机器学习技术与病媒生物种属鉴定  

The technology of machine learning and its application on identification of medical vectors

作　　者：裘炯良[1] 周力沛[1] 郑剑宁[1] 施惠祥[1] 李杰[1] 江滨 

机构地区：[1]宁波出入境检验检疫局,浙江宁波315012

出　　处：《中华卫生杀虫药械》2017年第5期436-439,共4页Chinese Journal of Hygienic Insecticides and Equipments

基　　金：国家质检总局科技计划项目(编号:2012B172)

摘　　要：目的探索应用机器学习技术开展病媒生物的种属鉴定并基于Python语言开发病媒生物机器鉴定系统。方法采用专家会商法提取宁波口岸常见蝇类的鉴别特征规则,构建特征与不同蝇种一一对应的训练样本数据集。应用k-近邻分类算法进行机器学习,并开发图形用户界面将整个机器学习运算及鉴定过程内嵌其中。结果抽提出复眼大小、颜色等7个特征,建立95×8维向量矩阵的训练样本数据集;开发病媒生物机器学习与鉴定系统,将80%的数据用于训练,20%的数据用于测试,正确率达到100%。由一名新手借助该信息系统对口岸新采集到的10只蝇进行种属鉴定,准确率达到90%。这些新的鉴定数据导入训练数据集再次进行自我学习、提升经验值。如此往复,逐步将该系统培育成长为病媒生物鉴定专家系统。结论以机器学习为特征的人工智能在病媒生物鉴定工作中的推广应用,将极大地提升工作效率,为我国的病媒生物防控工作奠定坚实基础。Objective To explore the technology of machine learning and its application in the field of medical vector identification,then to develop a machine identifying system of vectors（MISV） based on the Python language.Methods The method of experts consultation was employed for extracting the characteristics of 5 common fly species in Ningbo Port.The training dataset was constructed for machine learning using the k-Nearest Neighbor method.All of the calculation and identification was embedded within the GUI of MIS. Results The seven characteristics including the size and color of ommateum were extracted out,95 × 8 matrix was constructed as the training dataset. The correct rate attained 100 percent when 80 percent data was used for training and the remained for testing.90 percent of correct rate was also attained for the identification of ten newly collected flies.All the new data were imported for self-learning by the machine.Finally,the MIS could be cultivated as a genuine expert system.Conclusion With the popularization and application of the machine learning and artificial intelligence（AI）,it will improve the identification efficiency immensely and take a firm foundation for the prevention and control of medical vectors in our country.

关 键 词：病媒生物 鉴定 机器学习 人工智能 PYTHON 

分 类 号：R384[医药卫生—医学寄生虫学] R184.3[医药卫生—基础医学]