Rising stars prediction in reviewer network

Abstract

The prediction of rising stars is a challenging task in social networks. To this end, publicly available online social databases are considered to retrieve meaningful information for prediction of future trends. According to our knowledge, we are the first to identify the rising stars in review domain by applying machine leaning methods. More specifically, we predict rising reviewers in yelp review network. Metadata, Rrecency-frequency-activity, and temporal categories are considered, and various features are proposed for each type. In addition, two performance measures are developed to evaluate the rising stars. As a binary classification problem, three popular machine learning models, 10-fold cross validation and two newly designed yelp reviewer’s datasets are used. The proposed framework for rising stars is evaluated using various experiments, such as impact of individual feature, category-wise and model-wise performance. As an outcome, we find that our model demonstrates promising accuracy and f-measure values. In addition, two rankings of top-10 rising reviewers are presented (using weighted score and evolution score), and these rankings are validated against their current status from yelp.com.