Modeling Urban Land Markets in Data-Scarce Cities: A Spatial Big Data Mining Approach to Building Density Dynamics in Kigali

Background: Urban land markets significantly influence city development. However, analyzing them in data-scarce environments presents challenges. This study utilizes spatial big data mining techniques to analyze building density dynamics as a proxy for understanding urban land market behavior in Kigali, Rwanda. Methods: High-resolution building footprint data from 2023 were integrated with cadastral and administrative datasets to compute the housing density across 501,170 land parcels. Spatial autocorrelation analysis and k-means clustering were applied to segment the cities into distinct housing density zones. The clustering results were validated using Silhouette, Davies-Bouldin, and Calinski-Harabasz scores. Results: Spatial autocorrelation analysis revealed a significant clustering of housing density (Moran's I = 0.9780, p < 0.001). K-means clustering was used to identify the five distinct housing density zones. High-density clusters (average density 0.34) were concentrated in the central districts, encompassing 9.93% of the land area and 23.9% of the parcels. Extensive low-density zones (average density 0.03-0.09) dominated 90.07% of the land area. Conclusions: The spatial distribution of housing density clusters aligns with bid-rent theory and monocentric city models. This study demonstrates that building density, when analyzed through spatial big data mining, can provide critical insights into urban land market dynamics in data-scarce environments. This methodology offers a replicable framework for inferring land value gradients and market pressures in rapidly urbanizing cities lacking traditional valuation data.