Determination of parameters affecting thermal sensations using Support Vector Machine coupled with Firefly Algorithm

• Published in: Journal of thermal biology
• Main Authors: Kariminia, Shahab, Shervin Motamedi, Shahaboddin Shamshirband, Roslan Hashim, Chandrabhushan Roy, Dalibor Petkovic
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 2016
• Subjects: climate; demographic statistics; gender; Hot and dry climate; neural networks; Open urban area; prediction; questionnaires; sensation; Support vector machine; support vector machines; surveys; Thermal comfort; Iran
• ISSN: 0306-4565

Thermal environment in open urban spaces impacts its use. Thermal adaptation engages user's physiological, psychological and behavioral factors. This plays an important role in user's ability to assess thermal environments. Previous studies have rarely addressed the effect of factors such as gender, age and locality on thermal sensation particularly in hot and dry climate. This study investigated thermal comfort of visitors at two public squares in Iran against their demographics. In addition, the role of built environment within the squares was analyzed. Assessing thermal comfort of the subjects required taking physical measurement and questionnaire survey. Support Vector Machine (SVM) was further coupled with Firefly Algorithm (FFA) methodology to estimate thermal comfort of the visitors. The role of built environment within the squares was analyzed. Results from SVM-FFA were compared with conventional genetic programming (GP) and artificial neural network (ANN). It has been found that our SVM-FFA results were similar to the actual measured data. Based upon simulated results, it is evident that SVM-FFA can be employed effectively towards prediction of visitors’ thermal sensations approximation to actual values. Surveyed results illustrated that the SVM-FFA model as proposed here is suitable and precise to predict visitors’ thermal sensations. Based on this study we have proven that the predictability performance of our model is more reliable and superior compared to other approaches as discussed in this study.