Implementation and Evaluation of the K – Nearest Neighbors Algorithm in Badminton Movement Classification

To meet the needs of automated sports analysys, this study will develop and evaluated a bandminton motion analysis system that uses the K-nearest Neighbors (KNN) algorithm. This system will detect netting, smash, and serve motions and assess whether the labels are correct and inccprrect. The system uses MediaPipe Pose to extrac keypoints from 3-5 second videos, with data normalized using StandartScaler. Evaluation result show an eccuracy of 0.8438 for netting, 0.8276 for smashes, and 0.7778 for serves. Keypoints extraction time ranges from 4.53 to 25.44 seconds, influaced by lighting conditions, while prediction time is efficient at 0.03-0.05 second. Although this system can be used for sport training, additional data and features are needed to improve performance in low-ligh conditions.