JUST-INTIME MODELING FOR FUNCTION PREDICTION AND ITS APPLICATIONS

This paper introduces a variant of k Bipartite Neighbors (k-BN), called k-BN2, for use in function prediction. Like k-BN, k-BN2 selects k instances surrounding the query, i.e., the novel instance, and keeps them bipartitely. However, in order to improve the prediction precision, based on the bipartite neighborhood, k-BN2 combines local linear models and the global nonlinear model to predict the value of the novel instance. Applied to two real measured datasets, k-BN2 outperforms the typical k-BN and those methods in which kBN or a related approximate physical model alone is exploited.