Scheduled for Motor Behavior and Measurement Posters, Wednesday, April 2, 2003, 1:00 PM - 2:00 PM, Convention Center: Exhibit Hall A


Prediction of Physical Performance Using Data Mining

Lynn Fielitz, United States Military Academy, West Point, NY and David Scott, New Mexico/University Of, Albuquerque, NM

Data mining has become an innovative and powerful research tool in business for knowledge discovery and the development of predictive models from large volumes of historical data. Statistical techniques of data mining include linear and logistic regression, multivariate analysis, decision trees and neural networks. However, the application of data mining in physical education and sport is in its infancy. The purpose of this study was to examine and compare various data mining tools for analyzing physical aptitude test data as a predictor of physical performance in an academic setting. Historical data collected from freshman cadets at the United States Military Academy was analyzed to determine the level of physical performance required to successfully complete required physical education courses at the Academy. The focus of this study was on the methodology of data preparation, data import utility, and comparison of the data mining software relative to its ease of use and predictive power. Several data mining programs with reportedly similar function were used to build association rules between historical physical aptitude test data and actual cadet performance in a required physical fitness class. Each of the mining programs produced several similar association rules and identified performance standards that incoming cadets should achieve in order to accomplish the strenuous physical requirements of the service academy. However, there were notable differences in the utility and function of each of the software programs used in the study. Major differences included formatting issues for importing data, ability to delineate useful results, and report generation. Conclusions of this study are that analysis of large volumes of data through data mining is and will continue to be a very useful tool for academic research as well as for practitioner based decision making and strategic planning in educational and business settings. Knowledge of some of the tools available and an overview of their applicability, strengths and weaknesses will be presented and should be of interest to health and physical education teachers, researchers, and organizational leaders in our field. The results of the study have both practical and research implications in all dimensions of health, physical education, recreation and sport.

Back to the 2003 AAHPERD National Convention and Exposition