predictAutoforest {autocart}

Make a prediction using an autoforest model returned from the autoforest function.

Description

Make a prediction using an autoforest model returned from the autoforest function.

Usage

predictAutoforest(
  autoforestModel,
  newdata,
  newdataCoords = NULL,
  useSpatialNodes = FALSE,
  method = "idw",
  distpower = 2,
  distpowerRange = c(0, 2),
  modelByResidual = TRUE,
  decideByGC = FALSE
)

Arguments

autoforestModel

An S3 object of type "autoforest" returned from the autoforest function.

newdata

The dataframe of predictors for use in prediction.

newdataCoords

the matrix of locations for all the information in newdata. Required argument if you set "useSpatialNodes" to TRUE.

useSpatialNodes

If TRUE, instead of running all the observations through the autocart tree, use the spatialNodes function to make predictions.

method

If using the spatial nodes type of prediction, then the type of interpolation to use. The options are "idw" and "tps".

distpower

If using "idw" for the method, the power on distance. For example, setting this to 2 would mean inverse squared distance squared weighting.

distpowerRange

If using "idw" for the interpolation method, the range of distance powers to use on inverse distance weighting matched to terminal node Moran I measurements.

modelByResidual

When using interpolation, make a prediction using the region of interest's average and then interpolate the residual.

decideByGC

Use Geary's C in deciding to induce a local spatial process rather than Moran's I.

Value

A vector of predictions that correspond to the rows in newdata.

Examples

# Load some data for an autoforest example
snow <- na.omit(read.csv(system.file("extdata", "ut2017_snow.csv", package = "autocart")))
y <- snow$yr50[1:40]
X <- data.frame(snow$ELEVATION, snow$MCMT, snow$PPTWT, snow$HUC)[1:40, ]
locations <- as.matrix(cbind(snow$LONGITUDE, snow$LATITUDE))[1:40, ]

# Create a control object for the autoforest tree
snow_control <- autocartControl(spatialBandwidthProportion = 1.0)

# Create an autoforest model with 5 trees (low number chosen for computation time)
snow_model <- autoforest(y, X, locations, 0.30, 0, snow_control, numtrees = 5)

# Predict for a subset of the data
new_X <- X[1:10, ]
new_loc <- locations[1:10, ]
predicted_values <- predictAutoforest(snow_model, new_X, new_loc, TRUE)