SCALABLE MACHINE LEARNING ALGORITHMS FOR A TWITTER FOLLOWEE RECOMMENDER SYSTEM

Recently, machine learning (ML) algorithms have been employed in social networking recommender systems. In this paper, a Twitter recommender system is simulated by a multi-agent system that can be used to provide the users with a list of useful recommendations, specifically a list of users (i.e., followees) that a user is interested in following. The simulator is used to test the scalability of a machine learning algorithm (i.e., Neural Network, Multilayer Perceptron) for data analysis with parallel implementation on multi-node distributed systems. The distributed environment is simulated by a multi-agent modeling. The initial parameters that should be set up on the simulator include the number of nodes, the algorithm employed in the simulated recommender system, and the actual followees and followers information. The experimental results were obtained on three distinct datasets for evaluating the accuracy and the execution time of a simulated recommender system when testing the ML algorithm in different scenarios.