R解析具有JSON数组的dataframe列，并转换为one-hot编码

Question

我有一个dataframe，它的列具有字符串形式的JSON数组。我的目标是解析列并将其转换为one-hot编码，但在解析JSON时遇到错误。

library(jsonlite)
> df <- data_frame(Amenities=c("[\"Parking\", \"Lawn\", \"Garage\", \"Frontyard\"]", "[\"Parking\", \"Lawn\", \"Garage\", \"Backyard\"]", "[\"Parking\", \"Lawn\", \"Garage\"]"))
> df
# A tibble: 3 x 1
  Amenities                                           
  <chr>                                               
1 "[\"Parking\", \"Lawn\", \"Garage\", \"Frontyard\"]"
2 "[\"Parking\", \"Lawn\", \"Garage\", \"Backyard\"]" 
3 "[\"Parking\", \"Lawn\", \"Garage\"]"               
> df <- df %>% mutate(Amenities=fromJSON(Amenities))
Error: parse error: trailing garbage
          awn", "Garage", "Frontyard"] ["Parking", "Lawn", "Garage", "
                     (right here) ------^
> 

预期输出：

Parking  Lawn  Garage  Frontyard  Backyard
      1     1       1          1         0
      1     1       1          0         1
      1     1       1          0         0

解决方案:同时保留现有的数据帧。

library(qdapTools)
df <- cbind(df, +(mtabulate(str_extract_all(df$amenities, "\\w+( +\\w+)*"))))

Answer 1

我们可以使用mtabulate在一行代码中完成这项工作

library(qdapTools)
library(stringr)
mtabulate(str_extract_all(df$Amenities, "\\w+"))

-output

#  Backyard Frontyard Garage Lawn Parking
#1        0         1      1    1       1
#2        1         0      1    1       1
#3        0         0      1    1       1

Answer 2

您可以将json视为字符串，清理它们并展开dataset。

library(dplyr)

df %>%
  mutate(Amenities = gsub('\\[|\\]|"', '', Amenities)) %>%
  splitstackshape::cSplit_e("Amenities", sep = ',\\s*', 
                            type = 'character', fill = 0, fixed = FALSE) %>%
  rename_with(~sub('Amenities_', '', .))

#                         Amenities Backyard Frontyard Garage Lawn Parking
#1 Parking, Lawn, Garage, Frontyard        0         1      1    1       1
#2  Parking, Lawn, Garage, Backyard        1         0      1    1       1
#3            Parking, Lawn, Garage        0         0      1    1       1