Fusion of classifiers based on physical activities data from smartphone user

The behavior recognition of mobile phone users is based on different types of smartphone sensors. The accelerometer is one of these sensors which can represent the person activity. In this paper, in order to recognize and classify the physical activities of user's smartphones, we will use the Dempster-Shafer (DS) theory of belief functions. After applying decision tree machine learning algorithm on each of the three axis of movement, we will fuse different combinations of these results. Then, we will fuse different machine learning algorithms applied on three axis of movement together. We will show that using a classifier of each attribute give better results than applying on all attributes.