摘要
针对如何准确预测高校录取分数线,帮助高考生做出更加准确的志愿填报决策问题,提出一种基于Stacking集成思想的双层模型。该模型采用机器学习算法暴露特征重要性,融合3个单一算法并使用交叉检验法和网格搜索法进行参数优化。通过在贵州省2018-2022五年高考高校录取数据上进行实验结果表明,该双层融合模型的预测效果优于支持向量回归、决策树、随机森林等单一模型和其他集成模型;预测误差在5分以内的精度超过95%,平均绝对值误差低于2.43;较单一模型中表现最好的梯度提升指标分别提升44%和19%,提升了预测效果,为未来分数线预测提供了新的方向。
Aiming at how to accurately predict the college admission score line and help college entrance examination students make more accurate voluntary filling decisions,a two-layer model based on the idea of Stacking ensemble is proposed.The model uses machine learning algorithms to expose the importance of features,fuses three single algorithms,and uses cross-checking and grid search methods for parameter optimization.Through experiments on the admission data of colleges and universities in Guizhou Province in 2018—2022,the experimental results show that the prediction effect of the two-layer fusion model is better than that of single models such as Support Vector Regression,Decision Tree,Random Forest and other ensemble models.The accuracy of the prediction error within 5 points exceeds 95%,and the average absolute error is less than 2.43,which is 44%and 19%higher than the best performing gradient improvement index in a single model,respectively.The research improves the prediction effect,and provides the new direction for future score line prediction.
作者
干霞
魏嘉银
卢友军
秦信芳
来小孟
GAN Xia;WEI Jiayin;LU Youjun;QIN Xinfang;LAI Xiaomeng(School of Data Science and Information Engineering,Guizhou Minzu University,Guiyang 550025,China)
出处
《智能计算机与应用》
2024年第3期116-122,共7页
Intelligent Computer and Applications
基金
贵州省省级科技计划项目资助(黔科合基础[2018]1082号、[2019]1159号)
贵州省教育厅自然科学研究项目(黔教技[2022]015号、黔教技[2022]047号)。
