Friedman Frobenius Distributed Recommender Using Multi-Agent For Social Networks

Abstract

A social network has been a notable mechanism owing to the wide and convenient information stored in it. The social network is spread across many areas, it has become a  storehouse of millions of different applications like health oriented surveys, marketing service preferences and many more. Social network recommender model make use of this information and administer users in hand picking their possibilities via multi-agent platform. It is logical that a recommender models must be well organized to be adequately handle the enormous extent of data that has been originated in current years by social network users. In this work we propose an adaptable recommender system which will be able to process big data in a time efficient manner. To be apt to assess these models software agent-based framework is developed. This framework consist of three modules that is tweet collection agent, data mining agent and distributed recommender agent that utilize statistical test and Frobenius Matrix Collaborative Recommender algorithms. Extensive experiments on Sentiment140 dataset are conducted and the results are evaluated against the existing recommendation models. After the experiment the result shows that the proposed F-FMCR method achieves better performance compared to other methods, in recommendation accuracy, recommendation time and mean absolute error.