Friend Recommendation Engine for Facebook Users via Collaborative Filtering

Abstract

Today’s internet consists of an abundant amount of information that makes it difficult for recommendation engines to produce satisfying outputs. This huge stream of unrelated data increases its sparsity, which makes the recommender system’s job more challenging. Facebook’s main recommendation task is to recommend a friendship connection based on the idea that a friend of a friend is also a friend; however, the majority of recommendations using this approach lead to little to no interaction. We propose a model using the matrix factorization technique that leverages interactions between Facebook users and generates a list of friendship connections that are very likely to be interactive. We tested our model using a real dataset with over 33 million interactions between users. The accuracy of the proposed algorithm is measured using the error rate of the predicted number of interactions between possible friends in comparison to the actual values.

A Recommendation Engine Model for Giant Social Media Platforms using a Probabilistic Approach

“…Authors in [27] created a framework to analyze tweets in an effort to prevent the spread of SARS-CoV-2 Omicron variant, highlighting the significance of social media in machine learning development. A black box recommendation model was suggested in [28] utilizing the MF approach [29] and applying it to a real Facebook dataset. In the current research we extend that by introducing a probability methodology that makes use of the PMF technique [10], and we do so while utilizing the same data.…”