Reshaping with Melt

In the previous slide decks, we discussed how pandas provides 2 functions for reshaping data; .melt() and .pivot() .

We are going to spend this next section discussing .melt(), which is simply the reverse transformation of .pivot().

Instead of converting a long dataframe to a wider one as we do in .pivot(), we do the opposite and convert a wide dataframe into a longer one.

cereal

	name	mfr	calories	protein
0	Special K	K	110	6
1	Apple Jacks	K	110	2
2	Raisin Bran	K	120	3
3	Cheerios	G	110	6
4	Wheaties	G	100	3

Let’s take a look at a subset of our the cereal data we’ve been working with.

We can see there are 5 rows and 2 numerical columns; calories and protein.

Perhaps we would prefer if this was transformed into a long dataframe.

How would we do it?

We can use melt!

Melt

melted_cereal  = (cereal.melt(id_vars=['name', 'mfr'] , 
                              value_vars=['calories', 'protein'], 
                              var_name='nutrition', 
                              value_name='value')
                  )
melted_cereal

	name	mfr	nutrition	value
0	Special K	K	calories	110
1	Apple Jacks	K	calories	110
2	Raisin Bran	K	calories	120
...	...	...	...	...
7	Raisin Bran	K	protein	3
8	Cheerios	G	protein	6
9	Wheaties	G	protein	3

10 rows × 4 columns

Let’s attempt to melt the calories and protein columns into a single one named nutrition with the values expressed in a column named value.

This is the dataframe exactly like we started with when we used .pivot() in the last section.

After melting, we can see that the rows are not quite in the same order as before, but we can agree that its what we want.

melted_cereal  = (cereal.melt(id_vars=['name', 'mfr'], 
                              value_vars=['calories', 'protein'], 
                              var_name='nutrition', 
                              value_name='value')
                  )

Let’s try to understand what happened.

• The identifying column and the columns we wanted to keep are specified in the id_vars argument.

• the columns calories and protein are called in value_vars and melted down into a single column named nutrition.

• We named the new column using the argument var_name.

• The calorie and protein measurements are housed in the new column named value that we name using the value_name argument.

Although this verb takes 4 arguments and pivot only takes 3, we do not need to reset our index after melting.

Let’s apply what we learned!