Latent friends mining on social networks : a new friends recommendation system for Facebook

Abstract

Latent friends of a person mean the persons who have the potential to become friends with the target person. Finding latent friends of a user on Online Social Networks (OSN) and recommending to the user is important for growing a user’s network. Users posts things happening to them on OSN to keep their friends updated. It has been studied that the contents posted by Facebook users are expected to reflect their interests. Thus Facebook statuses can be used to identify one’s interests.

In this report, the author proposes a friends recommendation system by exploiting Facebook user’s statuses. The system identifies topics from users’ statuses and classifies them into categories. Users are categorized based on the classifications of their topics. Similarity is calculated using the categorization results and the system recommends friends based on the similarity values.

Experimental results using real Facebook users’ statuses provides the accuracy rate

of the system and indicates the usefulness of the new friends recommendation system.