Predictive Web Analytics
Discovery challenge co-located with ECML/PKDD 2014
The workshop took place on Friday, September 19th, 2014, in Nancy, France. All papers are freely available for download in the proceedings.
Welcome! Studying the behavior of visitors on web sites is extremely important: the number of visitors and social media reactions to a web site are indirect measures of its influence, and in the case of commercial operations, typically translate into income and profits.
Predictive tasks on the web are an active research topic (see references). This includes, for instance, predicting the number of visits an article will obtain using signals from early visits to it. Until now, such tasks have done on proprietary data because behavioral data is delicate from the point of view of privacy, and it is also a highly sensitive business secret for most large-scale web sites. This has made it difficult to compare and evaluate different approaches.
This predictive challenge will allow for the first time to compare predictive web analytic systems on a shared dataset »
Important dates:
Read the task description and download training datasets.
Read how to participate.