calculation of geographical distances vs jako-winkler distance

jw_vs_geo.R

pacman::p_load(tidyverse,jsonlite,stringdist)
res <- read_json("https://data.gov.il/api/3/action/datastore_search?resource_id=70ba1705-3b25-416f-939c-985999f87f35&limit=10000")
df <- res$result$records %>% map_df(~.x) %>% 
  filter(`סוג ישוב`!="מוקד תעסוקה") %>% 
  mutate(cntr = row_number())
dist_geo <- df %>% 
  select(`אורדינטה מזרחית`,`אורדינטה צפונית`) %>% 
  dist() %>% as.matrix()
res1 <- map(df$`שם ישוב`,~stringdist(.x,df$`שם ישוב`,  method='jw'))
matrix_from_list <- do.call(cbind, res1)
df_geo_dist <- dist_geo %>% 
  as.data.frame() %>% 
  rownames_to_column("from") %>% 
  gather(to,geo_dist,-from)
df_word_dist <- matrix_from_list %>% 
  as.data.frame() %>% 
  setNames(1:nrow(df)) %>% 
  rownames_to_column("from") %>% 
  gather(to,word_dist,-from)
df_geo_dist %>% 
  left_join(df_word_dist,join_by(from,to)) %>% 
  mutate(word_dist_x = (word_dist- min(word_dist))/(max(word_dist) - min(word_dist)),
         geo_dist_x = (geo_dist- min(geo_dist))/(max(geo_dist) - min(geo_dist)),
         dist = word_dist_x^2 + (geo_dist_x)^2) %>% 
  mutate_at(vars(from,to), as.integer) %>% 
  filter(from != to,geo_dist < 10000) %>%  
  left_join(df %>% select(cntr,`שם ישוב`),join_by(from == cntr)) %>% 
  left_join(df %>% select(cntr,`שם ישוב`),join_by(to == cntr)) %>% 
  select(-from,-to,-word_dist_x,-geo_dist_x) %>% 
  arrange(dist) %>% View()