R/fct_mod_text.R

In qiushiyan/nyclodging: Airbnb listings in New York City in 2019

#' mod_text
#'
#' @description predict price range classification
#'
#' @return The return value, if any, from executing the function.
#'
#' @import sf
#' @importFrom nycgeo nyc_boundaries
#' @importFrom readr read_rds
#' @import workflows
#' @import recipes
#' @import textrecipes
#' @importFrom broom augment
#' @importFrom shinyFeedback showFeedbackWarning
#' @noRd
predice_price <- function(classification_model, lon, lat, neighbourhood, description) {
  df_to_predict <- tibble(
    list_description = description,
    lon = lon,
    lat = lat,
    neighbourhood_group = neighbourhood
  )


  # predict
  df_predicted <- augment(classification_model, df_to_predict) %>%
    select(starts_with(".pred"), -`.pred_class`) |>
    tidyr::pivot_longer(everything(),
      names_to = "class",
      values_to = "prob",
      names_prefix = ".pred_"
    ) %>%
    mutate(class = factor(class, levels = rev(c("< 100", "100 to 200", "200 to 300", "300 to 400", "400 to 500", "500 to 1000", "> 1000"))))


  df_predicted
}

#'
#' @description get parent nyc neighbourhood based on user input
#' @noRd
get_neighbourhood <- function(nyc_borough, lon, lat) {
  selected <- data.frame(
    lon = lon,
    lat = lat
  ) %>%
    st_as_sf(coords = c(lon = "lon", lat = "lat")) %>%
    st_set_crs(4326)

  neighbourhood <- selected %>%
    st_join(nyc_borough) %>%
    .[["borough_name"]]

  neighbourhood
}