Finding Anxiety Based on On-line Communication in Internet

Abstract

Psychological anxiety is threatening people’s health. It is non-trivial to detect anxiety timely for proactive care. With thepopularity of on line media, people are used to sharing their daily activities and interacting with friends on on line media platforms, making it feasible to leverage online on line network data for anxiety detection. In this paper, we find that users anxiety state is closely related to that of his/her friends in on line media, and we employ a large-scale dataset from real-world on line platforms to systematically study the correlation of users’ anxiety states and on line communication. We first define a set of anxiety-related textual, visual, and on line attributes from various aspects, and then propose a novel hybrid model - a factor graph model combined with Convolutional Neural Network to leverage tweet content and on line interaction information for anxiety detection. Experimental results show that the proposed model canimprove the detection performance by 6-9 percent in F1-score. By further analyzing the on line interaction data, we also discover several intriguing phenomena, i.e., the number of on line structures of sparse connections (i.e., with no delta connections) of anxietyed users is around 14 percent higher than that of non-anxietyed users, indicating that the on line structure of anxietyed users’ friends tend to be less connected and less complicated than that of non-anxietyed users.

Index Terms: Anxiety detection, factor graph model, micro-blog, on line media, healthcare, on line interaction