Online Social Network (OSNs) users’ profile information is important in any big data research, but the information on gender is often undisclosed. One way to identify the gender of OSNs’ users is through their username. However, this method has a disadvantage if OSNs’ users are falsifying their identities and using nicknames instead. This study is a survey from the literature which extracted and analysed
related works to correlate both gender detection and stress detection of OSNs. Studies related to gender and stress detection were conducted separately by previous researchers. Nevertheless, the studies primarily focused on the implementation of data sources from OSNs and analysis methods using Machine Learning. Therefore, the basic design of this study focuses on the method of machine learning algorithm, sources of OSNs, and types of language. Major findings revealed that most previous
researchers implemented Natural Language Processing, traditional Machine Learning and Deep Learning in gender and stress detection. This study found some challenges which include length of post, varieties of language usage and style of writing related to gender, and correlation between gender and use of words. Therefore, determination of gender detection of OSNs’ users experiencing stress has great potential for in-depth exploration in the future.

Keywords: Gender Detection; Stress; Machine Learning; Deep Learning