Hierarchical Bayesian modeling for predicting ordinal responses of personalized thermal sensation: Application to outdoor thermal sensation data

Publication date: September 2018
Source:Building and Environment, Volume 142
Author(s): Jongyeon Lim, Yasunori Akashi, Doosam Song, Hyokeun Hwang, Yasuhiro Kuwahara, Shinji Yamamura, Naoki Yoshimoto, Kazuo Itahashi
A concept known as ‘nudge’ has recently received attention in many application domains. It implies influencing the behavior and decision-making of individuals by making indirect suggestions through the presentation of adequate information. We apply such a perspective to improve the value of a space. It can be measured by the number of visitors, and the predicted thermal sensation is considered as information offered to potential visitors. In the present study, we explain how to generate the information required for a successful nudge. This information must be specifically tailored towards personalized characteristics, rather than a one-fits-all approach. This study presents a new data-driven method for predicting individuals' thermal sensation by formulating the effect of both measured (thermal) and non-measured factors on thermal sensation votes. The proposed model is explicitly encoded based on a major premise that “different individuals have different thermal sensation characteristics; however, all individuals also have a common trend.” The inference model uses a Bayesian approach, and is hierarchically structured to represent dependencies across model parameters of the personalized characteristics of individual-level and the typical trend of group-level thermal sensations. The Markov chain Monte Carlo approach is used to approximate the posterior distribution and draw inferences on the model parameters. The results, based on data collected from outdoor spaces, show that the proposed model provides accurate predictions for personalized thermal sensation and improves the efficiency of parameter estimates. Our approach provides fresh insight into statistical models for predicting thermal sensation.