I have trained a **classification model** that predicts (returns a score) on how likely a customer is to buy from our company.
This model **does not contain any product specific data**, but we nevertheless apply it when determining which are the most likely customers to buy a certain product, e.g. product A (of course the results we get are the same if we use the model for product A,B,C...).
Additionally, we have performed **Association Analysis** on the Customer/Products set and have come up with certain rules (e.g. customer buying product C is more likely to buy A) and have computed the respective *lifts* for these rules.
**Is there a way to use the information derived in the Association Analysis to systematically "enhance" the use of our classification model?**
An example to illustrate this would be the following.
Suppose that, according to our Classification Model, these are the most likely customers to buy product A:
- Customer X (score 9.0)
- Customer Y (score 9.0)
- Customer Z (score 8.0)
Additionally, suppose that we have the following rule with a lift of 1.2: B => A.
If customer Y has bought product B (but X hasn't), it would be logical to increase customer's Y score.
Guilherme Costa
·
2020-10-01T21:13:15Z (about 4 years ago)