基于KNN算法的财政预算监督方法  

Financial Budget Supervision Method Based on K-Nearest NeighborAlgorithm

作　　者：沈斌[1] 赵重远 SHEN Bin;ZHAO Zhongyuan(School of Electrical and Information Engineering,Wuhan Institute of Technology,Wuhan 430205,China)

机构地区：[1]武汉工程大学电气信息学院,湖北武汉430205

出　　处：《武汉工程大学学报》2020年第1期108-112,共5页Journal of Wuhan Institute of Technology

基　　金：国家留学基金委(201408420066);湖北省自然科学基金(2013CFA049)。

摘　　要：为了解决预算单位不按照预算绩效使用财政资金的问题,提出一种基于K最近邻分类算法(KNN)的财政预算监督方法。首先利用报文得到初始结果集,然后改进传统K最近邻分类(T-KNN)算法,弱化训练集的噪声数据并对其特征值加权,最后将训练集分层得到报文分类结果。改进的K最近邻分类算法(I-KNN)使报文分类检测的真正类率(TPR)与真负类率(TNR)分别达到了89.67%和88.42%,且分类时间较短。实验结果表明,本文提出的方法为报文分类应用于预算绩效考核中提供了新思路。To solve the problem that budget units do not use financial funds according to budget performance,we proposed a financial budget monitoring method based on K-nearest neighbor classification algorithm.Firstly,an initial result was obtained based on messages,then the traditional K-nearest neighbor classification algorithm was improved,in which the noise data of the training set were weakened and the eigenvalues were weighted.Finally,the training set was divided into multiple layers to obtain the message classification results.The true negative rate and true negative rate of our approach in the message classification detection task reach 89.67%and 88.42%,respectively.Apart from that,our technique is also time efficient.Experimental results show that the proposed method provides a new idea for the application of message classification in budget performance appraisal.

关 键 词：报文 KNN算法 预算绩效 特征值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]