Classification of User Movement Data

Abstract

In this paper, a comprehensive analysis of various classification techniques is used on user movement data. The dataset used here utilizes a Nexus One Android smartphone, equipped with accelerometer sensor devices. The data was transformed and applied to two deep learning techniques: Convolutional Neural Network (CNN) and Long Short Term Memory-Recurrent Neural Network (LSTM-RNN). The results obtained via CNN and LSTM-RNN are compared with the traditional classifiers such as k-Nearest Neighbors (k-NN) and Feed Forward Neural Network (FFNN). The results show that the deep learning approaches were outperformed by the traditional classifiers on the applied dataset. While deep learning techniques reached a maximum accuracy of 84% utilizing LSTM, k-NN obtained an accuracy of 99.6%.