Table 1.
Datasets Used in This Study.
Fig 1.
Overall Process of the Methodology.
Fig 2.
Ensemble Learning Model Architecture.
Fig 3.
Ranking of the top 15 POI types based on their importance to population attraction.
Fig 4.
Pearson correlation coefficient matrix between the top 15 POI types based on importance.
Fig 5.
POI types with lower correlation (shown in red).
Table 2.
The abbreviations for the independent variables of building attributes.
Fig 6.
Results of the ensemble learning mode.
Table 3.
Comparison of accuracy results of various machine learning models on the training and test sets.
Fig 7.
SHAP feature importance plots for 6 base learners.
Fig 8.
SHAP feature contribution ranking plots for 6 base learners.
Fig 9.
SHAP summary plot and feature contribution ranking plot of the meta-learner.
Fig 10.
SHAP summary plot and feature contribution ranking plot for the overall contribution of the Stacking model.
Fig 11.
SHAP Summary Plot and Feature Contribution Ranking Plot for the Overall Contribution of the Stacking Model in Rural Areas.
Fig 12.
SHAP Summary Plot and Feature Contribution Ranking Plot for the Overall Contribution of the Stacking Model in Urban Areas.
Fig 13.
Spatial Distribution and Description of Yuxi City’s 2020 Population at 50m Resolution.
Fig 14.
Validation of Population Prediction Results at the Administrative Village Level for Four Datasets, (a) A linear fit plot of the log of village statistical population versus the simulation results of the validation set of the spatialized model of ensemble learning, (b) A linear fit plot of the log of village statistical population versus the log of the population from the WorldPop dataset, (c) A linear fit plot of the log of village statistical population versus the log of the population from the landscan dataset, (d) A linear fit plot of the log of village statistical population versus the log of the population from the GPWv4 dataset.
Fig 15.
Population spatial distribution maps for the central urban area of Yuxi derived from different data sources.