基于改进花朵授粉算法的极限学习机模型  

Extreme Learning Machine Model Based on Improved Flower Pollination Algorithm

作　　者：邵良杉 兰亭洋 李臣浩 SHAO Liangshan;LAN Tingyang;LI Chenhao(System Engineering Institute,Liaoning Technical University,Huludao,Liaoning 125105,China)

机构地区：[1]辽宁工程技术大学系统工程研究所

出　　处：《计算机工程》2019年第12期281-288,共8页Computer Engineering

基　　金：国家自然科学基金(71371091)

摘　　要：为提高瓦斯突出风险预测的准确率和效率,在极限学习机(ELM)模型的基础上构建预测模型ACFPA-ELM。采用核线性鉴别分析(KLDA)对瓦斯突出样本数据进行特征抽取,利用代价敏感思想修正ELM适应度函数,同时将Tent混沌搜索和自适应算子引入花朵授粉算法(FPA)中,优化ELM的初始输入权值和阈值,从而提高对瓦斯突出风险的预测能力。实验结果表明,相较于经典的SVM、BP和ELM单一预测模型以及改进的FPA-ELM和PSO-ELM复合预测模型,ACFPA-ELM模型在瓦斯突出风险预测的准确率、预测一致性以及运行效率方面均具有明显的优势。In order to improve the accuracy and efficiency of gas outburst risk prediction,this paper proposes a prediction model ACFPA-ELM based on Extreme Learning Machine(ELM).First,this paper adopts Kernel Linear Discriminant Analysis(KLDA)to extract the features of gas outburst sample data.Then,this paper utilizes the cost sensitive ideas to modify ELM fitness function.At the same time,the Tent chaotic search and adaptive operator are introduced into the Flower Pollination Algorithm(FPA)to optimize the initial input weight and threshold of the ELM,thus improving the prediction ability for gas outburst risk.Experimental results show that,compared with the classic SVM,BP and ELM single prediction models,as well as the improved FPA-ELM and PSO-ELM composite prediction models,the proposed model is superior in the accuracy,consistency and efficiency of gas outburst risk prediction.

关 键 词：瓦斯突出 花朵授粉算法 极限学习机 核线性鉴别分析 混沌映射 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]