美国棒球运动员薪资等级分类预测的多模型对比分析  

Multi-model Comparative Analysis on Salary Classification Forecasting of American Baseball Players

作　　者：严驰昊 YAN Chi-hao(School of Statistics&Mathematics,ZUEL,Wuhan 430073,China)

机构地区：[1]中南财经政法大学统计与数学学院,湖北武汉430073

出　　处：《南宁师范大学学报（自然科学版）》2023年第1期79-87,共9页Journal of Nanning Normal University：Natural Science Edition

摘　　要：随着竞技体育商业化、市场化,运动员薪资水平预测对解决运动员薪资、合约纠纷及制定运动员薪资制度都有积极作用。由于每年进入各商业联盟的运动员特征具一定的随机性、不可知性,因此,运动员数据样本分布常常是多变的。为了精确预测运动员薪资水平,该研究通过多模型对比分析,利用岭回归、Lasso回归及逐步回归进行变量筛选;球员薪资分类预测选择KNN、LDA法、Logistic回归CART决策树和SVM模型五类分别能够适应不同样本分布情况的分类预测模型进行对比分析,通过模型组合得到最优分类预测模型。结果表明:Stepwise-SVM组合模型无论从预测精确度还是稳健性角度考察均为本研究假设空间下最优的运动员薪资等级分类预测模型,能为竞技体育市场运动员薪资制度的制定提供重要启发。With the commercialization and marketization of competitive sports,the forecast of athletes salary level is of great significance to solve the disputes of athletes salary contract and formulate athletes salary system.However,since the characteristics of athletes entering each business league every year are random and unknowable,the distribution of athletesdata samples is often changeable.Therefore,to realize precise salary classification forecast of athlete,this study,based on data from American baseball players,attempts to screen variables through Ridge,Lasso and Stepwise regression.Then use KNN,LDA,Logistic regression,CART decision tree and SVM model which can adapt to different sample distribution for comparative analysis.Finaly,the optimal model was obtained by model combination.The results show that the Stepwise-SVMcombination model is the best model for forecasting athletes salary level in the hypothesis space,whichmay provide important inspiration for the formulation of athletes salary system in competitive sports market.

关 键 词：美国棒球运动员 薪资 分类预测 模型 多模型对比 

分 类 号：F832.39[经济管理—金融学] F272.92