A Hybrid Support Vector Machines and Decision Tree Model for Analyzing Basketball Games

• Main Authors: Lan-Hung Chang Liao, 張廖年鴻
• Other Authors: Ping-Feng Pai
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/90352771610279926935

碩士 === 國立暨南國際大學 === 資訊管理學系 === 101 === Support Vector Machines (SVM), which follows the principle of structural risk minimization, is an emerging and powerful technique in coping with classification problems. However, a lack of rule generation is a weakness of the SVM model, especially in analyzing sporting results. This investigation developed a hybrid model integrating the SVM technique and a decision tree approach (HSVMDT) to predict the results of basketball games, and to provide rules to aid coaches in developing strategies. The HSVMDT model employed the unique strength of SVM and decision tree in generating rules and predicting the outcomes of games. With predicted outcomes of games, and rules yielded from the HSVMDT model, coaches can easily and quickly learn essential factors increasing the chances to win games. Data collected from the National Basketball Association (NBA) were used to examine the performance of the designed HSVMDT model. Empirical results showed that the proposed HSVMDT model can obtain relatively satisfactory prediction accuracy by comparison with previous studies on analyzing basketball games. The developed model is therefore a promising alternative for analyzing the results of basketball competitions.