Classification of Elderly’s Home Activities using Tree-Based Model

Abstract

With the aging population, it becomes increasingly essential to prioritize the well-being and safety of older individuals by seeking innovative solutions. One promising approach is the integration of smart home technology equipped with sensors, which can significantly enhance independent living for the elderly. This research centres around the utilization of machine learning techniques to detect and classify activities within smart home environments, specifically focusing on older individuals. By analysing data on their activities and movements, valuable insights can be gained to identify potential behavioural patterns indicating cognitive decline or health issues. The primary goal of the research is to develop robust algorithms capable of accurately identifying and categorizing elderly activities by analysing sensor data collected from various devices. Tree-based approaches, namely Decision Tree and Random Forest, are adopted to achieve their balanced accuracy in categorizing activities and predicting trends and patterns. The results showed that the Random Forest model outperformed the Decision Tree model, achieving a higher balanced accuracy of 69.85% on the training set. This study highlights the immense potential of machine learning in conjunction with smart home technology to significantly improve the lives of the elderly. By accurately identifying and categorizing their activities, caregivers and healthcare professionals can proactively intervene, leading to better outcomes and increased independence for older adults.