Depth-based classification for relational data with multiple attributes

With the recent progress of data acquisition technology, classification exhibiting relational dependence, from online social interactions to multi-omics studies linkage electronic health records, continues gain an ever increasing attention. By introducing a robust and inherently geometric concept depth we propose new type geometrically-enhanced method for that are in form complex network with multiple node attributes. Starting logistic regression describe relationship between class labels attributes, key approach is based on modeling link probability any two nodes as function their depths within respective classes. The approximate prediction rule then obtained according posterior labels. Integrating into process allows us better capture underlying geometry and, result, enhance its finite sample performance. We derive asymptotic properties validate via extensive simulations. proposed illustrated application user analysis one largest Chinese media platforms, Sina Weibo.