摘要
教学质量评价是教学管理中的重要工作,其关键是建立影响教学质量的众多指标与评价结果间的复杂非线性关系模型。BP神经网络虽能建立相关模型,但却在建模过程中未考虑由专家经验知识所积累的各评价指标与评价结果的不确定性分布信息,导致模型预测的评价结果准确性差且模型泛化性能弱。为此,本文选用具备刻画不确定性分布信息功能的最大熵准则替换BP算法中的均方误差准则,从而获得教学质量的最大熵神经网络评价模型。数据仿真和重庆理工大学50名教师评价实例均显示,改进模型的预测评价结果相对误差均在6%内,且明显优于传统BP神经网络模型。表明了改进模型的评价结果具有很高的可信性和较强的泛化能力,从而为实现教学质量的准确评价提供了一个可行方法。
Teaching quality evaluation is important work in teaching management, the core of which is how to model the complex nonlinear relationship between many evaluation indicators and the evaluation results. The accumu- lation of expert experience knowledge was not considered in modeling process by BP neural network, which included uncertainty distribution information of the evaluation indicators and the evaluation results, this might lead to the model prediction accuracy of the results of the evaluation is poor and the weak performance of the model generalization. In view of this, mean square error criterion of the convention BP neural network was replaced by maximum entropy crite- rion which depicted the uncertainty distribution information in the paper, then this maximum entropy neural network modeling was used for education quality evaluation. Data simulation and fifty teachers＇ evaluation results by the mod- el both show that the relative errors of the evaluation results are under 5% and significantly better than convention BP neural network model. It indicates that the evaluation results by the model have high credibility, and the model has good generalization performance. It provides a feasible method to accurate evaluation of teaching quality.
出处
《计算机仿真》
CSCD
北大核心
2013年第5期284-287,共4页
Computer Simulation
基金
重庆市教育委员会教育改革项目(1203053)
重庆理工大学研究生创新基金重点项目(YCX2011206)
关键词
最大熵准则
均方误差准则
神经网络
教学质量评价
Maximum entropy criterion
Mean square error criterion
Neural network （ NN ）
Teaching quality evaluation
