SUPRA: a SUrfing Prediction fRAmework Files

Data 
--------------- 

We provide four datasets as benchmarks for testing the performance of our framework. All of them are fully anonymized, since each user and site (i.e., an object of class SimulationUnit) is actually represented as a series of page requests (i.e., objects of class Transaction). 

The first two datasets are suitable for the client-side revisitation prediction problem. Their technical characteristics are the following:

1) The HtUsers dataset comprises the navigational activity of 25 users, containing 137 thousand page requests in total. You can fetch it from here: http://www.l3s.de/~papadakis/revisitationPrediction/HtUsers.tar.gz.

2) The WhrUsers dataset comprises the navigational activity of 180 users, containing 2.47 million page requests in total. You can fetch it from here: http://www.l3s.de/~papadakis/revisitationPrediction/WhrUsers.tar.gz.

The other two datasets datasets are suitable for the server-side revisitation prediction problem. Their technical characteristics are the following:

1) The Wiki dataset comprises 35,223 page requests from 1,742 different IP addresses. You can fetch it from here: http://www.l3s.de/~papadakis/revisitationPrediction/Wiki.

2) The CMS dataset comprises 359,211 page requests from 24,614 different IP addresses. You can fetch it from here: http://www.l3s.de/~papadakis/revisitationPrediction/CMS.

You may freely use these datasets and code for research purposes, provided that you acknowledge the authors with the following reference:

George Papadakis, Ricardo Kawase, Eelco Herder. "Client- and Server-side Revisitation Prediction with SUPRA". In the 2nd International Conference on Web Intelligence, Mining and Semantics (WIMS), 2012.

The paper is available here: http://www.l3s.de/web/page25g.do?kcond12g.att1=1874&sp=page15g.