基于PCA-RBF网络的学生写作成绩预测模型  

Students' Writing Score Prediction Model Based on PCA-RBF Neural Network

作　　者：胡帅[1] 顾艳[1] 姜华[1] 

机构地区：[1]渤海大学大学外语教研部,辽宁锦州121013

出　　处：《计算机与现代化》2016年第1期69-72,126,共5页Computer and Modernization

基　　金：辽宁省教育厅科学研究一般项目(W2015015);辽宁省社会科学基金资助项目(L14CYY022)

摘　　要：为进一步提高学生英语写作成绩预测准确率,提出一种基于主成分分析(PCA)和径向基函数(RBF)神经网络相结合的写作成绩预测模型。先用主成分分析对所建立的学生写作评价体系作数据降维处理,提取前5个主成分,再将这些主成分作为RBF神经网络的输入,构建3层RBF神经网络预测模型。实验结果表明,与单一的RBF神经网络和BP神经网络相比,PCA-RBF预测模型的结构简单,收敛速度快,预测准确率高,泛化能力强,验证了本文提出模型的有效性。To improve the accuracy of students＇ writing score prediction, a prediction model based on principal component analy- sis （PCA） and radial basis function （RBF） was proposed. First the dimensions of an established assessment system of students＇ writings were reduced. The first five principal components were extracted and taken as inputs of the RBF neural network. Then a three-layered RBF network prediction model was created. The experiment results show that compared with a simple RBF neural network and a BP neural network, the PCA-RBF prediction model is of simpler structure, faster convergence speed, higher pre- diction accuracy and better generalization ability. The effectiveness of the proposed model is verified.

关 键 词：主成分分析 RBF神经网络 成绩预测 BP神经网络 

分 类 号：TP393.09[自动化与计算机技术—计算机应用技术]