MIWOA-LSSVM方法的构建及其在生物质炭模式分类中的应用  

Construction of MIWOA-LSSVM method and its application inbiochar pattern classification

作　　者：张烨峰 成忠 单胜道 ZHANG Yefeng;CHENG Zhong;SHAN Shengdao(School of Environment and Natural Resources,Zhejiang University of Science and Technology,Hangzhou 310023,Zhejiang,China;School of Biological and Chemical Engineering,Zhejiang University of Science and Technology,Hangzhou 310023,Zhejiang,China)

机构地区：[1]浙江科技大学环境与资源学院,杭州310023 [2]浙江科技大学生物与化学工程学院,杭州310023

出　　处：《浙江科技学院学报》2024年第3期228-238,共11页Journal of Zhejiang University of Science and Technology

基　　金：国家重点研发计划项目(2022YFE0196000);浙江省重点研发计划项目(2020C01017)。

摘　　要：【目的】最小二乘支持向量机(least square support vector machine,LSSVM)的性能受惩罚因子和核函数参数的影响较大,为了优化这些参数,提出一种基于多策略改进的鲸鱼优化算法(multi-strategy improved whale optimization algorithm,MIWOA)。【方法】首先,采用Logistic混沌初始化方法替代随机初始化,以提高种群的多样性,进而提高搜索效率;然后,引入了非线性收敛因子和动态惯性权重,以增强算法的全局搜索能力;最后,采用具有长尾分布的Lévy飞行策略,以跳出局部最优解,扩大搜索范围。【结果】将本研究所构建的MIWOA-LSSVM集成方法用于多类别生物质炭的模式分类,结果显示MIWOA算法在参数寻优上速度更快,仅迭代7次就能得到参数最优组合。随后,利用MIWOA算法优化的参数,结合LSSVM模型进行分类,成功将分类准确率提升至96.38%。【结论】本研究结果证明了MIWOA算法在参数寻优上的可行性和高效性,同时表明MIWOA-LSSVM集成方法在多类别模式识别中具有良好的应用前景,可为优化算法在参数寻优上提供一定的参考。[Objective]The performance of the least square support vector machine(LSSVM)is greatly affected by penalty factors and kernel function parameters.To optimize these parameters,a multi-strategy improved whale optimization algorithm(MIWOA)was proposed.[Method]First,the Logistic chaos initialization method was employed instead of random initialization to increase the diversity of the population and thereby improve search efficiency;then,nonlinear convergence factors and dynamic inertia weights were introduced to enhance the global search capability of the algorithm;finally,the Lévy flight strategy with long-tail distribution was applied to jump out of the local optimal solution and expand the search range.[Result]The MIWOA-LSSVM integration method constructed in this study was intended for pattern classification of multi-category biochar.The results show that the MIWOA algorithm is faster in parameter optimization,and the optimal combination of parameters can be obtained only after 7 iterations.Subsequently,the parameters optimized by the MIWOA algorithm are used and combined with the LSSVM model for classification,succeeding in raising the classification accuracy to 96.38%.[Conclusion]The results prove the feasibility and efficiency of the MIWOA algorithm in parameter optimization.At the same time,it is shown that the MIWOA-LSSVM integration method has sound application prospects in multi-category pattern recognition,which can provide a certain basis for the optimization algorithm in parameter optimization.

关 键 词：最小二乘支持向量机 鲸鱼优化算法 生物质炭 模式分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]