What is the "contamination" parameter in the `isolation.forest` function in R?

48 Views Asked by At

What is the "contamination" parameter in the isolation.forest function of the isotree package in R?

https://www.rdocumentation.org/packages/isotree/versions/0.1.24/topics/isolation.forest

isolation.forest(
  df,
  sample_weights = NULL,
  column_weights = NULL,
  sample_size = NROW(df),
  ntrees = 500,
  ndim = min(3, NCOL(df)),
  ntry = 3,
  max_depth = ceiling(log2(sample_size)),
  prob_pick_avg_gain = 0,
  prob_pick_pooled_gain = 0,
  prob_split_avg_gain = 0,
  prob_split_pooled_gain = 0,
  min_gain = 0,
  missing_action = ifelse(ndim > 1, "impute", "divide"),
  new_categ_action = ifelse(ndim > 1, "impute", "weighted"),
  categ_split_type = "subset",
  all_perm = FALSE,
  coef_by_prop = FALSE,
  recode_categ = TRUE,
  weights_as_sample_prob = TRUE,
  sample_with_replacement = FALSE,
  penalize_range = TRUE,
  weigh_by_kurtosis = FALSE,
  coefs = "normal",
  assume_full_distr = TRUE,
  build_imputer = FALSE,
  output_imputations = FALSE,
  min_imp_obs = 3,
  depth_imp = "higher",
  weigh_imp_rows = "inverse",
  output_score = FALSE,
  output_dist = FALSE,
  square_dist = FALSE,
  random_seed = 1,
  handle_interrupt = TRUE,
  nthreads = parallel::detectCores()
)
0

There are 0 best solutions below