In this paper we present a methodology to map individual occupants' thermal preference votes and indoor environmental variables into personalized preference models. Our modeling approach includes a new Bayesian classification and inference algorithm …