Using player sensor data for training load-injury modelling
Training in the lead up to competition season is crucial. Clubs do have to be wary of player injuries though, which when present can put undue pressure on a club and affect their performance during the season.
By working with an AFL club, this project will further our knowledge of the relationship between training load intensities and injury risks among individual players.
Different models, including predictive models, are being developed to assist coaches with their planned training loads, ensuring better outcomes. The data making our models viable incude;
- Players perception of exertion
- Body sensors, used during;
- Field training;
- Conditioning training
- Professor Meg Morris, La Trobe University and Healthscope
- Professor Kay Crossley, La Trobe University
- Mr David Carey, La Trobe University
Physical activity assessment using novel machine learning on free-living accelerometer measurements
We are using data from hip and wrist worn devices that measure acceleration rates among over 100 participants. Our research will:
- Assess the suitability of existing wrist-worn devices
- develop and assess novel machine learning models
Our study contains raw 100 Hz triaxial accelerometry files for 110 participants who have simultaneously worn ActiGraph accelerometers on their wrist and hip for two x 7-day periods. The study itself has been split into two parts:
- determine Physical Activity (PA) patterns from existing algorithms and machine learning models. The suitability of the existing wrist-worn models will be assessed against the hip-worn algorithms
- Using our computer modelling expertise and data from the first 7-day period, we will develop bespoke machine learning model(s) that determine PA patterns from wrist-worn models.
Given the widespread use of wrist-worn accelerometers in both population-based PA surveillance and research, this research will attract both commercial and research interests.