feat: add a "match success rate" plot

[jdhoffa]

@jacobvjk and I were talking, and this is something that seems to have been useful for supervisor projects, however, this seems like a useful exported function from r2dii.plot.

See function here: https://github.com/RMI-PACTA/workflow.aggregate.loanbooks/blob/cbf818d5467bd37506816c6cceb89497bb37028d/R/plots.R#L464

[jacobvjk]

as mentioned here: RMI-PACTA/pacta.multi.loanbook#236

I think there is an argument for making the match success rate available in r2dii.plot so that it can be used in individual analyses in low threshold way. I am thinking IFC applications for example. It would also be good to have something like that available if/when we set up a P4B cookbook. But definitely no urgency right now

[cjyetman]

@jacobvjk @jdhoffa flagging that this was mentioned as something useful for when a P4B cookbook was created

[jacobvjk]

Given that we are spending tech time of this grant on submitting releases of all P4B related packages on CRAN, I would tend to think that we risk going beyond the budget here by adding a new plot. I believe it was marked as interesting but not in scope. If we still have resources left at the end, it may be worth preparing it for the next release though?

[jdhoffa]

^ Exactly what @jacobvjk said!

[cjyetman]

@jacobvjk does this achieve the goal?

plot_match_success_rate <- function(.data, ..., .x = matched) {
  .data %>%
    summarize(
      loan_size_outstanding = sum(loan_size_outstanding, na.rm = TRUE),
      .by = c(matched, ...)
    ) %>%
    arrange(matched, ...) %>%
    ggplot() +
    geom_col(aes({{.x}}, loan_size_outstanding, fill = matched)) +
    r2dii.plot::theme_2dii() +
    r2dii.plot::scale_fill_r2dii() +
    theme(legend.position = "top")
}

calculate_match_success_rate <- function(matched, loanbook) {
  merge_by <-
    c(
      sector_classification_system = "code_system",
      sector_classification_direct_loantaker = "code"
    )

  loanbook_with_sectors <-
    loanbook %>%
    purrr::modify_at(names(merge_by), as.character) %>%
    dplyr::left_join(r2dii.data::sector_classifications, by = merge_by)

  by_cols <- names(matched)[names(matched) %in% names(loanbook_with_sectors)]

  dplyr::left_join(loanbook_with_sectors, matched, by = by_cols) %>%
    dplyr::mutate(
      loan_size_outstanding = as.numeric(loan_size_outstanding),
      loan_size_credit_limit = as.numeric(loan_size_credit_limit),
      matched = dplyr::case_when(
        score == 1   ~ "Matched",
        is.na(score) ~ "Not Matched",
        TRUE         ~ "Not Mached"
      ),
      sector = dplyr::case_when(
        borderline == TRUE & matched == "Not Matched" ~ "not in scope",
        TRUE ~ sector
      )
    )
}

library(dplyr)
library(ggplot2)
library(r2dii.data)
library(r2dii.match)

loanbook <- loanbook_demo %>%
  mutate(
    name_ultimate_parent =
      ifelse(id_loan == "L1", "unmatched company name", name_ultimate_parent),
    sector_classification_direct_loantaker =
      ifelse(id_loan == "L2", "99", sector_classification_direct_loantaker)
  )

matched <- loanbook %>%
  match_name(abcd_demo) %>%
  prioritize()


coverage <- calculate_match_success_rate(matched = matched, loanbook = loanbook)

coverage %>% plot_match_success_rate()

coverage %>%
  filter(sector != "not in scope") %>%
  plot_match_success_rate(matched, loan_size_outstanding)

coverage %>% plot_match_success_rate(sector, .x = sector)

coverage %>%
  mutate(matched = case_when(
    matched == "Matched" & level == "direct_loantaker"      ~ "Matched DL",
    matched == "Matched" & level == "intermediate_parent_1" ~ "Matched IP1",
    matched == "Matched" & level == "ultimate_parent"       ~ "Matched UP",
    matched == "Not Matched"                                ~ "Not Matched",
    TRUE                                                    ~ "Catch unknown"
  )) %>%
  plot_match_success_rate(sector, .x = sector)