Matching relational records. Is it possible using Data Minig?

Problem:
I am working on a price comparison system which matches the best prices for a purchase (or an order) from exisiting purchase data.
The order is stored in multiple tables including order details (stores major items purchased: e.g., PC) and order sub-details (optional items purchased with the major items: e.g., speakers, backup device, webcam etc.).
There could be a number of major items in an order and each major item could have multiple related sub items. The other variables that affect the price include trade-ins if any, sales going on at the time of order, number of units etc.

Now, for any new configuration (major items/related sub items), the system should be able to return a list of previous purchases made with similar configurations, and similar variables (quatities, trade-ins etc). Even if the same model is not present, similar pcs by the same vendor should be considered. etc etc.

Questions:
Is this possible using Data mining?
If yes, which algorithm is recommended?

Also, can I assign/modify any kind of weights to certain variables (if same model: .6 ; if same model not available but pcs made by same manufacturer available: .3 ; by other manufacturers: .1)?

Any help will be greatly appreciated.

Thanks,
Jojy

This seems like a reasonable problem for data mining. I would recommend decision trees, neural nets, or logistic regression. There is no way in SS2k5 to weight attributes, however, you can simulate gross weighting in NN and LR by duplicating columns. E.g. if you have a column "model" which you want to weight twice as much as other columns, you duplicate it to "model" and "model1" with the same data inside.