A systematic approach with data mining for analyzing physical activity for an activity recognition system

The increasing inclusion of plethora of sensors in sophisticated and latest generation smart phones opens new avenues for Data Mining applications for activity recognition, a task which involves identifying the physical activity a user is performing. In this paper, we describe and evaluate phone-based accelerometers to perform activity recognition. In order to implement our system, we collected labeled accelerometer data from twenty-three users as they performed daily activities such as strolling, running, climbing stairs, Relaxing (sitting inhaling), and Relaxing(standing exhaling), and then aggregated this time series data into examples that summarize the user activity over 10-second intervals. We transformed raw data into examples by tracing of action duration and segmented acceleration data by equal binning to make it a training data for input. We then used the resulting training data to induce a predictive model for activity recognition. We use WEKA data mining tools for data preprocessing and classification. Experimentation carried out based on our data classification stages eventually traces activities with finer accuracy.