ABSTRAC TLink prediction is a key technique in many applications in social networks; where it is necessary to foresee potential connections between entities. Typical link prediction techniques involve uniform entities, such as company-to-company, applicant-to-applicant links, or non-reciprocal relationships, such as company-to-applicant links. However, there is a challenging problem related to predicting links between composite entities and mutual links; such as accurately predicting matches on corporate datasets, jobs, or workers on job websites, where links are mutually determined by both entities that the composite entity belongs to disjoint groups. The causes of interactions in these domains make the prediction of composite and reciprocal links significantly different from the typical version of the problem. This work addresses these problems by proposing the Support Vector Machine model. By implementing the proposed algorithm, the accuracy is expected to increase in the link prediction problem. Keywords: link prediction, potential links, composite, reciprocal links, Support Vector Machine. INTRODUCTION A social networking service is an online service, platform or site that aims to facilitate the building of social networks or social relationships between people who, for example, share importance, activities, backgrounds or communications in real life. A social network service consists of a representation of each user, their social connections, and a variety of new services. It is used to model the interaction between communities on social networks. Where graphs are used to represent interactions between those communities, where nodes represent people in some communities and links represent the association between those people. Understanding the association between two specific nodes by predicting the probability of a future but not currently The association between them is a fundamental problem known as link prediction. Interaction on the social network involves both positive and negative relationships, for example, as attempts to establish a relationship may fail due to decline from the expected goal. This generates links that indicate declining invitations, disapproving of applications, or expressing disagreement with the opinions of others. Such social networks are reciprocal because the sign of a link that indicates whether it is positive or negative depends on the attitudes or beliefs of both entities forming the link. Furthermore, positive and negative reciprocal relationships have been even less studied. Fig.1: Collaborative information for predicting composite and reciprocal links Interaction Predicted preference Predicted match Recently, social network analysis has had a variety of applications, such as online dating sites, education admission portals as well as places of work, employment, career and hiring places, where people in networks have different roles and connections between them can only between people with different roles.