State of the art about GP applied to rule systems

[unintendedbear]

To whom it may want to do it

[JJ]

I am afraid this might be a bit too wide. There might be thousands of papers about it. I'll take a look anyway...

[unintendedbear]

Let's make a survey then :D

[JJ]

I think it's better to make the objectives clear, and also a few initial results, if possible...

[JJ]

There are some papers such as this one http://link.springer.com/article/10.1007/s11257-004-7961-2#/page-1 or this one http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1299832&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D1299832 In general, we are talking about "web usage mining" and genetic programming. I would say there is a whole bunch of them. If we're to push this forward, we have to find a way to make it one better.

[JJ]

This one too http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5359726&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5359726

In general, I would say our angle should be that we are looking for a novel application, authorization of usage or not.

[unintendedbear]

I don't agree with the statement that we are talking about "web usage mining", as we may have lots of different events, from opening a web browser to blocking the screen... Why is this web usage mining?

[JJ]

because we are dealing with web logs, are we not? If it's wider, then I'm wrong, but anyway the initial research on user activity mining was done with web logs.

[unintendedbear]

The initial research, yes, for this one I wanted to use MUSES data :(

[JJ]

OK. Maybe worth mentioning it, anyway. The general concept is the same: you have a resource with is identified uniquely and you allow or deny. In the case of the web it's URL in this case, well, whatever you use to identify the resource or asset.

[zeinebchelly]

We can also mention these if we want to talk in general about GP for rules extraction:

1. http://www.sciencedirect.com/science/article/pii/S1568494601000242
2. http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1004431&url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F7875%2F21687%2F01004431
3. http://link.springer.com/chapter/10.1007%2F3-540-44794-6_26

[unintendedbear]

Hi all, I've been reviewing these two:

• http://download.springer.com/static/pdf/887/chp%253A10.1007%252F978-3-642-17432-2_25.pdf?originUrl=http%3A%2F%2Flink.springer.com%2Fchapter%2F10.1007%2F978-3-642-17432-2_25&token2=exp=1461658558~acl=%2Fstatic%2Fpdf%2F887%2Fchp%25253A10.1007%25252F978-3-642-17432-2_25.pdf%3ForiginUrl%3Dhttp%253A%252F%252Flink.springer.com%252Fchapter%252F10.1007%252F978-3-642-17432-2_25*~hmac=a0ab7aa8ee0bff752a53eeb4c9715034175c8e9b3d0a041070123c7d821f4554
• http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=6029340

(The second it's an extension of the first one) Basically, my way of calculating the fitness is the worst possibe (yay! wrong again :D I'm aweso......nope), and the best (not always, but mostly) is Wilcoxon-Mann-Whitney which approximates the AUC. What do you think? Can somebody explain a bit about ROC ans AUC? :(