Data Mining for Automatic Identification and Analysis of Daily Indoor Environmental Patterns in Residential Buildings

With the development of the Internet of Things, smart devices have made a large amount of real-time Indoor Environmental Quality (IEQ) data available. In this context, the utilization of data mining clustering algorithms offers several advantages for knowledge discovery by extracting typical daily patterns from extensive raw data. However, literature research on clustering processes for IEQ analysis is limited. In this study, a room-level analysis was performed in 8 rooms of 4 flats, selected from two multistorey residential social housing buildings (Reggio Emilia, Italy) and monitored for the entire summer period in 2022. The dataset contained observations about IEQ, outdoor climatic conditions and building occupant behaviour. Firstly, based on a cluster analysis, meaningful daily indoor temperature patterns were extracted from 495 time-series data curves. Secondly, post-clustering analyses via classification trees were performed. The findings suggest that the proposed clustering method is suitable for detecting daily patterns in the IEQ domain. The outcomes of the post-clustering analysis offer valuable insights for enhancing the interpretability of clustering results and evaluating the influencing factors and connection between dynamic influencing factors (e.g., outdoor conditions, spatial characteristics, human factors) and IEQ patterns in residential buildings.