Mining input data

Legacy User

Hi,
I'm trying to know whether RapidMiner is suited for our needs. The point is that we are developing a web application that would like to mine data from a MySQL database. However, we need mining output to be reconsidered when things change (something like knowing the user preferences in a web-market). How can I do it with RapidMiner?

Thanks in advance

steffen

Hello and welcome to RapidMiner

Maybe this can help you: http://nemoz.org/joomla/content/blogcategory/21/53/lang,en/

greetings

Steffen

Legacy User

Thank you for the link. It really opened my eyes on how I can integrate RapidMiner with mu web application.
However, I still have a question: I have a list of products, a list of users and the products users bought. When a user enters in my webpage I would like to show him or her the products I think he or she would buy. How can I do this with RapidMiner?

Thanks again

IngoRM

Hi,

When a user enters in my webpage I would like to show him or her the products I think he or she would buy. How can I do this with RapidMiner?

So, that's the (more interesting, at least for me) analysis part. Usually, we are getting money from our customers for this type of information (hey - at least you also want to sell more via cross-selling and up-selling.... we just want to participate from that if you do this with our software  . And by the way: it is known that this type of recommendations increases sales about 20% so we clearly see why one should integrate something like this into his web application).

However, here is a starting point: you can either define this problem as a classification problem (label: a certain product was bought or not) or as a association rule mining problem (products frequently bought together). You can find examples for these tasks in the "sample" directory RapidMiner. The main questions are: how should the data be represented - which preprocessing should be performed - which are the best modelling schemes and their parameters - how should I evaluate my models.

RapidMiner supports everything for these analysis processes but as a generic data analysis software it is not an "out-of-the-box"-solution for every specific data mining problem you can think of. The good news, on the other hand, is: due to its generality and flexibility you can also hardly think of an analysis problem which cannot be solved (or at least you can extend RapidMiner so that is solved then). However, you still have to define the optimal processes with optimal settings and this will usually also lead to optimal results. Sorry, but there are only two options: learning all this yourself which might be applicable if you have enough time - or relying on others helping you setting up all those processes.

Cheers,
Ingo