Challenge Description
Visualize the athlete data to better understand athletes' skills and profile. Find the optimal run-time and time gates before the athletes even put on their skis.
Challenge Owner
Swissski is the National Ski Federation and trains Swiss winter sport athletes in eleven disciplines.
- Björn Bruhin - Scientific Assistant Swissski
- Marco Bäschlin - CEO Force8 AG
Solution
Pitch
Our solution is an interactive dashboard that allows users to analyze data. It includes a track visualization with the time and speed of a specific athlete for a specific run, a correlation matrix that can be adjusted dynamically, a t-sne clustering analysis, a descriptive data analysis, and a regression model with adjustable input parameters. The regression model is a multivariate linear regression, and the parameter analysis was done with a Random Forest regressor. The challenge owner will have all the code and can expand it with additional data.
The goal was to identify the most important parameters for athletes to improve their run. We analyzed the aggregated and raw data and then applied initial clustering methods and basic ML algorithms. As the dataset was at times inconsistent or incomplete, it was difficult to fit accurate ML models. However, we created insightful data visualizations and implemented basic ML models to find the most important parameters.
Links
Potential next steps
- Record more data and create higher accuracy for ML models
- Improve the dashboard
- Run more sophisticated algorithms and neural networks
Data
Swissski has collected the data of several athletes' runs on several courses, including time gates, turn velocities, and slalom angles.