基于原型优化方法的分类器设计  

Classifier Design Based on Prototype Optimization Method

作　　者：柳新强 徐欢 王栋[1] LIU Xinqiang;XU Huan;WANG Dong(School of Surveying&Testing,Shaanxi Railway Institute,Weinan 714099,China;State Key Laboratory of Continental Dynamics,Northwest University,Xi’an 710069,China)

机构地区：[1]陕西铁路工程职业技术学院,测绘与检测学院,陕西渭南714099 [2]西北大学,大陆动力学国家重点实验室,陕西西安710069

出　　处：《微型电脑应用》2024年第8期1-3,共3页Microcomputer Applications

基　　金：国家自然科学基金项目(42072231);陕西铁路工程职业技术学院科研基金项目(KY2020-53)。

摘　　要：常规分类器如k近邻、支持向量机等已经被广泛使用,但在大数据时代背景下,较多的训练量会大幅度降低分类器的训练效率和准确率。为了解决该问题,利用原型优化方法对已有训练数据进行筛选压缩,滤除大量冗余数据,将压缩后的数据集作为原型来训练分类器,提高训练效率和分类准确率。在已有方法基础上做出改进,设计新的基于原型优化方法的分类器,可以大幅度减小训练量,并保证分类准确率,测试结果验证了所提方法的有效性。Conventional classifiers such as k-nearest neighbor and support vector machine have been widely used,but in the big data era,more training will greatly reduce the training efficiency and accuracy of the classifier.To solve this problem,the prototype optimization method is used to filter and compress the existing training data to filter out a large number of redundant data.The compressed data set is used as the prototype to train the classifier to improve the training efficiency and classification accuracy.The existing methods,designs a new classifier based on prototype optimization method,which can greatly reduce the training amount and ensure the classification accuracy.The test results verify the effectiveness of the proposed method.

关 键 词：原型优化 分类器 K近邻 支持向量机 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程] P237[自动化与计算机技术—控制科学与工程]