The Classification of Badminton Strokes: A Feature Importance Investigation

This work employed the Mean Decrease Impurity (MDI) feature selection technique in classifying different badminton strokes. An online repository that consists of data acquired from an Inertial Measurement Unit of players executing five distinct strokes were used in the study. A total of 104 statistical features were extracted from the data. A vanilla Random Forest model was used to classify the strokes based on all the features extracted as well as features identified via the MDI technique. The dataset was split into an 80:20 ratio for training and testing. It was demonstrated from the study that a total of 59 features were identified to be significant that could yield a comparable testing accuracy. The findings suggest that MDI streamlined the important features whilst discarding redundant and less informative features. This allow for a more computationally efficient model to be developed and practically deployed without sacrificing its predictive power.