Towards Automated Occupant Profile Creation in Smart Buildings: A machine learning-enabled approach for user persona generation
The user persona is a communication tool that allows designers to generate a mental model representing the archetype of users. The development of building occupant personas has been proven effective in facilitating human-centered smart building design, which takes into account occupant comfort, behavior, and energy consumption. Optimizing building energy consumption requires an in-depth understanding of occupants' preferences and behaviors. However, current approaches to developing building occupant personas are significantly hindered by the need for manual data processing and analysis.
In this project, we propose a machine learning-based methodology to classify and predict occupant attributes, with the aim of streamlining the creation of building occupant personas. We have closely examined the 2015 Residential Energy Consumption Dataset using five machine learning techniques: Linear Discriminant Analysis, K-Nearest Neighbors, Random Forest, Support Vector Machine, and the AdaBoost classifier. These techniques have been applied to predict 16 distinct occupant attributes such as age, education, and thermal comfort. The models demonstrated reasonable accuracy in predicting most occupant traits, with particularly high accuracy (over 90%) for variables such as household occupancy, age group, and preferences for primary cooling equipment usage. These findings underline the potential of machine learning techniques in predicting occupant attributes, which could facilitate the automated development of building occupant personas and thus reduce the need for human intervention.
The research article can be found here.
Project Team: Sheik Murad Hassan Anik, Dr. Ray Gao (Xinghua), and Dr. Na Meng at Virginia Polytechnic Institute and State University (Virginia Tech).
Dr. Gao’s work in this project is under an umbrella research program Internet of Things Enabled Data Acquisition Framework for Smart Building Applications, which is initiated by Dr. Gao.
A previous study about Data-Driven Building Occupant Profile (Persona) Development can be found here.
code_book_v3.ipynb is the source code for analysis.
Resources contains the raw data.
output contains the analysis results.