<p>Social media users hold positive as well as negative emotions toward each other. Social networks comprising of users and their relationships often take into consideration only the positive associations such as friendship, trust, and liking, thereby ignoring the inherent hostility in form of disagreement, distrust, enmity, etc. However, both positive and negative relationships play pivotal role in structure, complexity, and nature of social networks. Recent times have inculcated the need to study signed social networks as they have a significant amount of information in the form of positive and negative links that can be leveraged for drawing useful insights. To this end, community detection is one of the important analysis tasks which focus on identifying densely connected subgroups of users. This work focuses on studying various embedding techniques that can be applied to signed social networks to achieve efficient representations of these networks. Once the latent feature representation of the users is obtained, we have applied clustering techniques such as <i>K</i>-Means and <i>K</i>-Mode to reveal the inherent community structure. Further, the performance of various embeddings and clustering techniques is investigated on various real-world datasets. To completely ensure our findings, statistical tests are also being employed that confirm not all embedding techniques perform same and the proposed methods are better than the conventional work done in this field.</p>