基于贝叶斯优化混合核极限学习机的真伪卷烟拉曼光谱预测研究

Cigarette Packaging Authentication Based on Bayesian Optimized Hybrid Kernel Extreme Learning Machine with Raman Spectra

为解决人工鉴别真伪卷烟存在的预测精度低和主观性强的问题,提出一种基于贝叶斯优化混合核极限学习机的真伪卷烟拉曼光谱鉴别方法。该方法通过采用混合核函数提高模型的学习能力和泛化性能,并采用贝叶斯算法对混合核函数的参数进行优化,使其不仅有良好的局部搜索能力,同时也加强了全局搜索能力。将该方法应用于某品牌的真伪卷烟预测,试验结果表明:该模型拥有更好的预测精度,为真伪卷烟拉曼光谱预测提供了一种新思路。

To solve the problems of low prediction accuracy and strong subjectivity in manual dis-crimination of cigarette authenticity,a raman spectrum discrimination method based on Bayesian op-timized hybrid kernel extreme learning machine(BO-HKELM)is proposed.The BO-HKELM uses mixed kernel function to improve the learning ability and generalization performance,and uses Bayesian algorithm to optimize parameters of the mixed kernel function,which not only has a good local search ability,but also strengthens the global search ability.Finally,the BO-HKELM is ap-plied to discriminate the authenticity of a certain brand of cigarette.The experimental results show that the BO-HKELM has better prediction accuracy and provides a new way to discriminate the ciga-rette authenticity with raman spectra.