Abstract
At the older age, the likelihood of disability increases and hence the increasing need for long-term care and facilities to assist elderly people who endure gradual loss of body function. Early detection of changes in health condition of elderly can increase safety for elderly people in emergency conditions. Alert prediction can be viewed as an assistive technology that will deliver appropriate escalation in the earliest time so that elderly can receive immediate responses. Supervised learning can be used as a tool to predict alert in emergency condition by training historical data of elderly behaviors and conditions. This paper proposed emergency alert prediction using supervised learning algorithms. Three algorithms of supervised learning, namely deep learning, k-NN, and LVQ were used to simulate the proposed system. The objective of this paper is to investigate the performance of three algorithms in making emergency alert prediction for elderly living independently. We conducted experiments for 30 days to elderly living independently and we obtained 1038 datasets. The simulation results showed deep learning performed the best accuracy 99.57% correct. Whereas k-NN obtained the best accuracy 90.79% correct, and LVQ obtained the best accuracy 80.32%.