Multimedia blog volume prediction using adaptive neuro fuzzy inference system and evolutionary algorithms

Abstract

Due to wide streaming multimedia blogs over the social networks, volume prediction has become indispensable for the analysis of blog popularity. As a rule base driven method, Adaptive Neuro Fuzzy Inference System has gained popularity in various prediction tasks for its efficiency and ease of implementation. In this paper, two modified Adaptive Neuro Fuzzy Inference System models have been proposed by tuning its premise and consequent parameters using (a) Particle swarm optimization and (b) Genetic algorithms, to improve its predictive performance. Particle Swarm Optimization helps in reducing the training and cross validation error of the predictive model whereas Genetic Algorithms optimize minimum clustering radius which aids in the formation of rule base. Comparative analysis of proposed method has been performed against Neural Networks, Support Vector Machines and basic Adaptive Neuro-Fuzzy Inference System. Both of the proposed variants have outperformed state-of-art techniques using Genetic algorithms and Particle swarm optimization when tested on UCI public dataset and real dataset of Twitter, making it well suitable for multimedia blog volume forecasting.