Artificial Intelligence and Mobile Phone Sensing Based User Activity Recognition

With the development of Micro Electro Mechanical Systems, a growing number of portable devices and wearable devices equipped with built-in sensors, which can detect the physical movements, such as identifying the action type and record the duration of exercise. Since the amount of data collected from sensors grows, automatic activity recognition becomes an important issue to living in a smart life. Therefore, this paper aims to use various kinds of machine learning techniques to build the automatic activity classification model, including Logistic regression, Decision tree, Random forest and Support vector machine algorism. Furthermore, we evaluated the prediction performance of four supervised machine learning classification models. Results of the experiments show that under specific acceptance of accuracy and minimum model training time, the decision tree algorithm creates the best model. However, if consider the accuracy as the only pursue, adopting the support vector machine algorithm will get the better model.