O. Roderick and I. Safro, "Polynomial Interpolation for Predicting Decisions and Recovering Missing Data," Preprint ANL/MCS-P1586-0209, February 2009. [pdf]
In this work we improve the existing tools for the recovery and prediction of human decisions based on multiple factors. We use a latent factor method and obtain the decision-influencing factors from the observed correlations in the statistical information by principal factor identification based on singular value decomposition (SVD). We generalize on widely used linear representations of decision-making functions by using adaptive high-order polynomial interpolation and applying iterative and adaptive post processing to obtain an estimated probability of every possible outcome of a decision. The novelty of the method consists in the use of flexible, nonlinear predictive functions and in the post processing procedure. Our experiments show that the approach is competitive with other SVD-based prediction methods and that the precision grows with the increase in the order of the polynomial basis. In particular, the method may be successfully applied when the objective is to get a high number of precisely exact predictions.