Predicting Tenant Behavior in Smart Households Using Machine Learning on Time-Series Data

As smart home technology advances, the ability to anticipate tenant behavior will be essential for improving home automation, personalized services, and energy efficiency. Traditional rule-based automation methods are limited in their adaptability, making it necessary to develop intelligent predictive solutions. This thesis will focus on developing a machine learning approach for predicting tenant behavior in smart household environments using time-series data. The study will involve designing a robust data acquisition and preprocessing pipeline to extract structured and reliable data from real-world smart home deployments. The core objective will be to create adaptive models capable of identifying and predicting behavioral patterns, ensuring long-term effectiveness as user habits evolve. Additionally, the research will examine computational efficiency, with the goal of developing a lightweight, resource-conscious implementation suitable for deployment on edge devices within smart home systems. By enabling proactive automation and optimization, this work aims to contribute to the advancement of intelligent behavior modeling in real-world smart environments.