BUILDING A BADMINTON COURT BOOKINGPLATFORM WITH A RECOMMENDATION SYSTEM

Abstract

Finding suitable playing partners remains a significant challenge on badminton platforms due to differences in skill levels, playing styles, and severely unbalanced user interaction outcomes. This study addresses this issue by developing a badminton court booking system integrated with a partner recommendation model. The recommendation method combines rule-based filtering with a Logistic Regression model trained on past invitation and acceptance data to estimate the likelihood of successful pairing. The recommendation model was evaluated on a severely unbalanced dataset using ROCAUC as the primary performance indicator. Experimental results show the model achieved an ROC-AUC of 0.861, PR-AUC of 0.061, a recall of 0.91 for accepted invitations, and an overall accuracy of 0.74. Although accuracy is still limited due to class imbalance, the model demonstrates strong ranking capabilities. Overall, the results confirm the effectiveness of the proposed system for the proposed partner scenarios.