Open-plan Glare Evaluator (OGE): A Demonstration of a New Glare Prediction Approach Using Machine Learning Algorithms

Abstract

Predicting discomfort glare in open-plan offices is a challenging problem. Although glare research has existed for more than 50 years, all current glare metrics have accuracy limitations, especially in open-plan offices with low lighting levels. Thus, it is crucial to develop a new method to predict glare more accurately. This paper is the first to adopt Machine Learning (ML) approaches in the prediction of glare. This research aims to demonstrate the validity of this approach by comparing the accuracy of the new ML model for open-plan offices (OGE) to the accuracy of the existing glare metrics using local dataset. To utilize and test this approach, Post-Occupancy Evaluation (POE) and High Dynamic Range (HDR) images were collected from 80 occupants (n=80) in four different open-plan offices in Brisbane, Australia. Consequently, various multi-region luminance values, luminance, and glare indices were calculated and examined as input features to train ML models. The accuracy of the ML model was compared to the accuracy of 24 indices which were also evaluated using a Receiver Operating Characteristic (ROC) analysis to identify the best cutoff values (thresholds) for each index in open-plan configurations. Results showed that the ML approach could predict glare with an accuracy of 83.8% (0.80 true positive rate and 0.86 true negative rate), which outperformed the accuracy of the previously developed glare metrics. OGE is applicable for open-plan office situations with low vertical illuminance (200 to 600 lux). However, ML models can be trained with more substantial datasets to achieve global model.