README

This repository includes all necessary files to replicate the research as presented in the work, 'Addressing the Cold User Problem for Model-based Recommender Systems', by Tomas Geurts and Flavius Frasincar. The data set used for this research can be classified as a webshop user implicit feedback data set. Interractions are based on user buy or return events where we define a positive interaction between a user and an item (coded by a 1) by 'number of purchases - number of returns > 0' and a negative interation between a user and an item (coded as a 0) by 'number of purchases = number of returns'. There are are also missing interactions (items that have been purchased nor returned by the user). The data set contains over 500,000 unique users and more than 200,000 unique items. The data set furthermore contains approximately 2.5 million interactions, where we have on average 4.55 interactions per user.

Note: the index (first column) provided in the dataset should be ignored.