Hearst Challenge : A Framework to Learn Magazine Sales Volume

This project is based on a 25000$ competition by the Hearst Magazines posted on www.kdnuggets.com . The goal of this project is to predict the sales of 10 different magazine titles published by the Hearst Magazines at each newsstand location across United States, to optimize the overall contribution of the newsstand locations. To predict the magazines’ sales volume, Hearst magazines has provided a database of the sales of 10 different magazines for almost 40390 newsstands across the US for the period of 2006−2009. The data set includes 10000000 training examples. In addition they have provided different statistical demographical information from the different sources. We need to find a learning algorithm which can use these information to learn the pattern of each title’s sales based on newsstand statistical demographic and time frame. There are 5 different category of Information provided by the Hearst Magazines as follows: