Classification analysis of the badminton five directional lunges
Loading...
Date
2018-06
Authors
Ho, Zhe Wei
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Badminton lunge motion is important skill for players in order to have a
fundamental footwork in badminton. Majority previous badminton studies on lunge
motions investigated male players. The gap was that the findings reported were not
applicable to the female players. There are no works conducted to mine the patterns of
directional badminton lunge motions. Therefore, this study attempted to (i) study the
patterns of lunge motion in the badminton game, (ii) classify badminton players’
postures by lunge type and (iii) compare the differences in the badminton lunge patterns
between university and national level players. The case study involved 11 university
level and 2 national level players in badminton singles captures. Five directional lunge
motions: center-forward, left-forward, right-forward, left-sideward and right-sideward
lunge and its corresponding attributes were tracked through Kinovea software. Data
mining concept is adopted in four stages: data pre-processing, data classification,
significant attribute analysis and knowledge discovery using the WEKA software. REP
Tree classifier is the best selected classifier for its strength and classification capability.
The highest classification accuracy obtained for experimental data-USM and public
data-SEA, were 93.75% and 93.01% respectively on REP Tree classifier. On selective
attribute configuration, the identity (ID), game reaction time (GT) and type of lunge
(LT) significantly enhanced the classification accuracy to 99.61% for experimental
data-USM and 100% for the public data-SEA. Lunge type patterns were related to ID
and GT. Conclusively, the identity, game reaction time and type of lunge were found
being the key determinants for badminton lunge classification accounting for highest
classification accuracy in REP Tree algorithm.