A machine learning approach for lighting perception analysis via crowdsourcing

In this paper, a new analytical scheme designed for crowd sourced subjective lighting evaluation system is proposed. Participants are gathered through crowdsourcing and evaluations are done based on an online system that shows CG (Computer Graphics) on a display. Data about preferable space brightness which is collected by the system is analyzed by machine learning techniques, the Bradley-Terry Mixture (BTM) model and the logistic regression. The results show that applying machine learning techniques enable us to extract important relationships between lighting preference and participants' attributes such as age.