Introduction:
Have you ever wondered that how do e-commerce sites show us a list of recommendations under: “Frequently bought together” when we are searching for products or when we are adding products to our cart?
The sole answer to this question is “Market Basket Analysis”, which is discussed in detail further.
Market Basket Analysis:
The market basket analysis approach is based on the theory that the customers who buy a certain item or a group of items are likely to buy another specific item or a group of items. For example, a customer who is buying bread from the grocery shop is likely to buy milk and butter as well.
In the business world, market basket analysis thus helps retailers better understand the consumers and their purchase behavior and ultimately serve them the best. It therefore is a key technique used by large retailers in order to determine associations between various items. These associations are determined by association rule mining process.
Association Rule Mining:
Association rules are the rules underlying market basket analysis. These rules are the ones that help to establish a pattern between the range of products.
Let us understand this with the help of an example:
· Assume that there are a total of 1000 customers.
· 100 of them bought a notebook, 80 bought a pen and 60 bought them both.
After receiving this data, we try to form an association between the products by calculating the following:
Support:
Support = probability of buying both notebook and pen
= P(notebook and pen)
= 60 / 1000
= 0.06
Confidence:
Confidence = support / probability of buying a pen
= 0.06 / 0.08
= 0.75
Lift:
Lift = confidence / probability of buying a notebook
= 0.75 / 0.10
= 7.5
When,
· Lift > 1, we say that a positive correlation exists and the customers who will buy the first product are likely to buy the second also.
· Lift < 1, we say that a negative correlation exists and the customers who will buy the first product will not buy the second along with it.
Lift = 1, we say that there exists no correlation between the products.
Author: Aditi Langar
Great work. Very useful data which mention above.