in my company we are working on a upset project in which we are trying to solve the following problem:
What we propose to our customer that he/she may be interested in based on the fact that he/she already bought A,B,C.
We have sales records but we also have many products.
We do not have many customer information (at least not in the dataset I'm working with) so as I first step I put all our sales records in a simple format:
CUST_ID A B C D E F G H I
Where each column is 1 if the customer has bought the item or not (One-Hot_Encoding)
mlxtend.frequent_patterns with Apriori and association_rules but I'm not getting good results.
I get so many rules that are quite useless at the moment.
I would love to predict, given a CUST_ID and the product he/she bought, the probability that he/she would buy the missing product.
Can someone point me in some good direction? Feature Extraction? The initial idea was to build a recommender system but I'm a bit lost at the moment...