A0163
Title: Towards interpretable and trustworthy network-assisted prediction
Authors: Liza Levina - University of Michigan (United States) [presenting]
Robert Lunde - Washington University in St Louis (United States)
Tiffany Tang - University of Notre Dame (United States)
Ji Zhu - University of Michigan (United States)
Abstract: Machine learning algorithms usually assume that training samples are independent. A network connecting the training samples tends to create dependency, reducing effective sample size but also creating an opportunity to leverage information from neighbors. Multiple prediction methods taking advantage of this opportunity have been developed, augmenting the usual node features with network features and/or neighborhood summaries. However, interpretability and inference are rarely available. Two contributions aiming to bridge this gap are covered. One is a conformal prediction method for network-assisted regression using estimated latent node positions in the network as additional features. We show that the usual conformal prediction offers finite-sample valid prediction intervals in this setting under a joint exchangeability condition and a mild regularity condition on the network statistics. The second contribution is a family of flexible network-assisted models built upon a generalization of random forests (RF+), which both achieve highly-competitive prediction accuracy and can be interpreted through importance measures, both for the features and the network. These tools help broaden the scope and applicability of network-assisted prediction for high-impact problems where interpretability and trustworthiness are essential.
