As sustainable buildings play a vital role in addressing urbanization, resource scarcity, and climate change, automating the identification and understanding of occupant’s activities within these environments becomes paramount. Automated identification of occupant activities allows for the intelligent management of resources such as water, electricity, and heating contributing to the conservation of resources. Adjusting ambient conditions like lighting and temperature based on recognized activities creates a more comfortable and personalized experience for occupants. Data analysis and machine learning (ML) techniques are great at extracting patterns from sensor data. Additionally, knowledge-based approaches such as hierarchical task network (HTN) planning provide a hierarchical way of recognizing activities. This master’s thesis explores the synergistic potential of a hybrid approach that seamlessly integrates ML and HTN planning to drive activity recognition in the domain of sustainable buildings. We make use of ML algorithms on labeled datasets to improve the initial state of the system and HTN planning recognizes activities making use of domain model. We find satisfactory results as our recognized activities relates to the ground truth. The study aims to enhance accuracy in activity recognition in the modeled domain and proves the scope of future research in the field of combining data-based and knowledge-based approaches for activity recognition.