format_observation: Format the occurrence dataset for usage in 'itsdm'

In itsdm: Isolation Forest-Based Presence-Only Species Distribution Modeling

Format the occurrence dataset for usage in itsdm

Description

The focus of this function is to format the dataset but to keep the dataset as original as possible. Then the users can modify the data if they want before put it into this function.

Usage

format_observation(
  obs_df,
  eval_df = NULL,
  split_perc = 0.3,
  seed = 123,
  obs_crs = 4326,
  eval_crs = 4326,
  x_col = "x",
  y_col = "y",
  obs_col = "observation",
  obs_type = "presence_only"
)

Arguments

obs_df	(data.frame). The data.frame style table that include x and y coordinate and observation of training dataset. This parameter is required as it is the training dataset. Note: it only takes data.frame to reduce the risk of column name mismatch between data.frame and other formats such as tibble.
eval_df	(data.frame or NULL) The data.frame style table that include x and y coordinate and observation of evaluation dataset. Note: it only takes data.frame to reduce the risk of column name mismatch between data.frame and other formats such as tibble.
split_perc	(numeric) a numeric between 0 and 1 corresponding to the percentage of data used to evaluate the models. Only required if eval_df is NULL.
seed	(integer) The seed to split train and evaluation set. The default value is 123. Only required if eval_df is NULL.
obs_crs	(integer, numeric, character, or crs) The EPSG code, CRS string, or sf::crs object of the coordinate system of the training dataset. It corresponds to x_col and y_col in obs_df.
eval_crs	(integer, numeric, character, or crs) The EPSG code, CRS string, or sf::crs object of the coordinate system of the evaluation dataset. Only required if eval_df is not NULL. It corresponds to x_col and y_col in eval_df if any.
x_col	(character) The name of column that is x coordinate in obs_df and eval_df if not NULL.
y_col	(character) The name of column that is y coordinate in obs_df and eval_df if not NULL.
obs_col	(character) The name of column that represents observations in obs_df and eval_df if not NULL.
obs_type	(character) The type of observation to be formatted to. Only can be one of c("presence_only", "presence_absence"). Note that if "presence_only" is set, the absences in obs_df will be deleted. This only affect obs_df, eval_df will keep the original type no matter it is an independent one or is split from eval_df.

Value

(FormatOccurrence) A list of

• obs (sf) the formatted pts of observations. The column of observation is "observation".

• obs_type (character) the type of the observations, presence_only or presence_absence.

• has_eval (logical) whether evaluation dataset is set or generated.

• eval (sf) the formatted pts of observations for evaluation if any. The column of observation is "observation".

• eval (eval_type) the type of the observations for evaluation, presence_only or presence_absence.

See Also

print.FormatOccurrence

Examples

library(dplyr)
library(itsdm)
data("occ_virtual_species")

# obs + eval, presence-absence
obs_df <- occ_virtual_species %>% filter(usage == "train")
eval_df <- occ_virtual_species %>% filter(usage == "eval")
x_col <- "x"
y_col <- "y"
obs_col <- "observation"
obs_type <- "presence_absence"

obs <- format_observation(
  obs_df = obs_df, eval_df = eval_df,
  x_col = x_col, y_col = y_col, obs_col = obs_col,
  obs_type = obs_type)

# obs + eval, presence-only
obs_df <- occ_virtual_species %>% filter(usage == "train")
eval_df <- occ_virtual_species %>% filter(usage == "eval")
x_col <- "x"
y_col <- "y"
obs_col <- "observation"
obs_type <- "presence_only"

obs <- format_observation(
  obs_df = obs_df, eval_df = eval_df,
  x_col = x_col, y_col = y_col, obs_col = obs_col,
  obs_type = obs_type)

# obs + eval, different crs, presence-only
obs_df <- occ_virtual_species %>% filter(usage == "train")
eval_df <- occ_virtual_species %>% filter(usage == "eval")
obs_crs <- 4326
# Fake one
eval_crs <- 20935
x_col <- "x"
y_col <- "y"
obs_col <- "observation"
obs_type <- "presence_only"

obs <- format_observation(
  obs_df = obs_df, eval_df = eval_df,
  obs_crs = obs_crs, eval_crs = eval_crs,
  x_col = x_col, y_col = y_col, obs_col = obs_col,
  obs_type = obs_type)

# obs + split, presence-absence
obs_df <- occ_virtual_species
split_perc <- 0.5
seed <- 123
obs_crs <- 4326
x_col <- "x"
y_col <- "y"
obs_col <- "observation"
obs_type <- "presence_absence"

obs <- format_observation(
  obs_df = obs_df, split_perc = split_perc,
  x_col = x_col, y_col = y_col,
  obs_col = obs_col, obs_type = obs_type)

# obs, presence-only, no eval
obs_df <- occ_virtual_species
eval_df <- NULL
split_perc <- 0
seed <- 123
obs_crs <- 4326
x_col <- "x"
y_col <- "y"
obs_col <- "observation"
obs_type <- "presence_only"

obs <- format_observation(
  obs_df = obs_df, eval_df = eval_df,
  split_perc = split_perc,
  x_col = x_col, y_col = y_col,
  obs_col = obs_col, obs_type = obs_type)

itsdm documentation built on July 9, 2023, 6:45 p.m.