Developing neural networks to investigate relationships between lighting quality and lighting glare indices

The present work compares the ability of the two most used glare indices, the Daylight Glare Probability (DGP) and the International Commission on Illumination (CIE) Glare Index (CGI), using Multiple Correspondence Analysis (MCA) and Artificial Neural Networks (ANN). The research investigates the efficiency of indexes in predictive indoor lighting quality. This study was carried out by analyzing data from a survey administered to ninety students in real design classrooms in the city of Biskra, Algeria. The experiment was conducted using three different lighting indoor conditions: natural and artificial lighting and mixed lighting. The true prediction of the Daylight Glare Probability for the variable Comfortable was 60.60%, and for (CIE) Glare Index the prediction values were equal to 44.60% for the same variable.