Integrating Adaptive Moving Window and Just-in-Time Learning Paradigms for Soft-Sensor Design

•Novel adaptive algorithm stacking moving window (MW) and just-in-time learning (JITL) models for continuous processes.•MW size adjusted using statistical monitoring of Mahalanobis distance and variance of model errors.•Relevant data for JITL models selected on the basis of similarity in predictor space and recent prediction performance.•Accurate and robust predictions obtained for various real datasets and a wide range of simulated concept drifts.