The Effectiveness of Using Deep Learning Algorithms in Predicting Daily Activities

Predicting Activities of Daily Living (ADL) for elder people could let them live actively, independently and healthy. In this paper, Authors perform a comparative study to address the effectiveness of deep learning algorithms on ADL. As a baseline structure, The Convolutional Neural Networks (CNN's) as a deep learning algorithm is employed to perform the experiments and conducting the comparative study with the very common used traditional machine learning algorithms. Several factors in the CNN are manipulated to gain the best result in predicting the ADL in comparison with the most ML result in this matter. To reduce the threat to validity, very common data set are used in several previous studies in term of ADL prediction is adopted in this paper. The dataset was collected from a wearable chest accelerometer. The total numbers of participants are 15 and they were performing 7 main activities namely standing up, working at the computer, going up downstairs, standing, walking, walking and talking with someone and talking while standing, walking and going up downstairs. Three experiments were conducted in this paper, and CNN provides promising result in term of ADL predictions for the very common data set in this field and ML algorithms.