Table 1.
Description of spatial datasets and multicollinearity.
Table 2.
Sample structure of the PU training dataset used for model development.
Fig 1.
Workflow of the PU learning framework for illegal dump risk mapping.
Table 3.
ML models and corresponding hyperparameter ranges used for tuning.
Fig 2.
Spatial distribution of IDSs.
Fig 3.
Spatial distribution of (a) population density, (b) poverty index, (c) household waste generation rate, (d) DEM, (e) LST, (f) STS, (g) road intersection, (h) buildings, (i) rail and road, (j) drain and waterbody.
Table 4.
Descriptive statistics of the predictor variables.
Table 5.
Performance comparison of ML models.
Table 6.
Model ranking across different weighting schemes.
Fig 4.
ROC curves for seven models.
Table 7.
Hold-out test set performance of RF.
Fig 5.
(a) Probability calibration curve and (b) prediction distribution of RF.
Fig 6.
Illegal dump risk score across KCC.
Fig 7.
Predictor importance analysis.