pandas groupby functions usage and examples

 Introduction:

 

Pandas is one of the most basic data processing libraries data enthusiasts learn and use frequently. We have discussed 10 most basic functions to know from pandas in a previous post. Now, although I have known and used groupby for quite a bit of time now, there are a lot of tricky things and actions around the groupby functions we need to learn, so that one can utilize groupby functions most.

The basics:

Now, if you are new to pandas, let's gloss over the pandas groupby basics first. groupby() is a method to group the data with respect to one or more columns and aggregate some other columns based on that. The normal syntax of using groupby is:
pandas.DataFrame.groupby(columns).aggregate_functions()

For example, you have a credit card transaction data for customers, each transaction for each day. Now, you want to know how much transaction is being done on a day level. Then in such a case, to know the transaction on a day level, you will want to group the data at a day level and then sum up the transactions.

Let's say our hypothetical dataset has the columns customer_id, date, and transaction_value. Now, to see the transaction on a daily value, we have to groupby using customer_id, date and sum the transaction_value. i.e. 

data = data.groupby(['customer_id','date']).sum('transaction_value')

will provide the transaction value summed at a day level. Now that you know how groupby works normally let's see what are the different functions we can use in a groupby, and how each of them work.

aggregate functions:

On using groupby, you can apply a number of different aggregate functions on the column or columns on which you can apply them. Few of them are:

(1) mean(): take the average of all the values

(2) sum(): take the sum of all the values

(3) first(): take the first entry of all the values

(4) last(): take the last entry of all the values

Here is a more comprehensive list of all the groupby aggregate terms; which you can read about. Such aggregate functions follow the simple syntax which is:

dataframe.groupby([columns_to_group_on]).aggregate_function()

here, in place of aggregate_function we can use the aggregate function like mean, sum, first and others. Now, once you do this, the columns you use to group becomes an index. In many cases, you will not want that; as you can't anymore reference those columns from your dataframe as normal columns. This can be solved very easily using reset_index(); as that resets the index as columns. Therefore, the safe syntax is:

df = dataframe.groupby([columns_to_group_on]).aggregate_function().reset_index()

where df is the new, returned grouped dataframe. 

We will discuss more difficult grouping styles in the later part of this blog post tomorrow.