Predicting Highlights in Badminton Matches: A Knowledge- Aware Model Analysis of Rally-by-Rally Statistics

Abstract In the realm of sports analytics, predicting highlights in badminton matches plays a crucial role in enhancing fan engagement and broadcasting. This study introduces a Knowledge-Aware Model (KAM) that integrates domain-specific knowledge and data-driven techniques to predict game highlights. Analyzing an extensive dataset from the 2017 World Championships and the 2018 Thomas Cup, comprising over 5,180 individual rallies from 140 singles matches, the KAM considers serving and receiving zones, technical stroke nuances, total strokes, rally time and point outcomes. Comparative evaluations against baseline models and state-of-the-art approaches demonstrate the KAM's superiority, achieving an F1-score of 0.793. By combining comprehensive match statistics with rally-specific data, the KAM offers an innovative approach to predicting highlights, with implications extending beyond badminton to multimedia analysis and recommendation systems. This research presents a pivotal step towards more precise and engaging sports analytics.