Avoiding for loops is using pandas

[rajanikant-ghate]

Consider the following example of an e-commerce data:

Cust ID	Order ID	Item category	Amount	Order Date
A	xyz	Electronics	30	20220726
B	abc	Home appliance	40	20220725

A ML problem statement requires, to identify “UserChurn” prediction. The business team feels has proposed following features:

1. The last order category placed
2. The number of orders placed in the last week
3. The amount spent in the last week

Note here that, it’s logically quite convenient to start with:

for i in df[‘Cust ID’].unique:

However, the performance of such a code will be sub-optimal. Each of the above three features can be calculated by a combination of group by and aggregation. If not, in one line, as separate lines, grouped by Cust ID, as separate tables, which then can be merged. This will still be faster than a for loop.

Would any learner like to volunteer to submit the answer?