基于支持向量机的NBA季后赛预测方法
曾 磐,朱安民

深圳大学计算机与软件学院,广东深圳 518060

计算机感知; NBA季后赛; 因子分析; 支持向量机; 不平衡数据; 欠采样; 过采样

A SVM-based model for NBA playoffs prediction
Zeng Pan and Zhu Anmin

College of Computer Science and Software Engineering, Shenzhen University, Shenzhen 518060, Guangdong Province, P.R.China

computer perception; NBA playoffs; factor analysis; support vector machine; unbalanced data; undersampling; oversampling

DOI: 10.3724/SP.J.1249.2016.01062

备注

通过对参加NBA赛事的每支球队在常规赛阶段的数据统计,整合出球队常规赛综合得分、球员常规赛综合得分、主教练水平及主客场因素4项指标作为一支球队的综合实力体现,构建训练样本,使用适合小样本数据的基于结构风险最小化的支持向量机(support vector machine, SVM)来训练一个预测模型,并预测NBA季后赛每场比赛的胜负.通过结合欠采样和过采样技术消除训练样本中的不平衡数据,更好地发挥SVM的学习能力.实验表明,所提方法具有较好的预测效果.

We use four statistics data from the NBA regular season including team comprehensive score, player comprehensive score, chief coach capability, and the impact of home and away games as the indicators to build up training samples. Then, we construct a prediction model based on a support vector machine(SVM), which is suitable for solving the problem of a small sample size and has the theoretical basis of structural risk minimization. We combine the technique of undersampling and oversampling to remove the unbalanced data in training datasets, thus improving the leaning ability of SVM. Experimental results show that the proposed method has a relatively preferable performance.

·