The slow integration of digital tools in the real estate industry, particularly for analyzing building data, presents significant yet underexploited potential. This thesis explores the use of occupancy sensor data and room attributes from three office buildings, demonstrating how analytical methods can enhance architectural planning and inform design decisions. Room features such as size, floor plan placement, presence of screens, video solutions, whiteboards, windows, table shapes, restricted access, and proximity to amenities like coffee machines and printers were examined for their influence on space utilization. Two datasets were analyzed: one recording daily room usage and the other summarizing usage over a consistent timeframe. Analytical methods included centered moving averages, seasonal decomposition, panel data analysis models such as between and mixed effects models, various regression techniques, decision trees, random forests, Extreme Gradient Boosting (XGBoost), and K-means clustering. Results revealed consistent seasonal patterns and identified key room attributes affecting usage, such as proximity to amenities, screen availability, floor level, and room size. Basic techniques proved valuable for initial data exploration, while advanced models uncovered critical patterns, with random forest and XGBoost showing high predictive accuracy. The findings emphasize the importance of diverse analytical techniques in understanding room usage. This study underscores the value of further exploration in refining models, incorporating additional factors, and improving prediction accuracy. It highlights the significant potential for cost reduction, time savings, and innovative solutions in the real estate industry.

Teknisk-naturvetenskapliga fakulteten, Uppsala universitet. Utgivningsort Uppsala/Visby. Handledare: Henrik Svensson, Ämnesgranskare: Justin Pearson, Examinator: Elisabet Andrésdóttir