PDA

View Full Version : Market Basket Tutorial



scmjea
12-18-2007, 06:07 AM
Hi All,

I'm just starting to feel my way with Weka and am looking for a good tutorial on how to do a market basket analysis. Does anyone have any links or advice. I'm planning on spending my Christmas hacking away :).

I'm thinking of buying this book from the Weka guys http://www.amazon.co.uk/gp/product/0120884070/ref=sib_rdr_dp
but I'm not sure it has enough practical examples for me has anyone read it?

Thanks in advance

John

UK

lucsorel
12-18-2007, 06:14 PM
Hi John,

Hi All,

I'm just starting to feel my way with Weka and am looking for a good tutorial on how to do a market basket analysis. Does anyone have any links or advice.
Well I'm not very familiar with market basket analysis. But it has to do with association rules (or maybe with decision trees as well), doesn't it ?

However, concerning Weka, i would suggest you these links :

the Weka documentation mainpage (http://www.cs.waikato.ac.nz/%7Eml/weka/index_documentation.html)
WekaWik (http://weka.sourceforge.net/wiki/)i: a general Wiki containing code snippets, HOWTOs, etc.
WekaDoc Wiki (http://weka.sourceforge.net/wekadoc/): a documentation project for Weka and its different versions

I'm thinking of buying this book from the Weka guys http://www.amazon.co.uk/gp/product/0120884070/ref=sib_rdr_dp
but I'm not sure it has enough practical examples for me has anyone read it?
I asked my research unit to buy it, and I was satisfied with it because it gives basics and fundamentals for all the data mining approaches covered by Weka, data set samples and methods to mine them according to different purposes.

There are around 15 pages specifically about "association rules", but although there is a 2-pages introduction to market basket analysis, I haven't found a specific example about it. It also depends on the kind of examples you're looking for:

application screenshots?
Java code snipplets?
interpretation of Weka results?You'll find all these kinds of example in this book, but mainly illustrations of datasets, problematics, data processing. I mean that the illustrations mainly highlights the concepts and the approaches (which was great for me since I'm not a hardcore data-miner yet :rolleyes:).

But there are a few screenshots of the application. However, they are based on the 3.4.x Weka branch (which is a freeze done when this book was written), but I would recommend you to use the 3.5.x branch since many functionalities were implemented, which make Weka even more ergonomic or powerful.

I hope this helps!
Cordially, Luc

scmjea
12-18-2007, 06:51 PM
Hi Luc,

Thanks very much for your reply,it's useful to have someone else's experiences. You're correct, market basket is about association rules - I'd hoped to find both Java Source code and interpretations of results. I think I'll get the book and work from there.

Merci beaucoup, vous etes tres gentile.

John

Taqua
12-18-2007, 07:20 PM
Well, the original paper "Fast Algorithms for Mining Association Rules" is in itself a very understandable description of the algorithm. It even contains a pen-and-paper example.

http://www.cs.wisc.edu/~nil/764/DADS/39_agrafa94.pdf

For WEKA I found it the easiest approach to start with the GUI tools and then to convert the configurations you got there into some Java-Code. The GUI itself is just a thin wrapper around the Java-components, but it helps a lot when trying to find the correct parameters and values. The Agrawal algorithm is known as Apriori-Algorithm and thats exactly the name of the implementation in Weka (class weka.associations.Apriori).

Have fun,
said Thomas