一种基于黏液菌觅食机制的特征选择算法及其在文本情感识别中的应用  

Slime mold foraging inspired feature selection algorithm and its application in sentiment recognition

作　　者：徐济惠[1] 颜晨阳[1] Xu Jihui;Yan Chenyang(School of Information and Intelligent Engineering,Ningbo City College of Vocational Technology,Ningbo 315211,China)

机构地区：[1]宁波城市职业技术学院信息与智能工程学院,浙江宁波315211

出　　处：《南京理工大学学报》2021年第5期596-605,共10页Journal of Nanjing University of Science and Technology

基　　金：全国教育信息技术研究课题(166243001);浙江省教育厅科研课题(Y201327646);浙江省教育技术研究规划课题(JB117)。

摘　　要：黏液菌(Physarum polycephalum)由于其展现出的迷宫寻径、路径寻优甚至构建与人工设计媲美的复杂交通网络等特殊能力而备受关注。该文正是受启发于黏液菌构建复杂鲁棒网络的行为,提出了一种仿生特征选择算法Slime-FS。Slime-FS将特征选择转化成一类最优特征子图求解问题,同时模仿黏液菌觅食机制,结合粗糙集理论构建了一种策略来指导最优特征子图的搜索。算法被应用于文本情感识别问题中,在某慕课平台评论文本数据集上进行了测试,结果显示Slime-FS能有效地选择鉴别特征,去除冗余和无关特征,其表现要远远优于基准算法(不带选特征选择的SVC),也要优于若干结合了元启发搜索策略的混合算法(Sklearn-genetic、EWGA、MSPSO和ACO)。Slime mold(Physarum polycephalum)has received much attention since it demonstrates the abilities to solve mazes,find shortest length networks,and even construct transport networks that have similar efficiency to those designed by human engineers.This paper seeks inspiration from the study of the robust,complex network formation behavior of slime mold to introduce an advanced feature selection algorithm(Slime-FS).The algorithm converts the feature selection into an optimal subgraph problem and employs a slime mold foraging inspired strategy combined with rough set theory to guide subgraph search procedure.The experimental results on the real-world MOOCs review dataset show that Slime-FS can effectively get rid of redundant or irrelevant features as well as capture discriminative features.The Slime-FS performs better than either the baseline algorithm(SVC without feature selection)or certain meta-heuristic approaches(EWGA,MSPSO and ACO).

关 键 词：黏液菌 情感识别 特征选择 仿生算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]