摘要
针对支持向量机在成绩预测时面临的泛化能力不足问题,提出基于最小流形类内离散度支持向量机M2 SVM。实验选取2000-2009年59次刘翔110m栏成绩作为研究对象,首先将前54次成绩作为训练样本并对模型进行训练得到分类标准,然后将后5次成绩作为测试样本并依次输入模型,比较预测结果与实际结果之间的相似程度,从而说明所提方法的有效性。该方法对人才选拔、成绩提升和梯队建设等具有重要意义。
When dealing with the achievement prediction, SVM （Support Vector Machine） suffers from limitation of generalizationcapability. In view of this, Manifold-based within-class Scatter Support Vector Machine （Mz SVM） is proposed and is used in the a-chievement prediction of 110 m hurdle. 59 achievements of Liu Xiang from the year 2000 to 2009 are collected and construct the ex-perimental dataset. Firstly, the first 54 achievements are used as training set and applied to build the prediction model; the last 5 a-chievements are taken as test set. The effectiveness of Mz SVM is verified by the similarity of the expected results and the actual re-suits. The proposed method MzSVM is important to talent selection, achievement improvement and echelon construction.
出处
《重庆师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2015年第5期165-168,共4页
Journal of Chongqing Normal University:Natural Science
基金
国家自然科学基金(No.61202311)
山西省社会经济统计科研课题(No.KJ[2014]036)
关键词
支持向量机
最小流形类内离散度
110M栏
成绩预测
Support Vector Machine （SVM）
minimum manifold-based within-class scatter Support Vector Machine （MzSVM）
110 m hurdle- achievement prediction
