Fits several random forest models on the same data in order to capture the effect of the algorithm's stochasticity on the variable importance scores, predictions, residuals, and performance measures. The function relies on the median to aggregate performance and importance values across repetitions. It is recommended to use it after a model is fitted (rf() or rf_spatial()), tuned (rf_tuning()), and/or evaluated (rf_evaluate()). This function is designed to be used after fitting a model with rf() or rf_spatial(), tuning it with rf_tuning() and evaluating it with rf_evaluate().
rf_repeat(
model = NULL,
data = NULL,
dependent.variable.name = NULL,
predictor.variable.names = NULL,
distance.matrix = NULL,
distance.thresholds = NULL,
xy = NULL,
ranger.arguments = NULL,
scaled.importance = FALSE,
repetitions = 10,
keep.models = TRUE,
seed = 1,
verbose = TRUE,
n.cores = parallel::detectCores() - 1,
cluster = NULL
)A model fitted with rf(). If provided, the data and ranger arguments are taken directly from the model definition (stored in model$ranger.arguments). Default: NULL
Data frame with a response variable and a set of predictors. Default: NULL
Character string with the name of the response variable. Must be in the column names of data. If the dependent variable is binary with values 1 and 0, the argument case.weights of ranger is populated by the function case_weights(). Default: NULL
Character vector with the names of the predictive variables. Every element of this vector must be in the column names of data. Default: NULL
Squared matrix with the distances among the records in data. The number of rows of distance.matrix and data must be the same. If not provided, the computation of the Moran's I of the residuals is omitted. Default: NULL
Numeric vector with neighborhood distances. All distances in the distance matrix below each value in dustance.thresholds are set to 0 for the computation of Moran's I. If NULL, it defaults to seq(0, max(distance.matrix), length.out = 4). Default: NULL
(optional) Data frame or matrix with two columns containing coordinates and named "x" and "y". It is not used by this function, but it is stored in the slot ranger.arguments$xy of the model, so it can be used by rf_evaluate() and rf_tuning(). Default: NULL
Named list with ranger arguments (other arguments of this function can also go here). All ranger arguments are set to their default values except for 'importance', that is set to 'permutation' rather than 'none'. Please, consult the help file of ranger if you are not familiar with the arguments of this function.
Logical. If TRUE, and 'importance = "permutation', the function scales 'data' with scale and fits a new model to compute scaled variable importance scores. Default: FALSE
Integer, number of random forest models to fit. Default: 10
Logical, if TRUE, the fitted models are returned in the models slot. Set to FALSE if the accumulation of models is creating issues with the RAM memory available. Default: TRUE.
Integer, random seed to facilitate reproduciblity. If set to a given number, the results of the function are always the same. Default: 1.
Logical, ff TRUE, messages and plots generated during the execution of the function are displayed, Default: TRUE
Integer, number of cores to use for parallel execution. Creates a socket cluster with parallel::makeCluster(), runs operations in parallel with foreach and %dopar%, and stops the cluster with parallel::clusterStop() when the job is done. Default: parallel::detectCores() - 1
A cluster definition generated with parallel::makeCluster(). If provided, overrides n.cores. When cluster = NULL (default value), and model is provided, the cluster in model, if any, is used instead. If this cluster is NULL, then the function uses n.cores instead. The function does not stop a provided cluster, so it should be stopped with parallel::stopCluster() afterwards. The cluster definition is stored in the output list under the name "cluster" so it can be passed to other functions via the model argument, or using the %>% pipe. Default: NULL
A ranger model with several new slots:
ranger.arguments: Stores the values of the arguments used to fit the ranger model.
importance: A list containing a data frame with the predictors ordered by their importance, a ggplot showing the importance values, and local importance scores.
performance: out-of-bag performance scores: R squared, pseudo R squared, RMSE, and normalized RMSE (NRMSE).
pseudo.r.squared: computed as the correlation between the observations and the predictions.
residuals: residuals, normality test of the residuals computed with residuals_test(), and spatial autocorrelation of the residuals computed with moran_multithreshold().
if(interactive()){
#loading example data
data(plant_richness_df)
data(distance_matrix)
#fitting 5 random forest models
out <- rf_repeat(
data = plant_richness_df,
dependent.variable.name = "richness_species_vascular",
predictor.variable.names = colnames(plant_richness_df)[5:21],
distance.matrix = distance_matrix,
distance.thresholds = 0,
repetitions = 5,
n.cores = 1
)
#data frame with ordered variable importance
out$importance$per.variable
#per repetition
out$importance$per.repetition
#variable importance plot
out$importance$per.repetition.plot
#performance
out$performance
#spatial correlation of the residuals for different distance thresholds
out$spatial.correlation.residuals$per.distance
#plot of the Moran's I of the residuals for different distance thresholds
out$spatial.correlation.residuals$plot
#using a model as an input for rf_repeat()
rf.model <- rf(
data = plant_richness_df,
dependent.variable.name = "richness_species_vascular",
predictor.variable.names = colnames(plant_richness_df)[8:21],
distance.matrix = distance_matrix,
distance.thresholds = 0,
n.cores = 1
)
#repeating the model 5 times
rf.repeat <- rf_repeat(
model = rf.model,
n.cores = 1
)
rf.repeat$performance
rf.repeat$importance$per.repetition.plot
}