Development of an Artificial Intelligence-Based Application to Predict Air Quality and Environmental Comfort from Design Parameters in New Buildings

In this paper, the development of an AI-based application to predict levels of environmental comfort and air quality in new buildings from the design stage is presented. The objective of this approach is to improve construction quality and reduce their environmental impact. This project uses as knowledge base the experience of the enterprise Bioteckta, a bioclimatics consulting firm, and is funded and supported by the Colombian Ministry of Science (MINCIENCIAS) and the National Learning Service (SENA). The research is divided into three key stages: predictive model optimization, software development, and validation. As a result, a predictive model and a Web application were obtained. A model was generated using 18 design parameters as input variables. In the initial model assessment, an accuracy of 91% was obtained, which then was improved to 100% with machine learning techniques (decision trees). This model was then integrated into a practical and user-friendly Web platform, and a coherence with the base model of 99.7% was obtained, with only one case getting a different result from the optimized base model. These results indicate that the model and Web platform are reliable and can be used to predict environmental comfort in new buildings in a simple and practical way.