基于FOA-RBF神经网络的机械类实验课程目标达成度评价

Evaluation of Achievement Degree for Mechanical Experiment Curriculum Based on FOA-RBF Neural Network

在本科机械类专业教学体系中,面向解决复杂工程问题实验课程的课程目标达成度评价是工程教育认证中该类专业毕业要求、培养目标达成度评价的基石。针对重庆理工大学机械工程学院的机械类基础实验课程,构建了实验课程目标达成度评价体系,提出了一种基于果蝇优化算法(FOA)的径向基函数(RBF)神经网络评价模型,以学生实验项目成绩为输入,课程目标达成度为输出,通过实际学生样本对该评价模型及3种经典神经网络模型进行对比验证。结果表明,FOA-RBF评价模型的平均误差、均方差和最大相对误差均为最优,且其相对误差在49%以内,泛化能力强,对专业实验课程目标达成度预估效果理想,为专业工程教育认证工作提供了科学、量化的评估数据支撑。

The evaluation of objective achievement degree for the undergraduate mechanical experimental courses which are for solving complex engineering problem is the basis of assessment of achievement degree for graduation requirements and training objectives in the engineering education certification.An achievement degree evaluation system of the experimental courses is established for mechanical basic experimental courses of College of Mechanical Engineering in Chongqing University of Technology,an evaluation model based on fruit fly optimization RBF neural network is proposed.The model takes students’experimental course scores as input and objective achievement degree as output,is trained by actual student samples,and compared with three classical neural network models.The result shows that the mean error,RMSE and relative error of FOA-RBF evaluation model are the best and the relative error is within 4.9%,which means the generalization ability of the model is strong.The predicting effect of the result is ideal,and provides scientific and quantitative support for the professional engineering education certification.