Behavioral analysis from online data using temporal graphs

Abstract

The Internet over and above social media is the basis of human interaction, information exchange, and communication nowadays, which has resulted in prodigious data footprints. If prediction techniques are efficiently employed this data can be put to appropriate utilization for deducing human behavior. We in our work have proposed a methodology for collecting data from social media by assessing the user interactions online, using time-varying attributed or temporal graphs. Initially, we have discussed temporal graphs and how temporal and structural properties of users can be modeled using these evolving graphs for predicting the personality type of the user. The online platforms from where the datasets have been used for the deductions are Stack Exchange and Twitter. Moreover, the secondary research question addressed in this paper is How temporal or time-varying features impact our user behavior prediction. The graphs plotted using the provided datasets show the interactive behavior of users on different platforms.