[Python] How do I create product-level propensity scores for prospects from customer purchase...

I have two datasets in SQL

• 
  customers – companies that have already bought exactly one of five products (product_1, product_2, product_3, product_4, product_5).
  
  Columns are mostly categorical (SIC_Code, businesssize, Country, County, etc.).
  
  Label column: SoldProduct.
  
  
• 
  prospects – same feature columns, but no SoldProduct.
  

Goal: produce a table where every prospect gets five probability columns—one per product—so the sales team can rank who is most likely to buy each product. I’m new to ML and have seen CatBoost and XGBoost recommended, but I’m open to any library or workflow that is:

• 
  Beginner-friendly (minimal encoding / preprocessing).
  
  
• 
  Works well with mostly categorical data and some missing values.
  
  
• 
  Outputs calibrated probabilities, not just class labels.
  

For this use-case, which Python library or pipeline would you choose?

Continue reading...