智能机内测试技术中KNN算法的改进  

An Improved KNN Algorithm of Intelligent Built-in Test

作　　者：姬东朝[1] 宋笔锋[1] 韩斐[1] 

机构地区：[1]西北工业大学航空学院,西安710072

出　　处：《弹箭与制导学报》2008年第4期230-233,共4页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：教育部优秀博士论文作者专项基金(200051)资助

摘　　要：针对复杂装备机内测试技术中广泛采用的KNN算法的缺陷,提出一种改进的算法——IKNN。首先用神经网络技术进行输入属性权重的计算,使重要输入属性赋予较大的权值,提高算法的分类精度;其次将训练集中的每一类样本进行聚类,从而减少训练事例集的数据量,提高算法的分类速度。实例验证表明该算法有效地提高了KNN算法的分类精度和速度。Focused on the deficiency of KNN algorithm in complex equipment＇s buik-in test, an improved KNN （IKNN） algorithm was brought forward. Firstly, the weight of each input feature was learned using neural network to make important features contribute more in the classifications, this improved the precision of classification. Secondly, clustering each sample of the training set to reduce the data volume of training set, this improved the running speed of the algorithm, Experiments show the efficiency of the improved algorithm.

关 键 词：机内测试 故障论断 K-最近邻 聚类 

分 类 号：O159[理学—数学]