LAUSR

This study makes use of Wi-Fi connectivity data to understand how physical spaces are utilized and how it can be segmented, from which the insight gained can facilitate spatial planning and design. To carry out this study, we used a Wi-Fi connectivity data collected from a university network of 291,124 devices from 2,980 access points located across three campuses. For space segmentation, we’ve defined three features that characterize space utilization: crowdedness, mobility, and connectivity entropy. We’ve developed a new method called Xplaces that employs PCA to reduce high dimensionality of the features, eigendecomposition to extract behavioral signatures of the access points, and X-means to cluster access points without predefined number of clusters. Silhouette value was used to measure how well clusters were formed for our evaluation. Our method outperforms the state-of-the-art model i.e., eigenplaces. Our further investigation on the impact of area usage temporality on space segmentation shows that the Xplaces performs better with specific features for different temporal observation windows. For example, Xplaces works well with the crowdedness feature for the weekend’s space segmentation. A set of recommended features for different temporal windows is thus also part of our study’s contributions in addition to the development of the Xplaces.