低资源场景下基于DA优化FNN的学习者分类方法  

Optimized FNN via DA for learner classification in low-resource scenarios

作　　者：张雅雯 张丽萍 闫盛 曹亚如 ZHANG Ya-wen;ZHANG Li-ping;YAN Sheng;CAO Ya-ru(College of Computer Science and Technology,Inner Mongolia Normal University,Hohhot 010022,China)

机构地区：[1]内蒙古师范大学计算机科学技术学院,内蒙古呼和浩特010022

出　　处：《计算机工程与设计》2025年第3期895-902,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61462071);内蒙古自然科学基金项目(2023LHMS06009);内蒙古自治区教育科学研究“十四五”规划2023年度课题基金项目(2023NGHZX-ZH119、NGJGH2023234);内蒙古师范大学基本科研业务费专项基金项目(2022JBXC018);内蒙古师范大学研究生创新基金项目(CXJJS22137);内蒙古自治区直属高校基本科研业务费基金项目(CXJJS23067)。

摘　　要：针对现有方法在分类学习者时面临过拟合、缺乏可解释性等问题,提出一种低资源场景下基于蜻蜓算法(DA)优化模糊神经网络(FNN)的学习者分类方法。面向有限的学习者标注样本,构建基于模糊神经网络的学习者分类模型;针对该网络收敛速度慢、容易陷入局部最优等问题,采用蜻蜓算法对隶属度函数参数寻优;捕获输入与输出之间的模糊规则,对学习者分类过程和结果展开分析。实验结果表明,利用蜻蜓算法进行优化能够提高模糊神经网络的收敛速度,其过程及结果具有可解释性。Aiming at the problems of overfitting and lacking of interpretability faced by the existing methods in classifying lear-ners,a learner classification method based on fuzzy neural network(FNN) optimized by the dragonfly algorithm(DA) was proposed for low-resource scenarios.The model was constructed for the limited learner annotated samples.To address the problems of low convergence speed and easiness to fall into the local optimum,the dragonfly algorithm was used to optimize the parameters of the affiliation function.The fuzzy rules between the inputs and the outputs were captured to analyze the process and results of the learner classification.Experimental results show that the optimization using the dragonfly algorithm can improve the convergence speed of the fuzzy neural network,and its processes as well as results are interpretable.

关 键 词：学习者分类 模糊神经网络 蜻蜓算法 模糊规则 可解释性 学习者分析 低资源场景 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]