Development of player position clustering and passing path visualization system for basketball tactical analysis

Objective: To develop an integrated system with both data mining and visualization functions in traditional basketball tactical analysis, so as to provide data support for coaching tactics formulation and player training optimization. Methods: Firstly, the coordinate standardization method is used to preprocess player position data, and an improved K-means clustering algorithm by introducing a local density factor is proposed to achieve accurate clustering of player positions. Secondly, combined with spatiotemporal data modeling, two-dimensional static and three-dimensional dynamic visualization models of the passing path are constructed using D3.js and Three.js technologies. Finally, an end-to-end system is developed based on Vue.js and Python Flask frameworks. Results: Through verification with the NBA open game dataset, the Rand index (RI) of the improved K-means algorithm reached 0.89, which was 17.1% higher than that of the traditional K-means algorithm, and the clustering accuracy reached 88.6%. The developed visualization system can realize spatiotemporal correlation display, interactive query, and critical path screening of passing paths, and the average satisfaction score is 89.3 points (out of 100) after evaluation by 10 basketball professionals. Conclusion: The system effectively solves the core requirements of position quantification and path visualization in basketball tactical analysis, and the algorithm performance and visualization effect meet practical application scenarios, providing a new technical means for basketball tactical analysis.