An Evaluation of General Classification Models for Multi Resident Activity Recognition
Human activity recognition has become a popular research field in smart environments such as smart homes, classrooms, and offices. Most of the research has been focused on single resident environment of activity recognition. However, in real life the live environment us usually inhabited by more than one person. So the research on multi-resident activity recognition is vitally important. The aim is to recognise human motions based on data collected from different types of sensors. Most works have considered multi-resident activity recognition utilizing various classification models. We believe that the existing methods alone cannot efficiently recognise multi-resident actions in the complicated situations of multi-resident environments. In this research, we address the research question of what is optimal general machine learning classification model for multi-resident activity recognition. We evaluate six general classification models in four datasets. We found that in six classification models the linear SVM has highest accuracy which obtained 88.57%. Second only to linear SVM is HMM which achieved 81.88% in terms of accuracy. After adopting the statistical analysis test, we conclude that the model will influence the classifying of results. Thus, creating or training an efficient and stable classification method remains an open challenge requiring further study.