The immense growth of social media has shifted online platforms into effective resources for determining and categorizing potential customers. Relational learning (RL) adopts network connectivity and social media content to establish predictions about users. It uses the influences of social network structures and user content to model a supervised learning task. However, current RL methods consider all data instances as significant.
Outlier data greatly affects the overall output, making this a flawed system. Furthermore, these methods overlook social status of users. According to the self-categorization theory, social status influences goals individuals pursue, further affects the way they pursue and the positive social identity they obtain. Delving into social status information on how to strategically apply this information opens an opportunity to advance targeted advertisements. Therefore, there is a need to develop a more robust system that applies social psychological findings to investigate social status for social network users.
Researchers at Arizona State University have developed a system that is capable of identifying the social status of social media users. This system, the relational learning framework based on social status analysis (RESA), extracts robust and intrinsic social dimensions of users. RESA then exploits this information as features for supervised learning. The invention estimates social status of social network users by learning who their friends are. These aspects are applied towards real-world relational learning problems where noise exists. The obtained social status can effectively facilitate online targeted advertising on social media platforms.
- Targeted advertising
- Social status to facilitate relational learning
- Information retrieval
Benefits and Advantages
- Accuracy – Robust framework that is applicable to real-world applications.
- Automation – Determines social status based on links associated with social media user for targeted advertising.
- Performance – Improves quality of responses through self-categorization theory.
For more information about the inventor(s) and their research, please see