Abstract
This paper presents a method to predict occupants individual thermal comfort based on environmental information and occupants' machine operation record. We construct sensor networks with which we can obtain environmental information and operation record. A part of values are chosen from the sensor networks and are associated with the value which needs to be predicted. We utilize a dynamically changing neural network to associate the values for predicting a human's thermal comfort. We conduct an experiment in which the sensor values and the thermal sensation are associated. Our experimental results shows that sensors are chosen properly and used for predicting desired. value.
