Skip to main content

Table 4 Important variables modeled by the quantitative Random forests and evaluation of predicted maps modeled by combined models. The variable importance (%) used to predict pig types and pig farm scales and the evaluation of the combined models. Predictor variables include, travel time to the capital city (Bangkok), travel time to the provincial capitals (Meung districts), rainfed croplands irrigated croplands, elevation, and human density)

From: Spatial analysis and characteristics of pig farming in Thailand

Categories Response variablesa The variable importanceb Evaluation
TCapCity TProCap RaCrop IrCrop Elev HuDen RMSEc Correlation RMSE Correlation
(sub-district) (sub-district) (pixel) (pixel)
Pig types (heads/km2) Native pigs 41.97 27.54 29.44 23.86 34.64 66.89 0.12 0.94 1.19 0.78
Breeding pigs 58.74 38.84 36.76 30.46 44.82 75.38 0.23 0.91 1.32 0.79
Fattening pigs 61.16 37.57 47.71 33.07 63.81 63.55 0.31 0.87 1.28 0.83
Pig farm scales (farms/10 km2) SM 100.27 46.86 62.10 50.90 77.83 148.82 0.14 0.95 1.43 0.80
LF 21.85 21.68 25.33 15.39 36.33 62.09 0.07 0.92 0.74 0.74
  1. aResponse variables include: number of native pigs, number of breeding pigs, number of fattening pig, number of smallholders (SM), and number of large-scale farming systems (LF)
  2. bPredictor variables include: travel time to the capital city (TCapCity), travel time to the provincial capitals (TProCap), rainfed croplands (RaCrop), irrigated croplands (IrCrop), elevation (Elev), and human density (HuDen)
  3. cRMSE stands for root mean square error