Discovering Influential Users in Online Social Networks Based on Multi-Features

Abstract

As online social contact has become the main means of social communication and cooperation, users, as the basic element of social contact, also present multiple characteristics. How to capture the characteristics of users among mass social data and evaluate their influence has gained the attention of many researchers. This paper, based on the improved co-evolutionary algorithm, conducts collaborative exploration on user characteristics and introduces neural network to optimize the algorithm, to improve the accuracy and fault tolerant ability of the algorithm. Based on the improved co-evolutionary algorithm, this paper conducts exploration and empirical analysis on the influence of multi-feature users, and gets a user rating model with good robustness and high accuracy. And through comparing three different algorithms, it conducts sensitivity analysis, showing that the optimized co-evolutionary algorithm has higher adaptability.