Recommender System based on Multi Datasets

Recommender systems are one of the tools designed to help users deal with the information explosion by giving information recommendation according to their information needs. The cold-start problem refers to the situations where insufficient initial information of data sources for recommendations to make suggestions to users. When a new user extends into the system initially nothing is known about their preference and this need to be discovered. The process of how to get about it in the quickest and most accurate way is a challenge. This paper were designed in two phases, an off-line phase, where non-redundant multi-level and cross-level association rules and rare-item association rules are built and an on-line phase, where the rules are applied to real situations providing recommendations to customers to solve cold-start problem.