xgb_stack: xgb predictions
In mengluchu/APMtools: statistical air pollution mapping tools

XGboost prediction on rasterstacks, allowing missing value

xgb_stack(
  sr,
  df_var,
  y_var,
  xgbname = "xgb.tif",
  max_depth,
  eta,
  gamma,
  nrounds,
  subsample,
  verbose = 0,
  xgb_lambda,
  xgb_alpha
)

`sr`	raster stack of predictors
`df_var`	dataframe containing predictors and responses
`y_var`	specify the name of the response
`xgbname`	specify the name/directoray of the predicted raster file to save

use as.matrix instead of a sparse matrix model, the R Matrix package is therefore not needed and missing data is kept in, though in situations with lots of zeros, it is less efficient than using a sparse matrix. The same dense array is used in the xgboost_imp, xgboost_LUR, and xgb_pre. The other funciton, predicLA_RF_XGB_tiles used a sparse matrix, which means rows with missing values are missed.

mengluchu/APMtools documentation built on Jan. 27, 2022, 2:41 a.m.