Samenvatting

Daskan Inc. manages a diverse portfolio of engineering projects and faces a recurring challenge inaccurately estimating the engineering effort required and understanding differences in project workload and difficulty at the early stage of a project. Currently, these estimations rely heavily on expert judgment without systematic validation against historical outcomes. This lack of a structured, data-driven approach leads to significant variability in workload distribution, pricing inefficiencies, and the risk of employee overload during peak periods. Furthermore, seasonal fluctuations in both workforce availability and project demand introduce additional layers of uncertainty into the planning process. Over the past four years, the company has accumulated a rich repository of internal data, including employee timesheets, project folders, and calculation logs. However, these records are stored across different sources and have not yet been systematically analysed. This represents a significant missed opportunity; the company possesses the necessary data but lacks the analytical tools required to transform these records into actionable insights for workforce planning and cost control. The increasing variability in projects and complexity has made it difficult for the company to accurately estimate engineering effort using experience-based methods alone. This created a need for a more structured and data-driven approach to support planning and decisions.

The objective of this research is to apply a structured data science approach, governed by the CRISP-DM methodology, to develop a predictive framework for project management. The primary goal is twofold: first, to build a regression model that predicts total engineering effort (man-hours); and second, to develop a classification model to categorise project complexity (low, medium, and high), where complexity is derived from project characteristics and engineering effort patterns. By integrating structural variables such as square footage (m2) and the number of levels, the model aims to provide a more reliable foundation for resource allocation, helping the organisation anticipate requirements and strengthen its competitive position. The project fits within the data science domain. Focusing on supervised machine learning techniques for the prediction and support in construction project management.