Over the last decade, the world’s population structure has significantly changed where the proportion of elderly people in the population increases and that of young people declines considerably. This change of demographic landscape has caused a severe manpower shortage. It reduces the pool of people in their productive years. Several countries have conducted plans to extend the retirement age to keep elderly workers in the labor market longer. However, maintaining health and physical ability among elderly workers is a primary issue. To address the problem, regular physical exercise is crucial. In this paper, we propose an innovative software system that support the elderly to participate in walking exercise regularly. Our system uses a mobile phone together with a fitness tracker as data collection devices. It applies an ANN algorithm to construct a regression-based model that can predict exercise minutes based on walking statistics. The exercise minutes predicted are then used to generate a personalized exercise plan designed for each elder. The experimental result indicates that our predictive model is practical to be used in real-world settings with reasonable prediction accuracy, about +/− 17 weekly exercise minutes observed.
Paper
Full text
Personalized Walking Exercise Support System for Elderly Based on Machine Learning
Semantic Scholar · Medicine · 2020
Abstract
Over the last decade, the world’s population structure has significantly changed where the proportion of elderly people in the population increases and that of young people declines considerably. This change of demographic landscape has caused a severe manpower shortage. It reduces the pool of people in their productive years. Several countries have conducted plans to extend the retirement age to keep elderly workers in the labor market longer. However, maintaining health and physical ability among elderly workers is a primary issue. To address the problem, regular physical exercise is crucial. In this paper, we propose an innovative software system that support the elderly to participate in walking exercise regularly. Our system uses a mobile phone together with a fitness tracker as data collection devices. It applies an ANN algorithm to construct a regression-based model that can predict exercise minutes based on walking statistics. The exercise minutes predicted are then used to generate a personalized exercise plan designed for each elder. The experimental result indicates that our predictive model is practical to be used in real-world settings with reasonable prediction accuracy, about +/− 17 weekly exercise minutes observed.