data-raw/fifa.R

In comeetie/greed: Clustering and Model Selection with the Integrated Classification Likelihood

library(readr)
library(dplyr)
library(stringr)
# data from https://www.kaggle.com/stefanoleone992/fifa-20-complete-player-dataset?select=players_20.csv
fifa <- read_csv("./data-raw/players_20.csv")


positions <- fifa$player_positions %>%
  str_remove_all(" ") %>%
  str_split(",") %>%
  unlist() %>%
  unique()
fifa <- fifa %>% mutate(player_positions_vec = str_split(str_remove_all(player_positions, " "), ","))
create_pos_feature <- function(data, position) {
  position_var <- enquo(position)
  data %>% mutate(!!position := factor(sapply(player_positions_vec, \(pvec){
    position %in% pvec
  })))
}


for (p in positions) {
  fifa <- create_pos_feature(fifa, p)
}



Fifa <- fifa %>%
  select(short_name, nationality, preferred_foot, age, height_cm, weight_kg, value_eur, pace, shooting, passing, dribbling, defending, physic, !!!syms(positions)) %>%
  filter(!is.na(pace)) %>%
  mutate(preferred_foot = factor(preferred_foot))


poscolnames <- colnames(Fifa)[14:28]

pos = read_delim("./data-raw/fifaposition.csv",delim=",")

pos <- pos[match(poscolnames, pos$pos), ]

Xp <- matrix(rep(pos$x, each = nrow(Fifa)), nrow(Fifa))
Yp <- matrix(rep(pos$y, each = nrow(Fifa)), nrow(Fifa))
Fposbin <- as.matrix(Fifa[, toupper(poscolnames)] %>% mutate_all(\(f){
  if_else(f == TRUE, 1, 0)
}))

mean_pos_x <- rowSums(Fposbin * Xp) / rowSums(Fposbin)
mean_pos_y <- rowSums(Fposbin * Yp) / rowSums(Fposbin)
Fifa$pos_y <- mean_pos_y
Fifa$pos_x <- mean_pos_x