| Summary: | Some of our capacities and functional mechanisms tend to become impaired with aging. What kind of impact can a fall have on someone's life? Falls have major consequences over time and may even lead to institutionalization or death. They carry lots of costs on treatment, not only for the one that falls, but also to the national medical system. It is possible to know who has a greater risk for falling and then, through some intervention, avoid falls. Some examples of risk factors for falls are: muscle weakness, poor gait/balance, and many others, having multiple factorial sources and many measuring formats. A good combination and evaluation of risk factors is needed to assess the risk for falling and an easy to apply, not time consuming with good accuracy and precision solution that combines multiple factors is not yet in the ideal stage of development. This work was developed in partnership with Fraunhofer Portugal AICOS. The Association has developed a wireless device with inertial sensors that allows the extraction of movement metrics in the evaluation tests of the risk factors along with balance metrics extracted with the pressure platform developed by Sensing Future, partner of their FallSensing project. In the scope of the same project, in partnership with ESTeSC - Coimbra Health School, data relative to the risk factors was collected from various persons in various contexts and conditions. All the data collected was stored on a database, including the data collected from the sensors and the information on the occurrence of falls in the following six months after the initial collection of data. The objective of the present work is to determine which are the most relevant factors and metrics from the sensors and how they can be combined to distinguish between persons that will fall in the future from the ones who won't. The approach was based on the application of Machine Learning methods, namely supervised learning classification algorithms, trying to understand which of the algorithms and combination of methods would allow a better distinction between the group of people that had fallen from the remaining, having as basis the comparison of their performance metrics.
|
|---|