Regression tree classification for activity prediction in smart homes
Bryan Minor, Diane J. Cook · 2014
The growing number of older adults in the population has created an increasing need for health-assistive systems, including prompting interventions to provide activity reminders. In this paper, we present a new regression-tree-based activity forecasting algorithm to predict the occurrence of future activities for prompting initiation of such activities. This automated algorithm extracts high-level features from sensor events and inputs these features to a machine learning algorithm which forecasts when a target activity will next occur. We compare this system to a standard linear regression classification using real data from smart homes. The forecasting algorithm is shown to provide lower error rates over the linear regression model.