Predição da produtividade a partir da qualidade ambiental interior: uma abordagem com regressão linear múltipla e k-means

Abstract

Due to changes in people's living and working patterns, studies on Indoor Environmental Quality (IEQ) have become increasingly relevant, as they directly influence people's health, comfort, and, above all, productivity. This study aimed to develop and apply a predictive productivity model based on the four tangible aspects of IEQ: thermal comfort, air quality, acoustic comfort, and visual comfort. The application scenario consisted of a Federal University in Brazil, in the city of Ponta Grossa. Multiple Linear Regression and K-means clustering were used with data collected in university classrooms. The main results were: (i) the regression obtained a good fit (R² = 0.560), with the operating temperature presenting a positive standardized coefficient (β = 0.524), followed by lighting (β = 0.226), while noise (β = - 0.261) and CO₂ (β = -0.197) had a negative impact, indicating that temperature and lighting are factors with the most significant positive influence on productivity; (ii) the K-means analysis formed two distinct clusters, where the cluster with the best environmental conditions obtained an average productivity score of 48.71%, compared to 46.63% in the cluster with the worst conditions; a direct correlation was identified between more comfortable environments and higher levels of perceived productivity, with statistically significant differences (p < 0.001); (iii) Through an analysis of the variation in perceived productivity between periods of the day, it was found that the highest perception of productivity is concentrated between 3:50 p.m. and 5:30 p.m., corresponding to the end of the school day. The results reinforce the importance of IEQ as a strategic factor in planning school environments, especially in Brazilian contexts, where climatic and structural variables can amplify its effects. The study contributes to advancing predictive models applicable to the educational context. It highlights the need to consider contextual, subjective, and methodological factors to ensure the reliability of results and to guide improvements in learning spaces.