基于MK-RVM学习方法的大学生科研能力预测模型  

MK-RVM-based prediction model of university students′ scientific research ability

作　　者：傅翠霞 罗亦泳[2] FU Cuixia;LUO Yiyong(Research and Evaluation Center for Higher Education,East China University of Technology,Nanchang 330013,China;School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,China)

机构地区：[1]东华理工大学高等教育研究与评估中心,江西南昌330013 [2]武汉大学测绘学院,湖北武汉430079

出　　处：《现代电子技术》2019年第23期163-167,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(41861058);江西省教育科学“十二五”规划课题(13YB126)~~

摘　　要：学生科研能力预测模型构建是大学生科研能力评价体系的关键问题,文中对机器学习新算法相关向量机进行改进,构建遗传算法优化的多核函数相关向量机(MK-RVM)学习算法,并基于该算法建立大学生科研能力预测新模型,分析结果的有效性及可靠性。由实例分析结果可知,MK-RVM对大学生科研能力评价指数(?)预测精度达到"好"级,对大学生科研能力等级η预测准确率达到100%,较大程度优于其他三种BP神经网络算法。MK-RVM建立的大学生科研能力预测模型具有很好的稀疏性,致使模型具有出色的计算效率,算法运行时间远小于其他三种BP神经网络。Establishment of the prediction model of students′ scientific research ability is the key in the evaluation system of University Students′ scientific research ability. In this paper,RVM is improved,and MK-RVM optimized by genetic algorithm is constructed. On the basis of this algorithm,a new prediction model of university students′ scientific research ability is established. The validity and reliability of the model are analyzed. The results of instance analysis show that the prediction accuracy of the evaluation index (?) of university students′ scientific research ability evaluated with MK-RVM reaches the "good" level,and the prediction accuracy of University Students′ scientific research ability level reaches 100%,which is better than the three BP neural network algorithms to a great extent. The MK-RVM-based prediction model of University Students′scientific research ability has good sparsity,which makes the model have excellent computational efficiency. The running time of the new algorithm is much less than the three kinds of BP neural networks.

关 键 词：预测模型 相关向量机 大学生 精度分析 科研能力评价 遗传算法 

分 类 号：TN911.1-34[电子电信—通信与信息系统]