I want to add a column to my dataframe (with repetitive values) which contains the name of previous group (float and sortable). any idea?

Before:

After:

CodePudding user response：

you can use groupby with shift which shifts the Group 1 place for each group:

df['Previous Group'] = df.groupby("value")['Group'].shift()

print(df)

   Group   value  Previous Group
0    1.1  value1             NaN
1    1.1  value2             NaN
2    1.2  value1             1.1
3    1.2  value2             1.1

EDIT:

We can use a helper column with groupby ngroup and then use it as a mapping column and filter out rank = 1:

a = df.assign(helper=df.groupby(["Group","value"]).ngroup())
rnk = a.groupby("Group")['helper'].rank(method='dense')
df['Previous Group'] = (a['helper'].sub(1).map(dict(zip(a['helper'],a['value'])))
                         .mask(rnk.eq(1)))

print(df)

   Group   value Previous Group
0    1.1  value1            NaN
1    1.1  value2         value1
2    1.1  value3         value2
3    1.1  value3         value2
4    1.2  value4            NaN
5    1.2  value4            NaN
6    1.2  value5         value4