ĐNĐorđe Nedeljković

Đorđe Nedeljković

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Đorđe Nedeljković has a total of 11 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Project Delivery Method Labels to Delivery Characteristics: A Unified Framework for Analysing Construction Project Delivery

    by Zorana Petojević, Đorđe Nedeljković, Konrad Graser

    Project delivery methods are usually discussed through method labels such as Design-Bid-Build, Design-Build, Construction Management, Integrated Project Delivery, Alliancing, and Public-Private Partnership. However, these labels often conceal substantial variation in contractual interfaces, governance, timing of involvement, pricing, risk allocation, and collaboration mechanisms. This paper systematises project delivery method characteristics extracted from a structured analysis of construction project delivery literature. The analysis covered 50 project delivery method and delivery-approach codes. From these, 224 method-specific characteristics were extracted and consolidated into 61 substantive unified characteristics. In addition, two auxiliary taxonomy/evidence records were retained to preserve terminology traceability during cross-model normalization. The resulting framework groups characteristics into contract, governance, process, timing, procurement, pricing, risk, collaboration, lifecycle, finance, supply chain, technology, implementation, cost, liability, and context categories. The paper contributes a reusable dictionary of project delivery characteristics that is independent of method labels and clarifies what each characteristic means. This supports more precise comparison of delivery configurations and provides a foundation for subsequent analysis of how delivery characteristics influence project outcomes.

  • Information Overload in Construction Projects in the Age of Generative AI: Mental Health Risks and Knowledge Management Strategies

    by Đorđe Nedeljković, Diana Vranešević

    Construction projects depend on dense information exchange between clients, designers, contractors, consultants, suppliers, and regulators. Digital tools, Building Information Modelling, Common Data Environments, dashboards, e-mails and collaboration platforms have improved access to project information, but they have also increased the volume, speed, and fragmentation of communication. This paper offers a narrative synthesis of information overload as a mental health challenge in construction project teams, with particular attention to the emerging role of generative artificial intelligence. The literature suggests that AI and large language models (LLMs) can support teams through summarization, retrieval, drafting, document analysis, planning support and automation of repetitive information work. At the same time, poorly governed AI adoption may increase information production, accelerate response expectations and shift human effort toward continuous verification and high-level cognitive oversight. The paper argues that AI should be understood neither as a simple cure for overload nor as a direct threat to wellbeing. Its likely effect depends on knowledge management maturity, information governance, clear human review responsibilities and realistic workload design.

  • Open Access

    From Observed Damage to Executable Repair Quantities: The RELA Repair Module

    by Zoran Stojadinović, Đorđe Nedeljković, Marija Ivanović, Dejan Marinković, Nevena Simić, Marko Marinković, Božidar Stojadinović

    Postearthquake workflows currently support three distinct classes of decisions: safety clearance through rapid tagging, portfolio‐level economic damage estimation through probabilistic loss models, and field inspections based on locally defined survey forms, which may subsequently trigger project‐specific strengthening through retrofit design. However, commitment to executable repair work at the building level, what to repair, where, and in what quantities, remains largely ad hoc, despite being required early in recovery for the majority of lightly to moderately damaged buildings. This paper introduces the RELA Repair Module (RELA‐RM), a rapid, observation‐driven framework that translates visually observed damage into executable repair quantities through predefined decision logic and standardized repair norms. Damage is recorded at the level of individual building elements and transformed into a building‐level Repair Bill of Quantities (RBoQ) expressed in person‐hours and material quantities. Repair cost and duration are subsequently derived through externally supplied pricing and workforce assumptions, while optional qualitative escalation indications flag cases that may require detailed engineering assessment. Spatial localization of repair actions is retained throughout the workflow to preserve building‐level organization of repair scope. Uncertainty is treated as bounded variability arising from three identifiable sources within the workflow: photographic measurement effects, human annotation variability, and dispersion in standardized productivity norms. These contributions are propagated to produce an approximate combined dispersion of ≈26%, interpreted as an engineering bound on expected variability in executable repair quantities. The framework is implemented and demonstrated for masonry walls, while the underlying decision principle is intended to be extensible to other building elements and typologies.


  • Digital Support in the Construction Industry: The Role of Large Language Models in Risk Management

    by Srbija Građevinski fakultet Univerziteta u Beogradu, Marija Ivanović, Đorđe Nedeljković, Zoran Stojadinović

    The construction industry frequently suffers cost and schedule overruns. Contributing factors include fragmented risk processes and the laborious manual review of large volumes of unstructured text. In 2022, the DREAM model demonstrated that combining machine-learning classifiers with expert knowledge can uncover the root causes of delays more systematically than manual review. However, it required extensive retraining for each new context and was prone to bias from experts. Recent advances in large language models, particularly GPT-4, offer zero or few-shot adaptability across vast unstructured text, eliminating the need for bespoke annotation. This paper evaluates whether an off-the-shelf GPT-4 model, used in zero-shot mode, can accurately classify causes of delay (CoDs) from meeting minutes without fine-tuning. A protocol for prompt engineering with iterative self-consistency checks is implemented to maximize reliability. Experimental results, including confusion-matrix analysis, demonstrate that ChatGPT-4 outperforms DREAM by closing the performance gap while significantly reducing annotation effort. By quantifying annotation savings and characterizing residual biases, we demonstrate that modern large language models (LLMs) enable scalable, data-efficient AI support in construction risk management.