Sports Injuries Prediction and Monitoring using Kernel based Techniques
Amit Yadid, M.Sc. student at the Department of Industrial Engineering
26 April 2022, 14:30 PM, Room 206& via zoom
Abstract:
Sports Analytics has gained significant momentum in recent decades. The understanding that the use of statistics and data-driven methodologies to perform business decisions has a significant impact on various aspects of the sport. One of the most challenging aspects of Sports Analytics is the ability to predict and monitor athletes' injuries. A sports injury can be defined as a self-organizing complex system, i.e. multiple parameters that interact among each other, and shift the system from one phase to another. The physics of non-linear phase transitions (e.g., sudden changes in the behavior) in complex systems is a key component for the early detection and prevention of sports injuries. In this work, several types of kernel-based techniques are combined for identification and monitoring of athletes’ injuries. First, we propose an algorithm for injury identification applied on short overlapping trajectories from the athlete performance and demographic statistics. The normal behavior of these trajectories is compared to new sensed trajectories in terms of the data's geometric structure. We utilize diffusion maps, a manifold learning method, for identifying deviations from the normal behavior. Diffusion maps reduce the dimension of the data while preserving its geometric structure. The low-dimensional representation reveals patterns that indicate a possible upcoming injury. The outcome is a simplified representation of the player behavior throughout the sampled period, where anomalous data points are assigned with large values that can be detected using a pre-defined threshold. The proposed algorithm is applied to sport injury data as well as other types of sensor data.
Bio:
Amit Yadid, M.Sc student at the department of industrial Engineering in Tel Aviv University, specializing in Data Science. Amit holds a B.Sc degree in industrial Engineering from Afeka College of Engineering. In the past few years, Amit has been leading the development of objects validation methodologies as part of the Autonomous Vehicle project in Mobileye. His research focuses on manifold learning, pattern recognition and time series prediction. This work was conducted under the supervision of Dr. Neta Rabin.
