Modelling Training Adaptation in Swimming Using Artificial Neural Network Geometric Optimisation.

Area of Science:

• Sports Science
• Biomechanical Engineering
• Data Science in Athletics

Background:

• Monitoring training adaptation is crucial for athlete performance and injury prevention.
• Traditional methods may not fully capture the complex interplay between training load and recovery.
• Developing objective metrics for training response is an ongoing challenge in sports science.

Purpose of the Study:

• To model and graphically differentiate training adaptation from maladaptation using Artificial Neural Network (ANN) geometric optimization.
• To introduce and validate the Geometric Activity Performance Index (GAPI) as a measure of training response.
• To assess the correlation between GAPI and key performance indicators in swimmers.

Main Methods:

• Collected 26 weeks of training and recovery data from 38 swimmers via a web platform.
• Applied ANN geometric optimization to model adaptation and maladaptation.
• Utilized jittering and ensemble modeling to mitigate model overfitting.
• Performed Spearman rank and Blomqvist β tests to validate GAPI's relevance.

Main Results:

• Successfully modeled and graphically separated training adaptation from maladaptation in 13 swimmers.
• Introduced GAPI, a ratio of adaptation to maladaptation area, as a novel index.
• GAPI derived from external load (distance) and internal load (session-RPE) demonstrated the strongest correlation with performance.
• Validated the relevance of collected parameters through correlation and independence tests.

Conclusions:

• Artificial Neural Network geometric optimization is a promising technique for modeling training adaptation.
• GAPI serves as a potentially valuable numerical surrogate for tracking athlete training status throughout a season.
• Objective modeling of training adaptation can enhance athlete monitoring and performance optimization.