Related Experiment Video
Updated: Jul 2, 2025
07:25
An Objective and Child-friendly Assessment of Arm Function by Using a 3-D Sensor
Published on: February 12, 2018
7.0K
Engineering Features from Raw Sensor Data to Analyse Player Movements during Competition.
Valerio Antonini1, Alessandra Mileo1,2, Mark Roantree1,2
1School of Computing, Dublin City University, Dublin 9, D09 V209 Dublin, Ireland.
Sensors (Basel, Switzerland)
|February 24, 2024
Summary
This study introduces a new framework to process raw sports performance data. It transforms simple tracking data into valuable features for machine learning analysis.
Area of Science:
- Sports Science
- Data Science
- Machine Learning
Background:
- Field sports analysis commonly uses descriptive statistics from player tracking data.
- Wearable technologies capture high-volume, but raw, positional data (time, latitude, longitude).
- Raw tracking data is often unsuitable for advanced analysis or machine learning.
Purpose of the Study:
- To develop a multistep feature engineering framework.
- To transform sequential wearable sensor data into machine learning-ready feature sets.
- To enhance the analytical utility of sports performance data.
Main Methods:
- Development of a novel feature engineering framework.
- Application of multistep transformations to sequential positional data.
- Methodology focused on preparing raw data for machine learning algorithms.
Main Results:
- Successfully transformed raw time-series positional data into structured feature sets.
- Demonstrated the framework's capability to prepare data for machine learning.
- Enabled more sophisticated analysis of running performance in field sports.
Conclusions:
- The proposed feature engineering framework effectively enhances raw sports tracking data.
- This approach facilitates the application of machine learning to sports performance analysis.
- It provides a foundation for deeper insights into player dynamics and performance.
