摘要
传感器漂移是影响气体传感检测系统稳定性的重要问题之一。本文提出了一种基于支持向量机的在线漂移补偿的分类模型设计方法,并在此基础上引入遗忘系数以应对漂移影响和保证数据集平衡状态,引入起学系数来避免因数据集不平衡而导致模型一直无法得到训练的极端情况出现。经实验验证,改进的分类模型能够延长传感器的可靠使用时间,并对短中期的分类效果有一定程度的提升,模型自训练过程无须人工参与,符合现实应用场景。本文提出的研究思路和方法对相关领域的研究有一定的参考意义。
Sensor drift is one of the important problems affecting the stability of gas sensing detection system.In this paper,a support vector machine(SVM)self-training classification model for online drift compensation is proposed,in which the forgetting factor is introduced to cope with the influence of drift and ensure the balance state of the data set,and the priming factor is introduced to avoid the extreme situation that the model cannot be trained due to the imbalance of the training set.The experimental results show that the improved classification model can extend the reliable service time of the sensor and improve the classification effect in the short and medium term to a certain extent.The self-training process of the model does not require manual participation,which is in line with the practical application scenario.The research ideas and methods put forward in this paper have certain reference significance to the research in related fields.
作者
董晓睿
商凯
杨建磊
崔健
DONG Xiaorui;SHANG Kai;YANG Jianlei;CUI Jian(Shengli College,China University of Petroleum,Dongying,Shandong 257000,China)
出处
《南昌大学学报（理科版）》
CAS
北大核心
2020年第4期397-401,共5页
Journal of Nanchang University(Natural Science)
基金
国家自然科学基金资助项目(61262047)。
关键词
传感器漂移
支持向量机
自训练
化学气体传感器
在线漂移补偿
sensor drift
support vector machine
self-training
chemical gas sensor
online drift compensation
