How to implement very simple model

[deleted_user]

I'm currently trying to find out more about how to actually start using a data mining model that I have built in Enterprise Miner. I understand how to build models based on historical data, what I'm really interested in though is how to use these models to predict behaviour based on information given.

I'm working with an incredibly simplistic example to start with. Say I have an online store that sells computer games, jewelery, cars and houses. Based on historic data I have built a decision tree that shows that certain ages and genders will likely purchase particular things.

For example, male users between 15 and 25 will more likely buy computer games, while females users between 39 and 50 will be jewellery.

What I was trying to find out was what process you need to go through to implement this on a website so that it reads the user's age and gender from their profile when they log in, and display the product that best suits them.

This is probably more complicated then I am imagining, but I am unable to find any information on the process anywhere. All the information I've seen relates the building the model, not using it.

Hopefully my explanation of what I'm looking for isn't too convoluted and confusing. Please let me know if it is and I'll try and explain it better.

Thanks

Steve

[SusieQ324]

Steve,

It should be as simple as taking the decision tree ruleset that you ended up with and coding that on your web page. EM should even give you the resulting scoring code in PMML that you could load directly to your webpage...theoretically that is. I've not specifically done that in practice with this package, but other similar packages have similar functionality. If it's that simple, you could always code up the ruleset by hand.

So, for your example, when a potential customer visits your page, by asking them the questions that were on your tree, and in that order, you should be able to place them into the appropriate product page. So, I'm a female visiting your page....since AGE seems to be the primary differentiator, ask that question first. Let's say I'm 40, then you would pop up the jewelery page.

Does that make sense?

Sue

[deleted_user]

That's one thing I thought and would be simple to do in this case, however, what about models with a lot of levels, or where a regression or neural network is used?

Also, what about having the model keep learning with new information, eg. a 15-25 male looks are cars, so the model may change accordingly.