A variety of approaches have been recently proposed to automatically infer users' personality from their user generated content in social media. Approaches differ in terms of the machine learning algorithms and the feature sets used, type of utilized footprint, and the social media environment used to collect the data. In this paper, we perform a comparative analysis of state-of-the-art computational personality recognition methods on a varied set of social media ground truth data from Facebook, Twitter and YouTube. We answer three questions: (1) Should personality prediction be treated as a multi-label prediction task (i.e., all personality traits of a given user are predicted at once), or should each trait be identified separately? (2) Which predictive features work well across different online environments? and (3) What is the decay in accuracy when porting models trained in one social media environment to another?

Research in psychology has suggested that behavior of individuals can be explained to a great extent by their underlying personality traits. In this paper, we focus on predicting how the personality of YouTube video bloggers is perceived by their viewers. Our approach to personality recognition is multimodal in the sense that we use audio-video features, as well as textual (emotional and linguistic) features extracted from the transcripts of vlogs. Based on these features, we predict the extent to which the video blogger is perceived to exhibit each of the traits of the Big Five personality model. In addition, we explore 5 multivariate regression techniques and contrast them with a single target approach for predicting personality impression scores. All 6 algorithms are able to outperform the average baseline model for all 5 personality traits on a dataset of 404 YouTube videos. This is interesting because previously published methods for the same dataset show an improvement over the baseline for the majority of personality traits, but not for all simultaneously.