NBA Advanced Metrics (2004-2024)
收藏资源简介:
该数据集由布莱顿大学研究团队构建,覆盖2004-05至2024-25赛季的NBA常规赛数据,包含球队得分、篮板、助攻等48项高级指标。数据量达9,840场次(相当于8个完整赛季),来源于NBA官网统计。通过时序排列记录每队最近比赛的完整特征,旨在解决传统模型因阵容变更和规则演变导致的'概念漂移'问题。其核心价值在于支持长序列LSTM建模,提升跨赛季预测的稳健性,应用于教练策略优化、博彩分析和球迷互动等领域。
This dataset was constructed by a research team from the University of Brighton, covering NBA regular season game data from the 2004-05 season to the 2024-25 season. It includes 48 advanced metrics such as team points, rebounds, and assists, with a total of 9,840 games (equivalent to 8 full seasons) sourced from official NBA statistics. The dataset records the complete features of each team's recent games in chronological order, aiming to address the "concept drift" problem faced by traditional models caused by roster changes and rule evolutions. Its core value lies in enabling long-sequence LSTM modeling, enhancing the robustness of cross-season predictions, and supporting applications in fields including coaching strategy optimization, sports betting analysis, and fan engagement.

- 1Long-Sequence LSTM Modeling for NBA Game Outcome Prediction Using a Novel Multi-Season Dataset布莱顿大学建筑、技术与工程学院 · 2025年



