A rule induction framework for the determination of representative learning design in skilled performance

ABSTRACT

Representative learning design provides a framework for the extent to which practice simulates key elements of a performance setting. Improving both the measurement and analysis of representative learning design would allow for the refinement of sports training environments that seek to replicate competition conditions and provide additional context to the evaluation of athlete performance. Using rule induction, this study aimed to develop working models for the determination of high frequency, representative events in Australian Rules football kicking. A sample of 9005 kicks from the 2015 Australian Football League season were categorised and analysed according to the following constraints: type of pressure, kick distance, possession source, time in possession, velocity and kick target. The Apriori algorithm was used to develop two models. The first consisted of 10 rules containing the most commonly occurring constraint sets occurring during the kick in AF, with support values ranging from 0.15 to 0.22. None of the rules contained more than three constraints and confidence values ranged from 0.63 to 0.84. The second model considered ineffective and effective kick outcomes and displayed 70% classification accuracy. This research provides a measurement approach to determine the degree of representativeness of sports practice and is directly applicable to various team sports.

KEYWORDS:

• Machine learning
• skill acquisition
• constraints
• Australian rules football

Disclosure statement

No potential conflict of interest was reported by the authors.