Private Graph Sampling for Association Rules Mining
Recommender systems are heavily used in many online services (e.g., suggesting friends in Facebook). Association rules mining (ARM) is one of the techniques used to build recommender systems. Building the rules based on a huge amount of data (e.g., social networks) is costly in terms of time and computation. Sampling the data is a proposed solution. User data can contain sensitive information, therefore, the sampling must be applied in a privacy-preserving manner.
In this thesis, we aim to:
· Study different graph sampling approaches
· Assess the approaches in terms of (a) quality of ARM, and (b) privacy
· Propose adapted/new approach