| Summary: | Studies on development athlete were frequently evaluated through the subjective
opinions of coaches and scouts that can lead to repeated errors and misjudgements when
they are used in isolation. Due to this, the data ets of development athletes can be very
large and often containing many variable ,as well as the standard univariate statistical
analy is techniques which are focused on the identification of single variables to
differentiate between performance levels. The objectives of the study are to; 1) identify
the most significant parameters based on anthropometric and physical fitness
parameters, 2) develop and verify the index based on anthropometric and physical
fitness parameters, 3) develop the model based on anthropometric and physical fitness
parameters u ing Machine learning. A total of 309 Terengganu male athletes (16.8 ±
2.3 years) were drawn randomly from individual port categories. Principal component
analysis (PCA) was used to identify the most essential performance variables,
developing an index of Terengganu Athletes Capabilities Index (TACI) with five
classification namely, Excellent Performance Athletes (EPA), Good Performance
Athletes (GPA), Moderate Performance Athletes (MPA), Low Performance Athletes
(LP A) and Poor Perf rmance Athletes (PP A). The discriminant analysis (DA) was
applied to validate the index group. he -Neare t Neighbour (k-NN), Tree, Support
Vector Machine (SVM), Random Forest, Artificial Neural Network (ANN), Naive
Bayes and Logistic Regression I arning algorithms were trained to predict and classify
the EPA, GP A, MP A, LP A and PP A. PCA indicates that out of nine performance
variables evaluated, only a set of six variables are essentials with 66.4% of total
variance. Most of the athletes were grouped on MPA classification C 0.1 %), followed
by LPA (28.5%), GPA (21%), PPA (15.8%) and EPA (4.5%) based on TACI. Finally,
ANN learning algorithm outperformed all the performance metrics of machine learning
with 0.88 classification accuracies, 0.87 F1 score for the model, and 0.88 precisions for
all the evaluated parameters. A strong sensitivity and specificity are observed between
the classification and the selected performance variables for the model. The study
revealed that the benefits of multivariate analysis and machine learning techniques
permit the researcher to identify Terengganu athletes accurately and holistically, hence
reducing the time, cost and manpower on the talent identification program. Importantly,
the identification of the potential Terengganu athletes through the holistically reliable
battery test with the help of machine learning technique revealed more useful insights
within the field of talent identification and science specifically.
|
|---|