A comparison of classification accuracy for karate ability using neural networks and discriminant function analysis based on physiological and biomechanical measures of karate athletes

Tim Heazlewood, Hovik Keshishian

Research output: Chapter in Book/Report/Conference proceedingConference Paper published in Proceedings

Abstract

Neural networks can be applied to many predictive data mining applications due to their power, flexibility and relatively easy operations. Predictive neural networks are very useful for applications where the underlying process is complex, such as in classification using a mix of nominal and ratio level variables and for predictive validity based on classification modelling. A neural network can approximate -a wide range of statistical models without requiring the researcher to hypothesize in advance certain relationships between the dependent and independent variables. Neural networks and discriminant function analysis (a more traditional statistical approach), based on physiological and biomechanical measures of karate ability and collected within the ACU exercise physiology laboratory, were compared for there classification accuracy. Twenty four karate athletes were assessed, 12 were classified as high performance athletes with black belt or higher and 12 were classifiedas non-high performance athletes, green belt and lower. Ability level served as the classification variable. The dependent variables were height, weight, age; motor fitness variables were Margaria power test, standing longjump, isometric grip strength, sit-reach flexibility, arm crank, peak aerobic power, anaerobic Wingate power test for peak power, time to peak power, mean power and power/weight; and Karate specific motor fitnesstests were karate agility, power punch, speed punch, reaction time, balance and lower limb bilateral flexibility. ANOV A indicated the general motor fitness constructs of Margaria power test, sit-reach flexibility, arm crankand Wingate power test for peak power; and karate specific motor fitness tests for karate agility test, power punch, speed punch, balance and lower limb bilateral flexibility or lateral split were significantly different (p<0.05 to 0.001). These two data sets were used in the multilayer perceptron (MLP) neural networks and method enter discriminant function analysis. The neural network solution based on the training data set and testing (holdout) data set classified at 100% accuracy karate ability (high and non-high) for the karate specific tests, as well as general motor fitness tests. Discriminant analysis was marginally less effective in classifying ability level. The karate specific tests produced a 95.8% and general motor fitness tests 91.3% correctclassifications, respectively. Neural networks, specifically the multilayer perceptron (MLP) networks, were more effective in predicting group membership and displayed higher predictive validity when compared to discriminant analysis.
Original languageEnglish
Title of host publicationProceedings of the Tenth Australasian Conference on Mathematics and Computers in Sport
EditorsAnthony Bedford, Matthew Ovens
PublisherMathSport (ANZIAM)
Pages197-204
Number of pages8
ISBN (Print)9780957862357
Publication statusPublished - 2010
EventAustralasian Conference of Mathematics and Computers in Sport - Darwin, Australia
Duration: 5 Jul 20107 Jul 2010
Conference number: Tenth

Conference

ConferenceAustralasian Conference of Mathematics and Computers in Sport
CountryAustralia
CityDarwin
Period5/07/107/07/10

Fingerprint Dive into the research topics of 'A comparison of classification accuracy for karate ability using neural networks and discriminant function analysis based on physiological and biomechanical measures of karate athletes'. Together they form a unique fingerprint.

  • Cite this

    Heazlewood, T., & Keshishian, H. (2010). A comparison of classification accuracy for karate ability using neural networks and discriminant function analysis based on physiological and biomechanical measures of karate athletes. In A. Bedford, & M. Ovens (Eds.), Proceedings of the Tenth Australasian Conference on Mathematics and Computers in Sport (pp. 197-204). MathSport (ANZIAM).