A Population Spatialization Model at the Building Scale Using Random Forest

Population spatialization reveals the distribution and quantity of population in geographic space with gridded maps. Fine-scale is essential for urbanization disaster prevention. Previous approaches have used remotely sensed imagery to disaggregate census data, but this approach has limitations. For example, large-scale censuses cannot be conducted underdeveloped countries or regions, remote sensing data lack semantic information indicating different human activities occurring a precise location. Geospatial big machine learning provide new fine-scale mapping methods. In paper, 30 features are extracted using easily accessible multisource data. Then, building-scale estimation model trained by random forest (RF) regression algorithm. The results show that 91% buildings Lin’an District absolute error values less than six compared actual comparison multiple linear (ML) model, mean errors RF ML models 2.52 3.21, respectively, root squared 8.2 9.8, R2 0.44 0.18. performs better at Future work will improve accuracy densely populated areas.