A Data Mining Framework for Activity Recognition in Smart Environments

Recent years have witnessed the emergence of Smart Environments technology for assisting people with their daily routines and for remote health monitoring. A lot of work has been done in the past few years on Activity Recognition and the technology is not just at the stage of experimentation in the labs, but is ready to be deployed on a larger scale. In this paper, we design a data-mining framework to extract the useful features from sensor data collected in the smart home environment and select the most important features based on two different feature selection criterions, then utilize several machine learning techniques to recognize the activities. To validate these algorithms, we use real sensor data collected from volunteers living in our smart apartment test bed. We compare the performance between alternative learning algorithms and analyze the prediction results of two different group experiments performed in the smart home.