of hybrid approaches integrating agile principles with structured methodologies to address DS projects’ experimental nature. Six characteristics distinguish DS projects: (1) data centrality, (2) evolving requirements, (3) inherent experimentation, (4) multidisciplinary collaboration needs, (5) continuous post-deployment monitoring, and (6) non-linear lifecycles. These findings provide a comprehensive characterization of DS project implementation patterns, offering structured guidance for practitioners while recognizing projects’ exploratory, iterative, data-centric nature. Future research should develop DS-specific methodologies rather than adapting existing frameworks.